Skip to content

Latest commit

 

History

History
3033 lines (2529 loc) · 333 KB

README.md

File metadata and controls

3033 lines (2529 loc) · 333 KB
Raw

This document analyses data of World Triathlon, formerly International Triathlon Union (ITU), to try to answer questions such as:

  • 🩱 How much faster do elite triathletes swim when wearing a neoprene wetsuit? [link]
  • 🐧 Should weaker swimmers rejoice when wetsuits are allowed? [link]
  • 👟 Are runs faster since carbon plates shoes? [link]
  • ⚖️ Is the wetsuit worth for 300m? [link]
  • 3️⃣ Does each sport (swim - bike - run) accounts for 1/3 of race durations? What about transitions? [link]
  • 👫 How much faster are men over women? In which sport is the gap larger/smaller? How has the gap evolved over the years? [link]
  • ⏳ How much faster are athletes in sprint over olympic distance? [link]
  • 📈 Does the level increase over years? [link]
  • 🚿 How much time does the wetsuit add to T1? [link]
  • 🎯 How often does an athlete win from a bike breakaway? [link]
  • 🏃 How often does the best runner win? [link]
  • 🚀 How often is the win decided with a sprint finish? [link]
  • 🌡️ Do water and air temperatures affect swimming and running performance? [link]
  • 🏳️ What are the most represented nations? Which nations have serious problems for their Olympics selection? [link]
  • 💪 What is the typical age of performing athletes and how has it evolved over years? [link]
  • ⏹️ How old are athletes when they stop racing elite short distance triathlon? [link]
  • 🏋️ What is the body mass index of performing triathletes? [link]
  • 🎂 Are two kids, born the same year but on two different months, equally likely to become professional triathletes? [link]

📚 DATA

Data are collected from the Triathlon.org API.

"The Triathlon.org API Platform allows access to the entire Triathlon.org infrastructure and data"

Race results are processed as followed:

  • Year: from 2009 (start of the world-series) to mid-2024, just after the Paris Olympics.
  • Event: only world-cups, world-series (called WTCS) and games related events (Commonwealth, Olympics and Olympic test events).
  • Distance: only sprint (750 - 20 - 5) and olympic (also called "standard": 1500 - 40 - 10) formats.
  • Minimum number of finishers: 25.
  • ⚠️ IMPORTANT: How to summarize all split-times of one race?
    • For each leg (swim, t1, bike, t2, run), an average of 5 times is computed.
    • Specifically, the 5-th to 9-th best times in each sport are used to compute this average.
    • This choice is arbitrary but, as explained below, relevant for my goal: "Analysing the general competitive field in each sport".
    • Other settings, e.g. top-1, top-3, top-10 and top-50, could also yield valuable insights, by adjusting a few parameters in the provided scripts.
    • Notably, the PACES and LEVEL OVER THE YEARS sections include top-3 and top-10 analyses as well. 🏅

Why consider split-times based on the ranking in each sport?

  • The split times of the finish-top-10 athletes can vary greatly depending on the race scenario:
    • For instance, a race might feature a strong breakaway group of swimmers reaching T2 with a significant lead, filling the top-10 positions.
    • In the same race, a large pack might arrive at T2 together, with the strongest runners then dominating the overall finish.
    • See this dedicated section for an analysis of race scenarios.
  • On the other hand, considering the ranking in each sport allows for more consistent and reliable comparisons across races.

Why consider the 5-th to 9-th best times?

  • To mitigate outliers:
    • Outstanding swimmers, riders, or runners might miss a race due to injury, scheduling conflicts, or other reasons, or they might have unusually good or bad performances.
    • Conversely, the 5-th to 9-th times are usually denser, providing a more robust representation of the general competitive field.
  • Against strategic variability among top performers:
    • The top-4 athletes might engage in strategic tactics on the run, such as testing each other or waiting for a final sprint, leading to varying performances.
    • Conversely, I believe athletes ranked 5-th to 9-th are more likely to give their all without strategic calculations, resulting in more robust and consistent comparative times across races.

What about the data quality?

  • World Triathlon puts efforts to uniform/standardize race reports and race timings.
  • But some manual cleaning is required.
    • Some information, such as the permission of wearing wetsuit, is often missing.
    • Also, I could not find any way to access the rankings of past years via the API.
  • Obviously, variations in distances, weather conditions, athlete levels and scenarios make comparisons between race difficult.
    • Fortunately, the number of races is large enough to smooth out these variations and provide interesting insights.
  • Comparing swim-times can be tricky, not just because the distances vary between events, but also because the positions of timing mats are not consistent.
    • Most mats are placed directly at the exit of the water, while others are located at the entrance of transition area, which can be hundreds of meters further.
    • Future results should be more consistent: World Triathlon is currently working on standardising placement of timing mats for all to be at swim exit as well as T-In.

🗂️ CONTENT

My favourite sections are marked with ⭐.

❓ THREE SPORTS

This section:

  • Investigates the statement: "swim, bike and run account each for one third of the total duration".
  • Analyses some transition data: T1 and T2.
sport_proportion.png
Proportion of swim, t1, bike, t2 and run in the overall race duration. For women/men and sprint/olympic formats.

Clearly not 1/3 + 1/3 + 1/3!

  • During a sprint format, women spend twice as much time running (run+t1+t2) as they do swimming.
  • Athletes spend more than three times longer on their bikes than in the water.
  • The proportion of the three sports remains similar between sprint and olympic formats.
    • The proportion of transition time is almost halved: transitions are fixed durations while the race time doubles from sprint to olympic.
  • Women spend less relative time on the swim but more on the run.

Why 1.5+40+10 as a format?

  • The format was allegedly introduced by the producers of the U.S. Triathlon Series (USTS) ( 🇺🇸 ) in the mid-1980s
  • According to this article by the Bass Lake triathlon:
  • "A need was seen to standardize the distances and make them more in sync with each individual sport. USTS is credited with inventing the distances of the modern day Olympic distance triathlon. For the swim, 1500 meters was chosen because it is the standard long distance competitive swimming event. For the bike, USTS chose 40 kilometers because it was the standard solo time trial distance in bike racing. And the choice for the run was the standard road distance of 10 km. Note that the distances were not chosen to be symmetrical nor were they in direct ratio to Ironman distances."

Drafting is allowed on bike.

  • Otherwise, gaps would probably be much larger, and probably the bike skills would become much more decisive.
  • I could not find when the drafting was first allowed on elite races.
    • In the first ITU world championship in 1989 in Avignon ( 🇫🇷 ), drafting seems to be banned.
    • Banning drafting seems nowadays complicated, considering the density of the swim level. Individual starts, e.g. every minute or so like during cycling time-trials, would be an option, and would give the bike section a much higher importance.

T1 and T2 represent a tiny part of the overall racing time, yet they are crucial!

  • E.g. to catch a good bunch at bike.

Comparing to other triathlon formats:

  • Gustav Iden won the IRONMAN world championships in 2022 with the following times.
  • "00:48:23" (3.8k), "4:11:06" (180k), "2:36:15" (42k), which represents:
    • 10.6% swim
    • 55.1% bike
    • 34.3% run
  • I.e. More run and less swim, which was expected from the distances: 3.8-180-42 compared to 1.5-40-10.
Click to expand - 🐇🐢 Some of the shortest and longest transitions.

(using Men data)

T1

The duration of T1 depends the distance between the water and the transition area, as well as on the position of the timing mats.

  • E.g. when the swim exit happens on a sand beach (Mooloolaba ( 🇦🇺 ) , Huatulco ( 🇲🇽 ), Hy-Vee ( 🇺🇸 )), timing mats are rarely placed directly on the sand, but instead further, close to the transition area.
  • Short T1 are often related to unusually long swim times: the best swim time 19:14 at Hy-Vee ( 🇺🇸 ) 2010, and 22 min at Mooloolaba ( 🇦🇺 ) 2012!
    • That is a shame: in such cases swim split times are incorrect!
    • Mooloolaba ( 🇦🇺 ) has apparently corrected the position of the timing mats: the fastest T1 in 2015 was 01:15 (against 00:13 in 2012).
T1 EVENT DISTANCE
00:13 2012 Mooloolaba ITU Triathlon World Cup ( 🇦🇺 ) OLYMPIC
00:16 2010 Huatulco ITU Triathlon World Cup ( 🇲🇽 ) OLYMPIC
00:16 2009 Hy-Vee ITU Triathlon Elite Cup ( 🇺🇸 ) OLYMPIC
... ... ...
01:26 2017 Madrid ITU Triathlon World Cup ( 🇪🇸 ) OLYMPIC
01:33 2017 ITU World Triathlon Grand Final Rotterdam ( 🇳🇱 ) OLYMPIC
01:40 2023 World Triathlon Championship Series Montreal ( 🇨🇦 ) SPRINT
02:35 2011 Guatape ITU Triathlon World Cup ( 🇨🇴 ) SPRINT

T2

The duration of T2 is mainly related to the size of the transition area (impacted by the number of participants) and to the position of the timing mats.

  • Variations are smaller than for T1.
T2 EVENT DISTANCE
00:14 2011 Mooloolaba ITU Triathlon World Cup ( 🇦🇺 ) OLYMPIC
00:14 2012 Mooloolaba ITU Triathlon World Cup ( 🇦🇺 ) OLYMPIC
00:15 2010 Monterrey ITU Triathlon World Cup ( 🇲🇽 ) OLYMPIC
... ... ...
00:35 2019 Banyoles ITU Triathlon World Cup ( 🇪🇸 ) SPRINT
00:36 2016 Montreal ITU Triathlon World Cup ( 🇨🇦 ) SPRINT
00:36 2013 Tiszaujvaros ITU Triathlon World Cup ( 🇭🇺 ) SPRINT
00:41 2017 Salinas ITU Triathlon World Cup ( 🇪🇨 ) SPRINT

T1+T2

On average, T1+T2 takes 01:11 (men) and 01:18 (women).

  • As mentioned for T1, very short times are mainly due to wrong positions of the timing mats after the swim: they are placed at the entrance of the transition area instead of at the water exit.
  • Longer T1+T2 means athletes must run more distance to reach their bikes / shoes.
    • In Montreal ( 🇨🇦 ) 2023, it was 02:13: more than one minute than usual.
    • For a sprint format, this is substantial: compared to the 15:05 average men's 5k, this long transition makes the run 7% longer!
T1+T2 EVENT DISTANCE
00:26 2011 Ishigaki ITU Triathlon World Cup ( 🇯🇵 ) OLYMPIC
00:27 2012 Mooloolaba ITU Triathlon World Cup ( 🇦🇺 ) OLYMPIC
00:28 2010 Tongyeong ITU Triathlon World Cup ( 🇰🇷 ) OLYMPIC
00:31 2010 Mooloolaba ITU Triathlon World Cup ( 🇦🇺 ) OLYMPIC
... ... ...
01:48 2014 ITU World Triathlon Stockholm ( 🇸🇪 ) SPRINT
01:56 2017 ITU World Triathlon Grand Final Rotterdam ( 🇳🇱 ) OLYMPIC
02:13 2023 World Triathlon Championship Series Montreal ( 🇨🇦 ) SPRINT
02:57 2011 Guatape ITU Triathlon World Cup ( 🇨🇴 ) SPRINT

⏱️ PACES

It is worth recalling the data settings:

  • Year: from 2009 to mid-2024.
  • Event: only world-cups, world-series (called WTCS) and games related events.
  • Distance: only sprint and olympic formats.
  • At least 25 finishers.
  • ⚠️ IMPORTANT: for each leg of a race (swim, bike, run), an average of 5 times is computed, using the 5-th to 9-th best times of the leg.
sports_paces.png
Distributions of times and paces.

Note: The distribution of swim times includes races both with- and without wetsuit. A subsequent section does the distinction (see its "second method" subsection).

Click to expand - 🏅 Same plots for the Top-3.
sports_paces_top3.png
Times and paces, considering the Top-3 in each leg.
Click to expand - 🏁 Same plots for the Top-10.
sports_paces_top10.png
Times and paces, considering the Top-10 in each leg.
Click to expand - ⏱️ Pace/speed/5k/10k conversion for the run.
Run Pace (M:SS) Speed (km/h) 5k Time (MM:SS) 10k Time (MM:SS)
2:55 20.6 14:35 29:10
2:56 20.5 14:40 29:20
2:57 20.3 14:45 29:30
2:58 20.2 14:50 29:40
2:59 20.1 14:55 29:50
3:00 20.0 15:00 30:00
3:01 19.9 15:05 30:10
3:02 19.8 15:10 30:20
3:03 19.7 15:15 30:30
3:04 19.6 15:20 30:40
3:05 19.5 15:25 30:50
3:06 19.4 15:30 31:00
3:07 19.3 15:35 31:10
3:08 19.1 15:40 31:20
3:09 19.0 15:45 31:30
3:10 18.9 15:50 31:40
3:11 18.8 15:55 31:50
3:12 18.8 16:00 32:00
3:13 18.7 16:05 32:10
3:14 18.6 16:10 32:20
3:15 18.5 16:15 32:30
3:16 18.4 16:20 32:40
3:17 18.3 16:25 32:50
3:18 18.2 16:30 33:00
3:19 18.1 16:35 33:10
3:20 18.0 16:40 33:20
3:21 17.9 16:45 33:30
3:22 17.8 16:50 33:40
3:23 17.7 16:55 33:50
3:24 17.6 17:00 34:00
3:25 17.6 17:05 34:10
3:26 17.5 17:10 34:20
3:27 17.4 17:15 34:30
3:28 17.3 17:20 34:40
3:29 17.2 17:25 34:50
3:30 17.1 17:30 35:00
3:31 17.1 17:35 35:10
3:32 17.0 17:40 35:20
3:33 16.9 17:45 35:30
3:34 16.8 17:50 35:40
3:35 16.7 17:55 35:50
3:36 16.7 18:00 36:00
3:37 16.6 18:05 36:10
3:38 16.5 18:10 36:20
3:39 16.4 18:15 36:30
3:40 16.4 18:20 36:40

Some outliers have been excluded:

Sprint vs olympic format:

  • 🏊 Swim paces are almost identical for 750m and 1500m: about 1s / 100m difference.
  • 🚴 There is less than 1km/h difference between the 20k and 40k bike.
  • 🏃 The 10k run requires 7s/km more than the 5k.

The next section analyses the time differences between women's and men's performance.

👫 WOMEN VS MEN

The difference in percent is computed with:

diff_in_percent = (time_w - time_m) / time_m

I.e.
time_w = (1 + diff_in_percent) * time_m

This percentage says "by how much are women slower than men".

  • To know "by how much are men faster", use diff_in_percent / (1 + diff_in_percent).
wm.png
By how much are women slower?

Notes about swim data:

wm_over_years.png
By how much are women slower? Evolution over years.
Click to expand - 📈 Evolution over years, considering only WTCS and games-related events.
wm_over_years_no_world_cup.png
By how much are women slower? Evolution over years. Only WTCS and games-related events.

💡 FINDINGS:

  • 🏊 The swim is the sport where the relative difference between women and men is the smallest.
    • Swimming is highly technique-oriented.
    • Women often excel in technical sports because these rely less on raw strength and more on skill, coordination, and efficiency.
    • Women and men have different buoyancy, as explained by Maria Francesca Piacentini, in this episode (at 19:00) of the triathlon show podcast.
    • As reported by this 2019 article by Romuald Lepers: "Elite female athletes generally have 7–12% more body fat than males (Fleck, 1983; Heydenreich et al., 2017). As fat is buoyant in water, women are less penalized than men in swimming than they are within terrestrial events such as cycling and running. Male triathletes also possess a larger muscle mass, greater muscular strength and lower relative body fat than female triathletes (Knechtle et al., 2010a)."
  • 🏃 The run is where the difference is the largest.
    • Men typically have greater muscle mass and aerobic capacity, which can provide an advantage in endurance activities like running.
  • 📏 The standard deviation is higher for swim and lower for run.
    • Because swim conditions (wind, current, temperature) can vary and athletes may follow non-straight swim lines leading to larger swim distances?
  • 📉 The w/m difference has not significantly reduced on the years, except for the run leg of the sprint-format races (-0.13 % / year).
    • Note: Probably some data processing should be applied to the line fitting.
      • E.g. to account for the very low number of points during the covid pandemic?
    • In WTCS and games-related events, the w/m swim gap has reduced (-0.11 % / year) as well.

These percentages can be compared with those of swim, bike or run competitions:

Click to expand - 🏊🚴🏃 Comparisons to individual swim/bike/run.

diff_percent = (time_w - time_m) / time_m is computed for a couple of events taken individually.

  • It would be statistically more significant to consider many events and compute averages.
  • I have not taken the time to do that.

🏊 SWIM

  • 🇯🇵 2021 Tokyo 800m: diff_percent = 7.7% considering the 4th to 8th men and women:
    • times_m = [7:42.68, 7:45.00, 7:45.11, 7:49.14, 7:53.31]
    • times_w = [8:19.38, 8:21.93, 8:22.25, 8:24.56, 8:26.30]
    • In previous olympic games, there was no men's 800m.
      • In Rio, the 4th-8th women average was less than 1s different to Tokyo.
  • 🇯🇵 2021 Tokyo 10k: diff_percent = 8.2% considering the 5th to 9th men and women:
    • times_m = [1:49:29, 1:50:23, 1:51:30, 1:51:32, 1:51:37]
    • times_w = [1:59:35, 1:59:36, 1:59:37, 2:00:10, 2:00:57]
    • Open-water conditions are closer to the one of triathlon, but the 10k race involves more strategy.

🚴 BIKE

  • Which race format?
    • Non-TT cycling race often involve strategy, making time comparison between women and men irrelevant.
    • ICU time-trial (TT) world championships uses different distances for women and men, making the comparison difficult.
    • Fortunately regional and national TT championships use the same distance and can therefore give relevant examples of diff_percent.
  • 🇺🇸 2024 USA TT: diff_percent = 11.1 % between Taylor Knibb (41:54) and Brandon McNulty (37:42)
  • 🇪🇺 2023 European TT: diff_percent = 12.3% considering the 5th to 9th men and women:
    • times_m = [32:43.77, 32:45.91, 32:52.03, 32:55.19, 32:55.22]
    • times_w = [36:42.01, 36:49.02, 36:53.75, 37:02.27, 37:05.13]
  • 🇪🇺 2022 European TT: diff_percent = 14.8% considering the 5th to 9th men and women:
    • times_m = [27:42.81, 28:01.56, 28:17.61, 28:18.47, 28:27.19]
    • times_w = [32:00.87, 32:01.10, 32:30.76, 32:32.76, 32:37.85]

🏃 RUN

  • diff_percent is computed in the same way.
    • But keep in mind that running competitions are often very strategic: the time matters often less than the ranking.
      • For instance, on the athletic track the pace can be kept low until the last lap.
    • Comparing world/continental/national records for women vs men is an option, but the level of one outstanding person says nothing about the average level.
  • 🇪🇺 2024 European 10k: diff_percent = 13.8% considering the 5th to 9th men and women:
    • times_m = [31:34.90, 31:38.45, 32:15.91, 32:16.85, 32:17.24]
    • times_w = [28:01.42, 28:04.43, 28:09.87, 28:10.97, 28:11.61]
  • 🇪🇺 2024 European 5k: diff_percent = 11.6% considering the 5th to 9th men and women:
    • times_m = [14:44.72, 14:58.28, 15:00.05, 15:02.56, 15:05.66]
    • times_w = [13:23.26, 13:24.54, 13:24.80, 13:25.08, 13:25.65]
  • 🇫🇷 2024 Paris marathon: diff_percent = 13.7% considering the 5th to 9th men and women:
    • times_m = [2:24:48, 2:26:00, 2:26:01, 2:26:03, 2:26:08]
    • times_w = [2:07:37, 2:07:39, 2:07:44, 2:08:41, 2:09:04]

Focus on the swim: 🏊

wm_swim.png
By how much swim women slower, with / without wetsuit? 🏊 🩱

💡 FINDING:

  • The women/men time difference can be considered constant, regardless of the distance (sprint / olympic) and the equipment (wetsuit or not): women swim ~8.8% slower.
  • This finding is consistent with Vleck et al., 2011.
  • This result will be used in a subsequent section to determine the benefit provided by the wetsuit.

🏊 SWIM GAPS

This section tries to answer the question "Should worse swimmers be happy when the wetsuit is allowed?".

  • My a priori reflexion was:
    • The wetsuit makes everyone swim faster.
    • => The swim takes less time.
    • => => Gaps are smaller.
    • => => => Worse swimmers are happy.

Approach:

  • For each race, the gap between the 5-9th first swimmers and 5-9th last swimmers is computed.
  • These gaps are split into two groups: with-wetsuit and without-wetsuit.
  • The averages of both group are compared to determine if the wetsuit makes swim gaps smaller or larger.
swim_gaps.png
Comparing average gaps between good and "bad" swimmers with and without wetsuit.

💡 FINDINGS:

  • Variations in the swim-pack length between event with-wetsuit and without-wetsuit are very small.
  • There is no evidence that worse swimmers should be happy about the wetsuit.
    • The wetsuit even tends to stretch the swim pack, especially for women.
    • To be honest, I would have expected the opposite!

POSSIBLE INTERPRETATION #1:

  • Wetsuits are typically worn in cold waters, often in seas and oceans, where waves can make swimming more challenging, potentially spreading out the pack.
  • However, there are many examples of sea and ocean swims that occur without wetsuits!
  • Therefore, I would dismiss this hypothesis.

POSSIBLE INTERPRETATION #2:

  • True, the swim is shorter in time.
  • But gaps do not significantly reduce because the benefit provided by the wetsuit differs between good and worse swimmers:
    • For 5-9th top swimmer: 5.4%.
    • For 20-24th top swimmer: 5.3%.
    • For last 9-5th top swimmer: 4.9%.
    • More details on this derivation can be found in the dedicated section.
  • In other words, despite the shorter swim duration, gaps do not reduce because top swimmers gain more benefits from the wetsuit.

QUESTION:

  • Would the wetsuit enable the slowest swimmers (last 9-5th) to catch the good swimmers (first 5-9th)?
    • On the olympic format, the gap is about 56s and 49s without wetsuit, while the fast women and men swim on average in 19:30 and 17:57.
    • A 4.9% improvement for the slowest swimmers represent 0.049 * (19:30 + 0:56) = 60s, and 0.049 * (17:57 + 0:49) = 55s.
    • Conclusion: the slowest swimmers with the benefit of the wetsuit would be ~10s faster that the good swimmers without.
Click to expand - Other comparisons.

Considering the "front-to-middle" distance (using the 20-24th swimming times instead of the last 5-9th), results looks similar: No significant gap reduction.

  • men+sprint+no_wetsuit may suffer from outliers: gaps at Tiszaújváros ( 🇭🇺 ) 2013 and 2016 were larger than 33s.
swim_gaps_20-24.png
Same computation as above, this time with gaps between 5-9th to 20-24th swimmers.

When considering world-series events only, the opposite trend occurs: the wetsuit tends to reduce the swim gaps.

  • However, as noted earlier, variations are very small, indicating that no significant effect of the wetsuit on swim gaps can be concluded.
swim_gaps-wcs.png
With world-series events only.

🐧 WETSUIT BENEFIT

This section tries to estimate the benefit (in percent) offered by the wetsuit, defined as

improve_percent = (time_no_wetsuit - time_wetsuit) / time_no_wetsuit

I.e.
time_with_wetsuit = (1 - improve_percent) * time_without_wetsuit

Main challenge:

  • Only one of [time_no_wetsuit, time_wetsuit] is typically available: the one recorded during the race.
  • This sections introduces different methods to estimate the missing time_, enabling the calculation of improve_percent.

Reminder:

  • The scope here is elite triathletes, going 5-9th out of water on top World Triathlon events.
  • Results would certainly differ for beginners and hobby triathletes.

👫 Method 1: using women/men differences

The idea of the derivation is as follows:

  • Women have been found to swim on average ~8.8% slower than men, with the same equipment.

  • With examples where women had the wetsuit, but men did not, one can:

    • 1- Estimate, from the men's time, the time women WOULD HAVE done without the wetsuit (thanks to the ~8.8% rule).
    • 2- Compare the women's time with wetsuit (measured) with the women's time without (computed in 1-).
    • 3- Deduce the advantage provided by the wetsuit for the women.
    • 4- Note that improve_percent should be the same for women and men (because of the constant ~8.8% difference given the same equipment).

  • Advantages of this method:

    • Proper environment: data comes from real races, as opposed to studies in 50m or even 25m pools.
    • No need to know the exact swim distance: what matters is that men and women swim the exact same course. This assumption is reasonable because the buoys should not move between the two races.
    • Data-based: the 8.8% should be robust since it leverages results from many races (230 events).
    • Low cost: all the data is available online for free.

  • Limitations:

    • The swim conditions are not guaranteed to be equal.
    • The limited number of events where the "women-with / men-without" scenario occurs.
      • Only five races fit this scenario, but the variability is low: std = 0.5%.
Click to expand - ✏️ Derivation of the formula. 
# VARIABLES
  swim_w:             time of women. no wetsuit.
  swim_m:             time of men. no wetsuit.
  swim_w_wet:         time of women. with wetsuit.
  
  improve_percent:    advantate (in percent) brought by the wetsuit. UNKNOWN.
  wm_percent:         relative delay of women over men assuming same equipment.
  wm_percent_w_fast:  relative delay when women have wetsuit, but men do not.

# FORMULA
  would W have not had wetsuit:
    (1)  `swim_w * (1 - improve_percent) = swim_w_wet`  =>  `swim_w = swim_w_wet / (1 - improve_percent)` 
  and this time would be related to swim_m:
    (2)  `swim_w = swim_m * (1 + wm_percent)`
  using (1) == (2):
    (3)  `swim_w_wet = swim_m * (1 + wm_percent) * (1 - improve_percent)`
  
  on the other hand:
    (4)  `swim_w_wet = swim_m * (1 + wm_percent_w_fast)`
  
  using (4) == (3):
    (5)  `(1 + wm_percent) * (1 - improve_percent) = (1 + wm_percent_w_fast)`
  re-written:
         `(1 - improve_percent) = (1 + wm_percent_w_fast) / (1 + wm_percent)`
  hence
         `improve_percent = 1 - (1 + wm_percent_w_fast) / (1 + wm_percent)`

# EXAMPLE
  [wm_percent = 8.8%]
  [wm_percent_w_fast = 2.9%]
  => improve_percent = 1 - (1+0.029)/(1+0.088) = 5.4%
wetsuit.png
Estimating the benefit brought by the wetsuit, using results of races where women swam with but men without.

💡 FINDING:

  • The wetsuit brings an advantage of ~5.4% to top swimmers (5th-9th).
  • Put differently, top swimmers (top 5-9) swim ~5.7% slower without wetsuit.
    • 0.054 / (1-0.054) = 0.057.

⚠️ CRITICISMS AND IDEAS FOR IMPROVEMENT:

  • 1) Uncertainty:
    • How to account for uncertainties in wm_percent_w_fast and wm_percent in the improve_percent = 1 - (1 + wm_percent_w_fast) / (1 + wm_percent) formula?
      • So far, the standard deviations were computed (± 3.0% and ± 0.5%), telling how spread out the observed w/m-swim-diff percentages are:
        • wm_percent = 8.8% ± 3.0%.
        • wm_percent_w_fast = 2.9% ± 0.5%.
        • Concretely, in the case of wm_percent, ± 3.0% can be interpreted as: "~68% of the observed w/m-swim-diff percentages lie between 8.8%-3% and 8.8%+3%".
      • I am not sure, but from what I understood, in order to produce a confidence interval for improve_percent, the standard errors (SE) should be used instead:
        • SE(wm_percent) = 3.0% / sqrt(230) = 0.2%.
        • SE(wm_percent_w_fast) = 0.5% / sqrt(5) = 0.2%.
    • In statistics, this question is known as Propagation of Uncertainty.
      • Approach #1 (simple): perform calculations using the extremes of the error intervals to see where improve_percent falls.
        • Here, applying the combinations (-0.2%, -0.2%), (-0.2%, +0.2%), (+0.2%, -0.2%) and (+0.2%, +0.2%) to (wm_percent = 8.8%, wm_percent_w_fast = 2.9%).
        • This results in the interval improve_percent = 5.4% with 0.4% standard error.
      • Approach #2 (using partial derivatives):
        • Using this tool, I obtain improve_percent = 5.4% with 0.3% standard error.
  • 2) Events consistency:
    • wm_percent_w_fast = 2.9% was computed from five "women-with-wetsuit, men-without" examples that all have the following properties: WTCS and olympic-format.
      • The five venues are: Yokohama ( 🇯🇵 ) (twice), Cagliari ( 🇮🇹 ), Stockholm ( 🇸🇪 ) and Edmonton ( 🇨🇦 ).
    • In contrast, wm_percent = 8.8% was obtained by considering all the sprint- and olympic-format WCTS, world-cups and games-related events since 2009, totaling 220 events.
      • This is inconsistent.
    • Instead, wm_percent should be estimated considering events with similar swim conditions as those for wm_percent_w_fast.
      • Idea #1: Only consider events with the same event-category (WTCS), and possibly the same format (olympic). For comparison:
        • ⚠️ WTCS-only: improve_percent=4.8%, with a lower women/men difference (8.1%).
        • ⚠️ World-cups-only: improve_percent=4.0%, with a higher women/men difference (9.4%).
      • Idea #2: Further restrict Idea #1 (same format and event-category), by considering only with the same venues.
        • The four venues mentioned have hosted multiple olympic-WTCS: 20 times, from which 15 had women and men sharing the same equipment for the swim.
        • This gives wm_percent = 7.4% ± 1.6% (SE = 1.6/sqrt(15) = 0.4%), leading to improve_percent = 4.2% with 0.3% standard error.
      • IMPORTANT CONCLUSION:
        • There is a trade-off between data-quantity and data-quality.
        • ➡️ First refinements suggest wetsuit improvements closer to 4-5%. ⬅️
  • 3) Additional examples:
    • Several other events feature the "women-with / men-without" scenario.
    • It would be valuable to include events with the opposite scenario: "women-without / men-with".
    • Future events may provide additional examples to further refine the estimate.
  • 4) Function approximation:
    • Currently, the relationship between swim_w and swim_m is modelled as linear: swim_w = swim_m * (1 + wm_percent).
      • Could this relationship be better captured using more sophisticated models, such as neural networks?
      • What additional inputs might enhance prediction accuracy? For example, should the model consider the complete list of swim times?
Click to expand - 🌍 Events used for the derivation.
YEAR EVENT DIFF (%) WOMEN-WITH vs MEN-without BENEFIT (%) DISTANCE EVENT CATEGORY
2024 Cagliari ( 🇮🇹 ) 2.3% 6.0% olympic WTCS
2021 Edmonton ( 🇨🇦 ) 2.4% 5.9% olympic WTCS
2022 Yokohama ( 🇯🇵 ) 2.9% 5.4% olympic WTCS
2015 Stockholm ( 🇸🇪 ) 3.3% 5.0% olympic WTCS
2024 Yokohama ( 🇯🇵 ) 3.5% 4.9% olympic WTCS

🌍 Method 2: Using Recurring Events

This method identifies pairs of events held at the same venue, where wetsuits were required in one year but not in another.

  • The approach assumes that the course layout remains consistent across years.
  • For example, in the men's race at Yokohama ( 🇯🇵 ), wetsuits were worn in 2023 (swim time: 17:31) but not in 2022 (swim time: 18:35) and not in 2021 (swim time: 18:23).
  • Using these three events, two comparisons can be made to estimate improve_percent:
    • improve_percent = (time_no_wetsuit - time_wetsuit) / time_no_wetsuit
    • 2023 (wetsuit) vs 2021 (no-wetsuit): improve_percent = (18:23 - 17:31) / 18:23 = 4.7%.
    • 2023 (wetsuit) vs 2022 (no-wetsuit): improve_percent = (18:35 - 17:31) / 18:35 = 5.7%.
  • By collecting many such comparisons, a distribution of improve_percent is generated, see the figure below.
  • The mean and median of this distribution provide an answer to the initial question: "How much faster are top swimmers with the wetsuit?"
wetsuit_in_swim_year_to_year.png
Distribution of the estimated improve_percent using the method of recurring events.

Results from 186 comparisons:

  • improve_percent:
    • Women+Men : mean = 3.5%, median = 3.6%. (std = 3.5)
    • Women only : mean = 3.6%, median = 3.6%. (std = 3.2)
    • Men only : mean = 3.4%, median = 3.3%. (std = 3.7)
  • Women vs Men: 0.2% (with means) or 0.3% (with medians).

💡 DISCUSSION:

  • Challenge: the swim course varies across years.
    • The year-to-year comparison incorporates a substantial amount of data (nearly 200 estimates) and assumes that while some samples may yield low estimates and others high, the large dataset size ensures these variations balance out.
      • This approach justifies including negative estimates, as the presence of very large benefits is also unlikely.
      • Together, they create a near-normal distribution, leading to robust overall estimates.
      • Just to check: When applied to events using the same swim equipment, the year-to-year comparison yields also a symmetric distribution centered around ~0%, as expected: the swim level does not change, while swim course lengths vary slightly year to year (sometimes longer, sometimes shorter), but remain constant on average.
    • Outliers, however, are addressed by removing estimates below -5% and above 14%.
    • To further limit the impact of outliers, comparisons are restricted to years in close proximity.
      • The choice of 5 years as the cutoff is another parameter of the analysis.
Click to expand - ⚖️ Full list of comparisons used for this derivation.

Distributions:

  • Event category:
    • WTCS : 73% (135)
    • World-Cup : 27% (51)
  • Distance category:
    • Olympic : 48% (90)
    • Sprint : 52% (96)

Venues of the comparisons:

  • 52 (28.0%): Hamburg ( 🇩🇪 )
  • 40 (21.5%): Yokohama ( 🇯🇵 )
  • 16 ( 8.6%): Edmonton ( 🇨🇦 )
  • 12 ( 6.5%): New Plymouth ( 🇳🇿 )
  • 12 ( 6.5%): London ( 🇬🇧 )
  • 11 ( 5.9%): Tongyeong ( 🇰🇷 )
  • 10 ( 5.4%): Cagliari ( 🇮🇹 )
  • 10 ( 5.4%): Karlovy Vary ( 🇨🇿 )
  • 7 ( 3.8%): Auckland ( 🇳🇿 )
  • 4 ( 2.2%): Cannigione, Arzachena ( 🇮🇹 )
  • 2 ( 1.1%): Valencia ( 🇪🇸 )
  • 2 ( 1.1%): San Diego ( 🇺🇸 )
  • 2 ( 1.1%): Stockholm ( 🇸🇪 )
  • 2 ( 1.1%): Montreal ( 🇨🇦 )
  • 2 ( 1.1%): Sydney ( 🇦🇺 )
  • 2 ( 1.1%): Chicago ( 🇺🇸 )
EVENT FORMAT GENDER SWIM WITH WETSUIT SWIM WITHOUT WETSUIT WETSUIT GAIN (%)
Cagliari ( 🇮🇹 ) SPRINT W 10:11 (01:21 /100m) (2019) 09:47 (01:18 /100m) (2017) -4.0%
Hamburg ( 🇩🇪 ) OLYMPIC M 17:26 (01:10 /100m) (2011) 16:46 (01:07 /100m) (2010) -4.0%
Tongyeong ( 🇰🇷 ) OLYMPIC M 17:08 (01:09 /100m) (2010) 16:33 (01:06 /100m) (2014) -3.5%
Cagliari ( 🇮🇹 ) OLYMPIC W 19:01 (01:16 /100m) (2024) 18:23 (01:14 /100m) (2023) -3.4%
Tongyeong ( 🇰🇷 ) OLYMPIC M 18:12 (01:13 /100m) (2009) 17:36 (01:10 /100m) (2011) -3.4%
Cagliari ( 🇮🇹 ) SPRINT M 09:27 (01:16 /100m) (2019) 09:13 (01:14 /100m) (2018) -2.5%
Auckland ( 🇳🇿 ) OLYMPIC M 18:54 (01:16 /100m) (2011) 18:28 (01:14 /100m) (2015) -2.3%
Cagliari ( 🇮🇹 ) SPRINT M 09:27 (01:16 /100m) (2019) 09:15 (01:14 /100m) (2017) -2.2%
Tongyeong ( 🇰🇷 ) SPRINT M 09:07 (01:13 /100m) (2018) 08:55 (01:11 /100m) (2016) -2.2%
Cannigione, Arzachena ( 🇮🇹 ) SPRINT W 09:28 (01:16 /100m) (2020) 09:17 (01:14 /100m) (2022) -2.0%
New Plymouth ( 🇳🇿 ) SPRINT M 09:54 (01:19 /100m) (2018) 09:42 (01:18 /100m) (2019) -2.0%
Yokohama ( 🇯🇵 ) OLYMPIC W 19:12 (01:17 /100m) (2015) 18:50 (01:15 /100m) (2019) -1.9%
Auckland ( 🇳🇿 ) OLYMPIC W 20:16 (01:21 /100m) (2011) 19:55 (01:20 /100m) (2014) -1.7%
Hamburg ( 🇩🇪 ) SPRINT M 08:59 (01:12 /100m) (2016) 08:50 (01:11 /100m) (2013) -1.7%
Yokohama ( 🇯🇵 ) OLYMPIC M 18:02 (01:12 /100m) (2015) 17:45 (01:11 /100m) (2019) -1.7%
Cagliari ( 🇮🇹 ) SPRINT W 10:11 (01:21 /100m) (2019) 10:02 (01:20 /100m) (2018) -1.4%
Hamburg ( 🇩🇪 ) OLYMPIC W 18:35 (01:14 /100m) (2011) 18:23 (01:14 /100m) (2010) -1.1%
Hamburg ( 🇩🇪 ) SPRINT W 09:43 (01:18 /100m) (2016) 09:36 (01:17 /100m) (2018) -1.1%
Auckland ( 🇳🇿 ) OLYMPIC W 20:16 (01:21 /100m) (2011) 20:06 (01:20 /100m) (2015) -0.8%
Hamburg ( 🇩🇪 ) SPRINT M 08:56 (01:11 /100m) (2019) 08:52 (01:11 /100m) (2022) -0.8%
Yokohama ( 🇯🇵 ) OLYMPIC W 18:58 (01:16 /100m) (2014) 18:50 (01:15 /100m) (2019) -0.7%
Cagliari ( 🇮🇹 ) SPRINT W 09:51 (01:19 /100m) (2016) 09:47 (01:18 /100m) (2017) -0.6%
Hamburg ( 🇩🇪 ) SPRINT M 08:59 (01:12 /100m) (2016) 08:56 (01:11 /100m) (2014) -0.4%
Yokohama ( 🇯🇵 ) OLYMPIC W 18:55 (01:16 /100m) (2022) 18:50 (01:15 /100m) (2019) -0.4%
Tongyeong ( 🇰🇷 ) OLYMPIC W 18:50 (01:15 /100m) (2010) 18:46 (01:15 /100m) (2015) -0.3%
Yokohama ( 🇯🇵 ) OLYMPIC M 18:02 (01:12 /100m) (2015) 17:59 (01:12 /100m) (2012) -0.3%
Valencia ( 🇪🇸 ) SPRINT W 09:26 (01:15 /100m) (2020) 09:25 (01:15 /100m) (2022) -0.2%
Hamburg ( 🇩🇪 ) SPRINT M 08:59 (01:12 /100m) (2016) 08:58 (01:12 /100m) (2018) -0.1%
Yokohama ( 🇯🇵 ) OLYMPIC W 18:51 (01:15 /100m) (2018) 18:50 (01:15 /100m) (2019) -0.0%
Hamburg ( 🇩🇪 ) SPRINT M 08:50 (01:11 /100m) (2017) 08:50 (01:11 /100m) (2013) 0.0%
Yokohama ( 🇯🇵 ) OLYMPIC W 19:12 (01:17 /100m) (2015) 19:13 (01:17 /100m) (2017) 0.0%
Cagliari ( 🇮🇹 ) SPRINT M 09:12 (01:14 /100m) (2016) 09:13 (01:14 /100m) (2018) 0.1%
Hamburg ( 🇩🇪 ) SPRINT M 08:56 (01:11 /100m) (2019) 08:56 (01:11 /100m) (2014) 0.1%
Tongyeong ( 🇰🇷 ) OLYMPIC M 17:08 (01:09 /100m) (2010) 17:10 (01:09 /100m) (2015) 0.2%
Edmonton ( 🇨🇦 ) SPRINT M 08:39 (01:09 /100m) (2016) 08:40 (01:09 /100m) (2013) 0.2%
Valencia ( 🇪🇸 ) SPRINT M 08:35 (01:09 /100m) (2020) 08:37 (01:09 /100m) (2022) 0.3%
Hamburg ( 🇩🇪 ) SPRINT M 08:56 (01:11 /100m) (2019) 08:58 (01:12 /100m) (2018) 0.4%
Hamburg ( 🇩🇪 ) SPRINT W 09:43 (01:18 /100m) (2016) 09:45 (01:18 /100m) (2013) 0.4%
Hamburg ( 🇩🇪 ) SPRINT M 08:50 (01:11 /100m) (2017) 08:52 (01:11 /100m) (2022) 0.4%
Cagliari ( 🇮🇹 ) SPRINT M 09:12 (01:14 /100m) (2016) 09:15 (01:14 /100m) (2017) 0.4%
Hamburg ( 🇩🇪 ) SPRINT W 09:43 (01:18 /100m) (2016) 09:45 (01:18 /100m) (2014) 0.5%
New Plymouth ( 🇳🇿 ) SPRINT M 09:39 (01:17 /100m) (2016) 09:42 (01:18 /100m) (2019) 0.5%
Edmonton ( 🇨🇦 ) SPRINT M 08:39 (01:09 /100m) (2016) 08:43 (01:10 /100m) (2017) 0.7%
Yokohama ( 🇯🇵 ) OLYMPIC W 18:42 (01:15 /100m) (2023) 18:50 (01:15 /100m) (2019) 0.7%
Edmonton ( 🇨🇦 ) SPRINT W 09:09 (01:13 /100m) (2015) 09:14 (01:14 /100m) (2013) 0.8%
Edmonton ( 🇨🇦 ) SPRINT W 09:08 (01:13 /100m) (2016) 09:14 (01:14 /100m) (2013) 1.0%
Yokohama ( 🇯🇵 ) OLYMPIC M 17:33 (01:10 /100m) (2014) 17:45 (01:11 /100m) (2019) 1.1%
Edmonton ( 🇨🇦 ) SPRINT M 08:39 (01:09 /100m) (2016) 08:46 (01:10 /100m) (2019) 1.3%
Hamburg ( 🇩🇪 ) SPRINT M 08:50 (01:11 /100m) (2017) 08:56 (01:11 /100m) (2014) 1.3%
Yokohama ( 🇯🇵 ) OLYMPIC M 17:31 (01:10 /100m) (2018) 17:45 (01:11 /100m) (2019) 1.3%
Yokohama ( 🇯🇵 ) OLYMPIC W 18:58 (01:16 /100m) (2014) 19:13 (01:17 /100m) (2017) 1.3%
Edmonton ( 🇨🇦 ) SPRINT M 08:33 (01:08 /100m) (2015) 08:40 (01:09 /100m) (2013) 1.4%
Yokohama ( 🇯🇵 ) OLYMPIC W 18:55 (01:16 /100m) (2022) 19:13 (01:17 /100m) (2017) 1.5%
Hamburg ( 🇩🇪 ) SPRINT M 08:50 (01:11 /100m) (2017) 08:58 (01:12 /100m) (2018) 1.6%
Hamburg ( 🇩🇪 ) SPRINT M 08:56 (01:11 /100m) (2019) 09:05 (01:13 /100m) (2024) 1.6%
Cannigione, Arzachena ( 🇮🇹 ) SPRINT W 09:08 (01:13 /100m) (2021) 09:17 (01:14 /100m) (2022) 1.6%
Hamburg ( 🇩🇪 ) SPRINT W 09:43 (01:18 /100m) (2016) 09:53 (01:19 /100m) (2015) 1.7%
Auckland ( 🇳🇿 ) OLYMPIC M 17:37 (01:10 /100m) (2012) 17:57 (01:12 /100m) (2014) 1.8%
Hamburg ( 🇩🇪 ) SPRINT M 08:40 (01:09 /100m) (2012) 08:50 (01:11 /100m) (2013) 1.8%
Edmonton ( 🇨🇦 ) SPRINT M 08:33 (01:08 /100m) (2015) 08:43 (01:10 /100m) (2017) 1.9%
Hamburg ( 🇩🇪 ) OLYMPIC M 16:27 (01:06 /100m) (2009) 16:46 (01:07 /100m) (2010) 1.9%
Cagliari ( 🇮🇹 ) SPRINT W 09:51 (01:19 /100m) (2016) 10:02 (01:20 /100m) (2018) 1.9%
Yokohama ( 🇯🇵 ) OLYMPIC W 18:51 (01:15 /100m) (2018) 19:13 (01:17 /100m) (2017) 1.9%
Hamburg ( 🇩🇪 ) SPRINT M 08:59 (01:12 /100m) (2016) 09:10 (01:13 /100m) (2015) 2.0%
Hamburg ( 🇩🇪 ) SPRINT W 09:25 (01:15 /100m) (2019) 09:36 (01:17 /100m) (2018) 2.0%
Tongyeong ( 🇰🇷 ) OLYMPIC W 18:22 (01:13 /100m) (2011) 18:46 (01:15 /100m) (2015) 2.1%
Hamburg ( 🇩🇪 ) SPRINT W 09:24 (01:15 /100m) (2017) 09:36 (01:17 /100m) (2018) 2.2%
Yokohama ( 🇯🇵 ) OLYMPIC M 18:02 (01:12 /100m) (2015) 18:27 (01:14 /100m) (2017) 2.2%
Edmonton ( 🇨🇦 ) SPRINT M 08:33 (01:08 /100m) (2015) 08:46 (01:10 /100m) (2019) 2.4%
Yokohama ( 🇯🇵 ) OLYMPIC M 17:33 (01:10 /100m) (2014) 17:59 (01:12 /100m) (2012) 2.4%
Hamburg ( 🇩🇪 ) SPRINT M 08:56 (01:11 /100m) (2019) 09:10 (01:13 /100m) (2015) 2.5%
Edmonton ( 🇨🇦 ) SPRINT W 09:09 (01:13 /100m) (2015) 09:24 (01:15 /100m) (2017) 2.5%
Hamburg ( 🇩🇪 ) SPRINT W 09:22 (01:15 /100m) (2021) 09:36 (01:17 /100m) (2018) 2.5%
Yokohama ( 🇯🇵 ) OLYMPIC M 17:18 (01:09 /100m) (2023) 17:45 (01:11 /100m) (2019) 2.5%
Yokohama ( 🇯🇵 ) OLYMPIC W 18:22 (01:13 /100m) (2024) 18:50 (01:15 /100m) (2019) 2.5%
Yokohama ( 🇯🇵 ) OLYMPIC M 17:18 (01:09 /100m) (2023) 17:45 (01:11 /100m) (2024) 2.6%
Hamburg ( 🇩🇪 ) SPRINT W 09:29 (01:16 /100m) (2012) 09:45 (01:18 /100m) (2013) 2.7%
Tongyeong ( 🇰🇷 ) OLYMPIC M 17:08 (01:09 /100m) (2010) 17:36 (01:10 /100m) (2011) 2.7%
Hamburg ( 🇩🇪 ) SPRINT M 08:38 (01:09 /100m) (2021) 08:52 (01:11 /100m) (2022) 2.7%
Edmonton ( 🇨🇦 ) SPRINT W 09:08 (01:13 /100m) (2016) 09:24 (01:15 /100m) (2017) 2.7%
Hamburg ( 🇩🇪 ) SPRINT W 09:29 (01:16 /100m) (2012) 09:45 (01:18 /100m) (2014) 2.7%
Yokohama ( 🇯🇵 ) OLYMPIC W 19:12 (01:17 /100m) (2015) 19:45 (01:19 /100m) (2012) 2.8%
London ( 🇬🇧 ) OLYMPIC W 18:52 (01:15 /100m) (2012) 19:25 (01:18 /100m) (2011) 2.8%
London ( 🇬🇧 ) OLYMPIC W 18:51 (01:15 /100m) (2013) 19:25 (01:18 /100m) (2011) 2.9%
Hamburg ( 🇩🇪 ) SPRINT W 09:25 (01:15 /100m) (2019) 09:42 (01:18 /100m) (2022) 3.0%
Hamburg ( 🇩🇪 ) SPRINT M 08:40 (01:09 /100m) (2012) 08:56 (01:11 /100m) (2014) 3.1%
Hamburg ( 🇩🇪 ) SPRINT W 09:24 (01:15 /100m) (2017) 09:42 (01:18 /100m) (2022) 3.2%
London ( 🇬🇧 ) OLYMPIC W 18:52 (01:15 /100m) (2012) 19:30 (01:18 /100m) (2010) 3.2%
London ( 🇬🇧 ) OLYMPIC W 18:51 (01:15 /100m) (2013) 19:30 (01:18 /100m) (2010) 3.3%
Cannigione, Arzachena ( 🇮🇹 ) SPRINT M 08:34 (01:09 /100m) (2021) 08:52 (01:11 /100m) (2022) 3.5%
Hamburg ( 🇩🇪 ) SPRINT W 09:22 (01:15 /100m) (2021) 09:42 (01:18 /100m) (2022) 3.5%
Hamburg ( 🇩🇪 ) SPRINT W 09:25 (01:15 /100m) (2019) 09:45 (01:18 /100m) (2014) 3.6%
Cagliari ( 🇮🇹 ) OLYMPIC W 19:01 (01:16 /100m) (2024) 19:43 (01:19 /100m) (2022) 3.6%
San Diego ( 🇺🇸 ) OLYMPIC W 17:54 (01:12 /100m) (2013) 18:34 (01:14 /100m) (2012) 3.6%
Hamburg ( 🇩🇪 ) SPRINT W 09:24 (01:15 /100m) (2017) 09:45 (01:18 /100m) (2013) 3.6%
Yokohama ( 🇯🇵 ) OLYMPIC M 17:31 (01:10 /100m) (2018) 18:11 (01:13 /100m) (2021) 3.6%
Yokohama ( 🇯🇵 ) OLYMPIC W 18:55 (01:16 /100m) (2022) 19:38 (01:19 /100m) (2021) 3.6%
Hamburg ( 🇩🇪 ) SPRINT M 08:50 (01:11 /100m) (2017) 09:10 (01:13 /100m) (2015) 3.6%
Hamburg ( 🇩🇪 ) SPRINT W 09:24 (01:15 /100m) (2017) 09:45 (01:18 /100m) (2014) 3.7%
Hamburg ( 🇩🇪 ) SPRINT M 08:38 (01:09 /100m) (2021) 08:58 (01:12 /100m) (2018) 3.8%
Auckland ( 🇳🇿 ) OLYMPIC W 19:09 (01:17 /100m) (2012) 19:55 (01:20 /100m) (2014) 3.8%
Karlovy Vary ( 🇨🇿 ) OLYMPIC W 19:05 (01:16 /100m) (2021) 19:51 (01:19 /100m) (2019) 3.8%
New Plymouth ( 🇳🇿 ) SPRINT W 10:32 (01:24 /100m) (2016) 10:57 (01:28 /100m) (2019) 3.9%
Hamburg ( 🇩🇪 ) SPRINT W 09:29 (01:16 /100m) (2012) 09:53 (01:19 /100m) (2015) 3.9%
Yokohama ( 🇯🇵 ) OLYMPIC M 18:02 (01:12 /100m) (2015) 18:47 (01:15 /100m) (2011) 4.0%
Yokohama ( 🇯🇵 ) OLYMPIC W 18:58 (01:16 /100m) (2014) 19:45 (01:19 /100m) (2012) 4.0%
Yokohama ( 🇯🇵 ) OLYMPIC W 18:51 (01:15 /100m) (2018) 19:38 (01:19 /100m) (2021) 4.0%
Hamburg ( 🇩🇪 ) OLYMPIC W 17:38 (01:11 /100m) (2009) 18:23 (01:14 /100m) (2010) 4.0%
Hamburg ( 🇩🇪 ) SPRINT M 08:30 (01:08 /100m) (2020) 08:52 (01:11 /100m) (2022) 4.2%
Stockholm ( 🇸🇪 ) OLYMPIC M 18:42 (01:15 /100m) (2013) 19:32 (01:18 /100m) (2015) 4.2%
Edmonton ( 🇨🇦 ) SPRINT M 08:39 (01:09 /100m) (2016) 09:02 (01:12 /100m) (2018) 4.2%
London ( 🇬🇧 ) OLYMPIC W 18:52 (01:15 /100m) (2012) 19:42 (01:19 /100m) (2009) 4.3%
London ( 🇬🇧 ) OLYMPIC W 18:51 (01:15 /100m) (2013) 19:42 (01:19 /100m) (2009) 4.3%
Yokohama ( 🇯🇵 ) OLYMPIC W 19:12 (01:17 /100m) (2015) 20:05 (01:20 /100m) (2011) 4.4%
London ( 🇬🇧 ) OLYMPIC M 17:15 (01:09 /100m) (2013) 18:02 (01:12 /100m) (2010) 4.4%
Edmonton ( 🇨🇦 ) SPRINT W 09:09 (01:13 /100m) (2015) 09:35 (01:17 /100m) (2019) 4.5%
Tongyeong ( 🇰🇷 ) SPRINT M 08:31 (01:08 /100m) (2021) 08:55 (01:11 /100m) (2016) 4.6%
Auckland ( 🇳🇿 ) OLYMPIC M 17:37 (01:10 /100m) (2012) 18:28 (01:14 /100m) (2015) 4.6%
London ( 🇬🇧 ) OLYMPIC M 17:12 (01:09 /100m) (2012) 18:02 (01:12 /100m) (2010) 4.7%
Auckland ( 🇳🇿 ) OLYMPIC W 19:09 (01:17 /100m) (2012) 20:06 (01:20 /100m) (2015) 4.7%
Edmonton ( 🇨🇦 ) SPRINT W 09:08 (01:13 /100m) (2016) 09:35 (01:17 /100m) (2019) 4.7%
Yokohama ( 🇯🇵 ) OLYMPIC M 17:31 (01:10 /100m) (2018) 18:23 (01:14 /100m) (2022) 4.7%
Yokohama ( 🇯🇵 ) OLYMPIC W 18:42 (01:15 /100m) (2023) 19:38 (01:19 /100m) (2021) 4.7%
Hamburg ( 🇩🇪 ) SPRINT W 09:25 (01:15 /100m) (2019) 09:53 (01:19 /100m) (2015) 4.8%
Yokohama ( 🇯🇵 ) OLYMPIC M 17:18 (01:09 /100m) (2023) 18:11 (01:13 /100m) (2021) 4.9%
Hamburg ( 🇩🇪 ) SPRINT W 09:24 (01:15 /100m) (2017) 09:53 (01:19 /100m) (2015) 4.9%
Yokohama ( 🇯🇵 ) OLYMPIC M 17:33 (01:10 /100m) (2014) 18:27 (01:14 /100m) (2017) 4.9%
London ( 🇬🇧 ) OLYMPIC M 17:15 (01:09 /100m) (2013) 18:08 (01:13 /100m) (2009) 4.9%
Hamburg ( 🇩🇪 ) SPRINT M 08:38 (01:09 /100m) (2021) 09:05 (01:13 /100m) (2024) 5.0%
Hamburg ( 🇩🇪 ) SPRINT W 09:07 (01:13 /100m) (2020) 09:36 (01:17 /100m) (2018) 5.0%
London ( 🇬🇧 ) OLYMPIC M 17:15 (01:09 /100m) (2013) 18:10 (01:13 /100m) (2011) 5.0%
Yokohama ( 🇯🇵 ) OLYMPIC M 17:31 (01:10 /100m) (2018) 18:27 (01:14 /100m) (2017) 5.1%
London ( 🇬🇧 ) OLYMPIC M 17:12 (01:09 /100m) (2012) 18:08 (01:13 /100m) (2009) 5.2%
Karlovy Vary ( 🇨🇿 ) OLYMPIC M 17:29 (01:10 /100m) (2020) 18:27 (01:14 /100m) (2019) 5.2%
Hamburg ( 🇩🇪 ) SPRINT M 08:30 (01:08 /100m) (2020) 08:58 (01:12 /100m) (2018) 5.3%
London ( 🇬🇧 ) OLYMPIC M 17:12 (01:09 /100m) (2012) 18:10 (01:13 /100m) (2011) 5.3%
Hamburg ( 🇩🇪 ) SPRINT M 08:40 (01:09 /100m) (2012) 09:10 (01:13 /100m) (2015) 5.4%
Edmonton ( 🇨🇦 ) SPRINT M 08:33 (01:08 /100m) (2015) 09:02 (01:12 /100m) (2018) 5.4%
New Plymouth ( 🇳🇿 ) SPRINT M 09:11 (01:13 /100m) (2023) 09:42 (01:18 /100m) (2019) 5.4%
Montreal ( 🇨🇦 ) SPRINT W 08:55 (01:11 /100m) (2023) 09:26 (01:15 /100m) (2019) 5.4%
Karlovy Vary ( 🇨🇿 ) OLYMPIC W 18:45 (01:15 /100m) (2020) 19:51 (01:19 /100m) (2019) 5.5%
New Plymouth ( 🇳🇿 ) SPRINT W 10:32 (01:24 /100m) (2016) 11:09 (01:29 /100m) (2017) 5.5%
Yokohama ( 🇯🇵 ) OLYMPIC W 18:58 (01:16 /100m) (2014) 20:05 (01:20 /100m) (2011) 5.6%
Cannigione, Arzachena ( 🇮🇹 ) SPRINT M 08:34 (01:09 /100m) (2021) 09:04 (01:13 /100m) (2020) 5.6%
Yokohama ( 🇯🇵 ) OLYMPIC W 19:12 (01:17 /100m) (2015) 20:23 (01:22 /100m) (2016) 5.7%
Hamburg ( 🇩🇪 ) SPRINT W 09:25 (01:15 /100m) (2019) 09:59 (01:20 /100m) (2024) 5.8%
Edmonton ( 🇨🇦 ) SPRINT W 09:09 (01:13 /100m) (2015) 09:44 (01:18 /100m) (2018) 5.9%
Yokohama ( 🇯🇵 ) OLYMPIC M 17:18 (01:09 /100m) (2023) 18:23 (01:14 /100m) (2022) 5.9%
Hamburg ( 🇩🇪 ) SPRINT W 09:07 (01:13 /100m) (2020) 09:42 (01:18 /100m) (2022) 6.0%
Edmonton ( 🇨🇦 ) SPRINT W 09:08 (01:13 /100m) (2016) 09:44 (01:18 /100m) (2018) 6.1%
New Plymouth ( 🇳🇿 ) SPRINT W 10:17 (01:22 /100m) (2018) 10:57 (01:28 /100m) (2019) 6.1%
Hamburg ( 🇩🇪 ) SPRINT W 09:22 (01:15 /100m) (2021) 09:59 (01:20 /100m) (2024) 6.3%
Karlovy Vary ( 🇨🇿 ) OLYMPIC W 18:35 (01:14 /100m) (2018) 19:51 (01:19 /100m) (2019) 6.4%
Hamburg ( 🇩🇪 ) SPRINT M 08:30 (01:08 /100m) (2020) 09:05 (01:13 /100m) (2024) 6.4%
Yokohama ( 🇯🇵 ) OLYMPIC W 18:22 (01:13 /100m) (2024) 19:38 (01:19 /100m) (2021) 6.5%
Yokohama ( 🇯🇵 ) OLYMPIC M 18:02 (01:12 /100m) (2015) 19:18 (01:17 /100m) (2016) 6.5%
Yokohama ( 🇯🇵 ) OLYMPIC M 17:33 (01:10 /100m) (2014) 18:47 (01:15 /100m) (2011) 6.6%
Yokohama ( 🇯🇵 ) OLYMPIC W 18:58 (01:16 /100m) (2014) 20:23 (01:22 /100m) (2016) 6.9%
Tongyeong ( 🇰🇷 ) OLYMPIC W 17:26 (01:10 /100m) (2014) 18:46 (01:15 /100m) (2015) 7.2%
Hamburg ( 🇩🇪 ) SPRINT M 08:30 (01:08 /100m) (2020) 09:10 (01:13 /100m) (2015) 7.3%
Yokohama ( 🇯🇵 ) OLYMPIC W 18:51 (01:15 /100m) (2018) 20:23 (01:22 /100m) (2016) 7.5%
Sydney ( 🇦🇺 ) OLYMPIC W 19:38 (01:19 /100m) (2012) 21:15 (01:25 /100m) (2010) 7.6%
Hamburg ( 🇩🇪 ) SPRINT W 09:07 (01:13 /100m) (2020) 09:53 (01:19 /100m) (2015) 7.7%
New Plymouth ( 🇳🇿 ) SPRINT W 10:17 (01:22 /100m) (2018) 11:09 (01:29 /100m) (2017) 7.7%
Karlovy Vary ( 🇨🇿 ) OLYMPIC M 16:59 (01:08 /100m) (2018) 18:27 (01:14 /100m) (2019) 8.0%
Karlovy Vary ( 🇨🇿 ) OLYMPIC M 16:56 (01:08 /100m) (2021) 18:27 (01:14 /100m) (2019) 8.2%
Tongyeong ( 🇰🇷 ) SPRINT M 08:11 (01:05 /100m) (2017) 08:55 (01:11 /100m) (2016) 8.3%
New Plymouth ( 🇳🇿 ) SPRINT W 10:03 (01:20 /100m) (2023) 10:57 (01:28 /100m) (2019) 8.3%
San Diego ( 🇺🇸 ) OLYMPIC M 16:12 (01:05 /100m) (2013) 17:41 (01:11 /100m) (2012) 8.3%
Karlovy Vary ( 🇨🇿 ) OLYMPIC W 18:08 (01:13 /100m) (2023) 19:51 (01:19 /100m) (2019) 8.7%
Hamburg ( 🇩🇪 ) SPRINT W 09:07 (01:13 /100m) (2020) 09:59 (01:20 /100m) (2024) 8.7%
New Plymouth ( 🇳🇿 ) SPRINT M 08:51 (01:11 /100m) (2014) 09:42 (01:18 /100m) (2019) 8.8%
Karlovy Vary ( 🇨🇿 ) OLYMPIC W 18:06 (01:12 /100m) (2022) 19:51 (01:19 /100m) (2019) 8.8%
Yokohama ( 🇯🇵 ) OLYMPIC M 17:33 (01:10 /100m) (2014) 19:18 (01:17 /100m) (2016) 9.1%
Yokohama ( 🇯🇵 ) OLYMPIC M 17:31 (01:10 /100m) (2018) 19:18 (01:17 /100m) (2016) 9.2%
New Plymouth ( 🇳🇿 ) SPRINT W 09:56 (01:19 /100m) (2015) 10:57 (01:28 /100m) (2019) 9.3%
Stockholm ( 🇸🇪 ) OLYMPIC M 17:42 (01:11 /100m) (2016) 19:32 (01:18 /100m) (2015) 9.4%
Montreal ( 🇨🇦 ) SPRINT M 08:09 (01:05 /100m) (2023) 09:00 (01:12 /100m) (2019) 9.4%
Karlovy Vary ( 🇨🇿 ) OLYMPIC M 16:40 (01:07 /100m) (2022) 18:27 (01:14 /100m) (2019) 9.7%
Tongyeong ( 🇰🇷 ) SPRINT M 08:00 (01:04 /100m) (2019) 08:55 (01:11 /100m) (2016) 10.4%
Karlovy Vary ( 🇨🇿 ) OLYMPIC M 16:27 (01:06 /100m) (2023) 18:27 (01:14 /100m) (2019) 10.9%
New Plymouth ( 🇳🇿 ) SPRINT W 09:56 (01:19 /100m) (2015) 11:09 (01:29 /100m) (2017) 10.9%
Sydney ( 🇦🇺 ) OLYMPIC W 19:38 (01:19 /100m) (2012) 22:07 (01:28 /100m) (2011) 11.2%
New Plymouth ( 🇳🇿 ) SPRINT M 08:33 (01:08 /100m) (2015) 09:42 (01:18 /100m) (2019) 11.9%
Chicago ( 🇺🇸 ) OLYMPIC W 17:56 (01:12 /100m) (2015) 20:26 (01:22 /100m) (2014) 12.3%
Chicago ( 🇺🇸 ) OLYMPIC M 16:31 (01:06 /100m) (2015) 19:09 (01:17 /100m) (2014) 13.8%

The benefit of the wetsuit could be estimated with a third method (more naive):

Click to expand - Third approach.
wetsuit_2.png
Comparing duration of swims with and without wetsuit.

The idea is to compare the average swim durations, with- and without wetsuit:

  • improve_percent = (time_no_wetsuit - time_wetsuit) / time_no_wetsuit.
  • This formula is applied to four cases: W-sprint, W-olympic, M-sprint, M-olympic. And results are written in the title of the above figure.
  • Note how the swim histograms of the ⏱️ PACES section reveals two distinct modes: with- (in violet, with lower times) and without wetsuit (larger times).

Outliers:

⚠️ CRITICISMS.

  • This method incorporates more data, but yields less reliable results to the following factors:
    • 📏 Swim course distances vary between races.
    • ⏱️ Timing methods (e.g., timing mat locations) are not always consistent.
  • For exemple:
    • The difference between the fastest and the slowest sprint+men+same-equipment swim exceeds 2 minutes.
      • Specifically, swim times at Mooloolaba ( 🇦🇺 ) are unusually fast considering the absence of wetsuit.
    • For olympic+men+no-wetsuit, the difference is 5:45.
  • Clearly, swim distances or timing methods vary between races, even when they are labeled with the same format.
  • Nonetheless, we can hope that these variations will balance out thanks to the large amount of data.

📊 Conclusion

How much faster are top swimmers (5-9th) with the wetsuit?

  • Method 1: improve_percent = 5.4% (women only), leveraging events where the "women-without / men-with-wetsuit scenario" occurs.
    • Refinements of this method suggest lower benefits, around 4-5%.
  • Method 2: improve_percent = 3.6%, using on year-to-year comparisons.
  • Method 3: improve_percent = 4.1% (olympic format only), naively comparing the overall average times with and without wetsuit.
  • Which method do you think is the most appropriate? What could be improved? If you have any comments or ideas, please do not hesitate to contact me!

Does the wetsuit benefit equally women and men?

  • The methods 2 and 3 suggest that women benefit slightly more from the wetsuit, with a tiny improvement difference of 0.1-0.3%.
    • It that statistically significant?
    • This finding is somewhat surprising, given that women already have a higher natural buoyancy, which could suggest a smaller relative benefit.
    • The first method could only compute improve_percent for women, limiting direct comparability.

🎓 Literature review

Here are some findings of related scientific works:

  • This 2019 research by Gay et al. asked 33 swimmers to perform 2×400-m maximal front crawl in a 25-m swimming pool, with wetsuit and with swimsuit. Participants were good swimmers, but not as fast as ITU elite athletes: 1:27 / 100m average on the 400m with swimsuit.
    • The wetsuit allows for a ~6% improvement.
    • Interestingly: "Swimmers reduced stroke rate and increased stroke length (by 4%) to benefit from the hydrodynamic reduction of the wetsuit and increase their swimming efficiency."
  • This 2022 meta study by Gay et al. concludes from 26 studies, a "3.2–12.9% velocity increments in distances ranging from 25 to 1500 m" for the full-body wetsuit.
    • The range is broad: it depends on many factors such as the profile of the swimmer (age, level, triathlete or swimmer), the swimming conditions (temperature, 25m pool vs open water) and the wetsuit itself.
    • This interview of Ana Gay in triathlete.com gives a good introduction to her study.
  • In this episode by scientifictriathlon.com, Maria Francesca Piacentini mentions some of her research on wetsuit, and claims 6% to 11% improvements.
    • Apart from the time reduction, her team found that wetsuit usage can make athletes feel less fatigued: during the 2x7x200m tests, the stroke index and the stroke length significantly decreased in the swimsuit condition, whereas they remained relatively stable in the wetsuit condition.
  • This article from sports-performance-bulletin reports improvements by 3% to 7%.
    • Sources are not explicitly referenced, but the article probably mentions this 1995 research by Chatard et al. which apart from computing improvements, shows that the impact of the wetsuit is very different for competitive swimmers than for competitive triathletes.

The 4-5% improvement observed in top swimmers (ranked 5th-9th) based on World Triathlon race data appears to align reasonably well with the findings of related research publications.

  • ⚠️ It is important to note that the lab- or pool-conditions used by most studies differ significantly from the open-water race environment.

Is the wetsuit worth for 300m?

When they can choose, pro athletes decide to use the wetsuit for ~300m swim, e.g. during mixed team relay. Examples:

Is this decision sound?

Click to expand - 🐧 Answering the question "wetsuit for 300m?"

Two quantities must be compared:

  • (1) How much time is spent for the wetsuit during T1?
    • According to the next section, the extra-time added by the wetsuit during T1 is around 9 seconds.
    • Before this analysis, I would have estimated a lower impact, around 6s.
  • (2) How much time is saved by the wetsuit during the 300m swim?
    • According to the present section, the benefit of the wetsuit ranges between 3.6% and 5.4%.
    • These estimates are based on the 1500m swim in Olympic-distance triathlons. Are they applicable to a 300m swim?
    • A 300m swim typically takes between 3:30 and 4:00. See justification below.
    • Therefore, the time saved with a wetsuit is estimated to be 7.6 seconds (3.6% of 3:30) and 13 seconds (5.4% of 4:00).

💡 Conclusion

  • 7.6 - 9 = -1.4 < 0
  • 13 - 9 = 4 > 0
  • Depending on the selected parameters, the total wetsuit gain for a 300m swim ranges from -1.4 to 4.0 seconds.
  • Beyond time saving, wetsuits provide additional benefits:
    • Temperature comfort and potential energy savings in the legs.
    • Athletes may also gain a couple of seconds to catch their breath while removing it at T1?
  • Unique conditions.
    • The 300m swim is featured in events like the mixed team relay and supertri, with smaller fields compared to the 55-athlete mass starts typical of sprint or olympic formats.
    • Could it be more advantageous to try to draft without a wetsuit at the back of the small group, prioritizing a fast T1, rather than fighting for a front position and risking delays from wetsuit removal failures?
      • This is particularly relevant for the first relay athlete, as subsequent (2nd, 3rd, and 4th) often swim in smaller groups or even alone, reducing the drafting and position-fighting dynamics.
    • Further research is needed to determine if skipping the wetsuit in these particular contexts could provide a strategic advantage.
  • An interesting real-world experiment: At the 2024 Toulouse ( 🇫🇷 ) supertri some athletes opted out of wetsuits for the first 300m swim:

Appendix: Deriving that 300m swim takes about 3:30 - 4:00.

Benefit 03:20 03:25 03:30 03:35 03:40 03:45 03:50 03:55 04:00 04:05 04:10
3.0% 6.0 6.1 6.3 6.4 6.6 6.8 6.9 7.0 7.2 7.4 7.5
3.2% 6.4 6.6 6.7 6.9 7.0 7.2 7.4 7.5 7.7 7.8 8.0
3.4% 6.8 7.0 7.1 7.3 7.5 7.6 7.8 8.0 8.2 8.3 8.5
3.6% 7.2 7.4 7.6 7.7 7.9 8.1 8.3 8.5 8.6 8.8 9.0
3.8% 7.6 7.8 8.0 8.2 8.4 8.6 8.7 8.9 9.1 9.3 9.5
4.0% 8.0 8.2 8.4 8.6 8.8 9.0 9.2 9.4 9.6 9.8 10.0
4.2% 8.4 8.6 8.8 9.0 9.2 9.5 9.7 9.9 10.1 10.3 10.5
4.4% 8.8 9.0 9.2 9.5 9.7 9.9 10.1 10.3 10.6 10.8 11.0
4.6% 9.2 9.4 9.7 9.9 10.1 10.4 10.6 10.8 11.0 11.3 11.5
4.8% 9.6 9.8 10.1 10.3 10.6 10.8 11.0 11.3 11.5 11.8 12.0
5.0% 10.0 10.3 10.5 10.8 11.0 11.3 11.5 11.8 12.0 12.3 12.5
5.2% 10.4 10.7 10.9 11.2 11.4 11.7 12.0 12.2 12.5 12.7 13.0
5.4% 10.8 11.1 11.3 11.6 11.9 12.2 12.4 12.7 13.0 13.2 13.5
5.6% 11.2 11.5 11.8 12.0 12.3 12.6 12.9 13.2 13.4 13.7 14.0
5.8% 11.6 11.9 12.2 12.5 12.8 13.1 13.3 13.6 13.9 14.2 14.5
6.0% 12.0 12.3 12.6 12.9 13.2 13.5 13.8 14.1 14.4 14.7 15.0
6.2% 12.4 12.7 13.0 13.3 13.6 14.0 14.3 14.6 14.9 15.2 15.5
6.4% 12.8 13.1 13.4 13.8 14.1 14.4 14.7 15.0 15.4 15.7 16.0
6.6% 13.2 13.5 13.9 14.2 14.5 14.9 15.2 15.5 15.8 16.2 16.5
6.8% 13.6 13.9 14.3 14.6 15.0 15.3 15.6 16.0 16.3 16.7 17.0

🚿 WETSUIT AT T1

This section addresses the question:

"How much time does the wetsuit add to T1?"

To answer this, the goal is to calculate:

extra_time_for_wetsuit = t1_with_wetsuit - t1_without_wetsuit

The challenge is that only one of these two t1_ values is typically available: the one recorded during the race.

  • In practice, when the wetsuit is allowed, everyone wears it.
    • It does not happen for a race to have groups of athletes split between wearing and not wearing wetsuits, which would have allowed direct measurement of both t1_ times under the same conditions.
  • Consequently, this section introduces two methods to estimate the missing t1_, enabling the calculation of the extra_time_for_wetsuit difference.

Note: For each race analysed, the average T1 duration is calculated, excluding the 5 slowest transition times to reduce the influence of outliers.

🌍 Method 1: Using Recurring Events

This method identifies pairs of events held at the same venue, where wetsuits were required in one year but not in another.

  • The approach assumes that the course layout remains consistent across years and that timing method for T1 do not change significantly.
  • For example, in the women’s race at Cagliari ( :italy: ), wetsuits were worn in 2024 (T1 time: 46.9s) but not in 2023 (T1 time: 39.1s) and not in 2022 (T1 time: 37.8s).
  • Using these three events, two comparisons can be made to estimate extra_time_for_wetsuit:
    • 2024 (wetsuit) vs 2023 (no-wetsuit): extra_time_for_wetsuit = 46.9s - 39.1s = 7.9s.
    • 2024 (wetsuit) vs 2022 (no-wetsuit): extra_time_for_wetsuit = 46.9s - 37.8s = 9.1s.
  • By collecting many such comparisons, a distribution of extra_time_for_wetsuit is generated, see the following figure.
  • The mean and median of this distribution provide an answer to the initial question: "How much time does the wetsuit add to T1?"

Limitations:

  • The exact length of T1 may vary if the course layout changes between years or if timing mats are moved.
    • Such differences can render comparisons inaccurate.
    • For instance, an extra 15 meters in T1 would add about 3 seconds at a speed of 18 km/h.
    • One option considered was to exclude comparisons between events that were several years apart. However, since the final results were not significantly affected, this rule was not applied to maintain a larger dataset.
  • To address this, data cleaning is applied to remove implausible entries.
    • Determining the appropriate upper threshold for such exclusions remains a challenge: 15s? 20s? 30s?
    • But this choice is not critical, as explained below.

Advantages:

  • extra_time_for_wetsuit can be calculated separately for men, women, or combined datasets.
  • The method benefits from a large number of comparisons (more than 300), improving the robustness of the estimate:
    • Variations in course length are mitigated since some comparisons will involve longer and others shorter T1 courses, balancing the overall distribution.
    • Outliers have minimal impact on the final aggregated estimate due to the high volume of data points. Testing with different thresholds did not show any significant difference in the final estimate.
t1_with_wetsuit.png
Distribution of the estimated extra_time_for_wetsuit, i.e. the additional time charged to the wetsuit at T1, using the method of recurring events.

Results from 270 comparisons:

  • The wetsuit adds 9.3 (mean) or 9.2 (median) seconds in T1. (std = 2.7)

  • Women are 0.1 (with means) or 0.4 (with medians) seconds slower than men.

    • Women only: mean = 9.4, median = 9.4. (std = 2.6)
    • Men only: mean = 9.3, median = 9.0. (std = 2.7)

  • Restricting to world-series yields ~9.6 extra seconds (from 206 comparisons) both for men and women.

Click to expand - ⚖️ Full list of comparisons used for this derivation.

Distributions:

  • Event category:
    • WTCS : 80% (215)
    • World-Cup : 20% (55)
  • Distance category:
    • Olympic : 52% (140)
    • Sprint : 48% (130)

Venues of the comparisons:

  • 95 (35.2%): Hamburg ( 🇩🇪 )
  • 63 (23.3%): Yokohama ( 🇯🇵 )
  • 28 (10.4%): Edmonton ( 🇨🇦 )
  • 21 ( 7.8%): London ( 🇬🇧 )
  • 20 ( 7.4%): Tongyeong ( 🇰🇷 )
  • 13 ( 4.8%): New Plymouth ( 🇳🇿 )
  • 10 ( 3.7%): Karlovy Vary ( 🇨🇿 )
  • 7 ( 2.6%): Auckland ( 🇳🇿 )
  • 4 ( 1.5%): Cannigione, Arzachena ( 🇮🇹 )
  • 4 ( 1.5%): Cagliari ( 🇮🇹 )
  • 2 ( 0.7%): San Diego ( 🇺🇸 )
  • 2 ( 0.7%): Valencia ( 🇪🇸 )
  • 1 ( 0.4%): Stockholm ( 🇸🇪 )
EVENT GENDER T1 WITH WETSUIT (s) T1 WITHOUT WETSUIT (s) EXTRA TIME FOR WETSUIT (s)
London ( 🇬🇧 ) W 41.3 (2012) 37.3 (2009) 4.0
Yokohama ( 🇯🇵 ) W 66.7 (2023) 62.4 (2009) 4.3
Karlovy Vary ( 🇨🇿 ) M 27.2 (2022) 22.8 (2019) 4.3
Tongyeong ( 🇰🇷 ) M 48.5 (2017) 44.1 (2015) 4.5
Tongyeong ( 🇰🇷 ) M 37.0 (2021) 32.5 (2022) 4.5
Yokohama ( 🇯🇵 ) W 67.1 (2022) 62.4 (2009) 4.8
Karlovy Vary ( 🇨🇿 ) M 27.7 (2023) 22.8 (2019) 4.9
London ( 🇬🇧 ) W 45.4 (2013) 40.3 (2011) 5.0
Hamburg ( 🇩🇪 ) W 39.9 (2019) 34.8 (2010) 5.0
Tongyeong ( 🇰🇷 ) M 48.5 (2017) 43.3 (2016) 5.2
Hamburg ( 🇩🇪 ) M 38.8 (2016) 33.6 (2010) 5.2
Hamburg ( 🇩🇪 ) M 38.8 (2016) 33.6 (2022) 5.3
Tongyeong ( 🇰🇷 ) W 52.2 (2017) 46.9 (2015) 5.3
London ( 🇬🇧 ) M 42.2 (2014) 36.8 (2011) 5.4
London ( 🇬🇧 ) W 45.7 (2014) 40.3 (2011) 5.4
Hamburg ( 🇩🇪 ) M 39.1 (2017) 33.6 (2010) 5.5
Hamburg ( 🇩🇪 ) M 39.1 (2017) 33.6 (2022) 5.5
Hamburg ( 🇩🇪 ) M 36.2 (2019) 30.6 (2024) 5.6
Hamburg ( 🇩🇪 ) M 39.2 (2009) 33.6 (2010) 5.6
Hamburg ( 🇩🇪 ) M 39.2 (2009) 33.6 (2022) 5.6
Yokohama ( 🇯🇵 ) M 61.6 (2023) 55.9 (2009) 5.6
Karlovy Vary ( 🇨🇿 ) W 30.2 (2022) 24.6 (2019) 5.7
Karlovy Vary ( 🇨🇿 ) W 30.3 (2023) 24.6 (2019) 5.7
Edmonton ( 🇨🇦 ) W 68.0 (2014) 62.2 (2013) 5.8
Hamburg ( 🇩🇪 ) W 42.2 (2009) 36.4 (2022) 5.9
Hamburg ( 🇩🇪 ) M 39.5 (2021) 33.6 (2010) 5.9
Hamburg ( 🇩🇪 ) M 39.5 (2021) 33.6 (2022) 5.9
Hamburg ( 🇩🇪 ) W 42.3 (2017) 36.4 (2022) 5.9
Tongyeong ( 🇰🇷 ) M 48.5 (2017) 42.6 (2014) 6.0
Hamburg ( 🇩🇪 ) W 42.4 (2021) 36.4 (2022) 6.0
London ( 🇬🇧 ) M 39.8 (2012) 33.8 (2009) 6.1
Hamburg ( 🇩🇪 ) M 36.2 (2019) 30.1 (2015) 6.1
Yokohama ( 🇯🇵 ) M 61.6 (2023) 55.5 (2021) 6.1
Hamburg ( 🇩🇪 ) W 42.6 (2016) 36.4 (2022) 6.2
New Plymouth ( 🇳🇿 ) W 30.5 (2018) 24.2 (2017) 6.3
Tongyeong ( 🇰🇷 ) M 38.9 (2019) 32.5 (2022) 6.3
Cannigione, Arzachena ( 🇮🇹 ) W 48.2 (2020) 41.9 (2022) 6.4
Tongyeong ( 🇰🇷 ) W 41.5 (2021) 35.1 (2022) 6.4
Hamburg ( 🇩🇪 ) W 39.9 (2019) 33.4 (2015) 6.5
Edmonton ( 🇨🇦 ) W 68.8 (2021) 62.2 (2013) 6.5
New Plymouth ( 🇳🇿 ) M 28.2 (2017) 21.6 (2019) 6.5
Tongyeong ( 🇰🇷 ) W 41.7 (2019) 35.1 (2022) 6.5
Karlovy Vary ( 🇨🇿 ) W 31.1 (2021) 24.6 (2019) 6.6
New Plymouth ( 🇳🇿 ) W 30.8 (2014) 24.2 (2017) 6.6
Stockholm ( 🇸🇪 ) M 51.7 (2016) 45.1 (2015) 6.6
Yokohama ( 🇯🇵 ) W 66.7 (2023) 60.1 (2021) 6.6
Hamburg ( 🇩🇪 ) M 36.2 (2019) 29.5 (2013) 6.6
Tongyeong ( 🇰🇷 ) M 52.6 (2009) 45.9 (2011) 6.6
Yokohama ( 🇯🇵 ) W 69.1 (2015) 62.4 (2009) 6.7
Hamburg ( 🇩🇪 ) M 36.2 (2019) 29.4 (2018) 6.7
New Plymouth ( 🇳🇿 ) M 28.4 (2018) 21.6 (2019) 6.7
Yokohama ( 🇯🇵 ) M 61.6 (2023) 54.8 (2017) 6.8
Yokohama ( 🇯🇵 ) W 69.2 (2024) 62.4 (2009) 6.8
Hamburg ( 🇩🇪 ) W 43.3 (2012) 36.4 (2022) 6.9
Tongyeong ( 🇰🇷 ) M 39.4 (2023) 32.5 (2022) 6.9
Edmonton ( 🇨🇦 ) W 68.0 (2014) 61.1 (2019) 7.0
Hamburg ( 🇩🇪 ) M 36.2 (2019) 29.2 (2014) 7.0
Yokohama ( 🇯🇵 ) W 67.1 (2022) 60.1 (2021) 7.1
London ( 🇬🇧 ) M 43.9 (2013) 36.8 (2011) 7.1
Yokohama ( 🇯🇵 ) M 61.6 (2023) 54.5 (2024) 7.1
Yokohama ( 🇯🇵 ) W 69.6 (2018) 62.4 (2009) 7.2
Karlovy Vary ( 🇨🇿 ) M 30.0 (2020) 22.8 (2019) 7.2
New Plymouth ( 🇳🇿 ) W 30.5 (2018) 23.3 (2019) 7.2
Hamburg ( 🇩🇪 ) W 39.9 (2019) 32.7 (2024) 7.2
Hamburg ( 🇩🇪 ) W 39.9 (2019) 32.6 (2013) 7.2
Karlovy Vary ( 🇨🇿 ) M 30.1 (2021) 22.8 (2019) 7.3
New Plymouth ( 🇳🇿 ) W 31.5 (2016) 24.2 (2017) 7.3
Yokohama ( 🇯🇵 ) W 69.7 (2014) 62.4 (2009) 7.3
Hamburg ( 🇩🇪 ) W 42.2 (2009) 34.8 (2010) 7.4
Yokohama ( 🇯🇵 ) M 63.4 (2015) 55.9 (2009) 7.4
Hamburg ( 🇩🇪 ) W 42.3 (2017) 34.8 (2010) 7.5
Yokohama ( 🇯🇵 ) M 63.4 (2018) 55.9 (2009) 7.5
New Plymouth ( 🇳🇿 ) W 30.8 (2014) 23.3 (2019) 7.5
Yokohama ( 🇯🇵 ) M 61.6 (2023) 54.0 (2016) 7.5
Hamburg ( 🇩🇪 ) W 42.4 (2021) 34.8 (2010) 7.6
Cagliari ( 🇮🇹 ) W 46.3 (2024) 38.8 (2023) 7.6
Karlovy Vary ( 🇨🇿 ) W 32.2 (2020) 24.6 (2019) 7.6
London ( 🇬🇧 ) M 44.4 (2015) 36.8 (2011) 7.6
Yokohama ( 🇯🇵 ) W 66.7 (2023) 59.0 (2017) 7.7
Yokohama ( 🇯🇵 ) W 66.7 (2023) 59.0 (2016) 7.7
Edmonton ( 🇨🇦 ) W 68.8 (2021) 61.1 (2019) 7.7
Tongyeong ( 🇰🇷 ) W 54.6 (2014) 46.9 (2015) 7.7
New Plymouth ( 🇳🇿 ) M 29.5 (2014) 21.6 (2019) 7.8
Hamburg ( 🇩🇪 ) W 42.6 (2016) 34.8 (2010) 7.8
Hamburg ( 🇩🇪 ) W 39.9 (2019) 32.0 (2018) 7.9
Yokohama ( 🇯🇵 ) M 63.4 (2015) 55.5 (2021) 7.9
Edmonton ( 🇨🇦 ) M 62.5 (2014) 54.6 (2019) 8.0
Yokohama ( 🇯🇵 ) M 63.4 (2018) 55.5 (2021) 8.0
Cannigione, Arzachena ( 🇮🇹 ) W 49.9 (2021) 41.9 (2022) 8.0
London ( 🇬🇧 ) W 45.4 (2013) 37.3 (2009) 8.1
Yokohama ( 🇯🇵 ) W 67.1 (2022) 59.0 (2017) 8.1
Yokohama ( 🇯🇵 ) W 67.1 (2022) 59.0 (2016) 8.1
Hamburg ( 🇩🇪 ) W 39.9 (2019) 31.7 (2014) 8.1
Karlovy Vary ( 🇨🇿 ) M 31.0 (2018) 22.8 (2019) 8.2
Hamburg ( 🇩🇪 ) M 38.8 (2016) 30.6 (2024) 8.2
New Plymouth ( 🇳🇿 ) W 31.5 (2016) 23.3 (2019) 8.2
Tongyeong ( 🇰🇷 ) W 43.4 (2023) 35.1 (2022) 8.3
London ( 🇬🇧 ) M 42.2 (2014) 33.8 (2009) 8.4
London ( 🇬🇧 ) W 45.7 (2014) 37.3 (2009) 8.4
Hamburg ( 🇩🇪 ) W 43.3 (2012) 34.8 (2010) 8.5
Hamburg ( 🇩🇪 ) M 39.1 (2017) 30.6 (2024) 8.5
Cannigione, Arzachena ( 🇮🇹 ) M 46.4 (2021) 37.9 (2020) 8.5
Hamburg ( 🇩🇪 ) M 42.1 (2012) 33.6 (2010) 8.5
Tongyeong ( 🇰🇷 ) M 52.6 (2009) 44.1 (2015) 8.5
Hamburg ( 🇩🇪 ) M 42.1 (2012) 33.6 (2022) 8.5
Hamburg ( 🇩🇪 ) M 39.2 (2009) 30.6 (2024) 8.6
Yokohama ( 🇯🇵 ) W 66.7 (2023) 58.1 (2019) 8.6
Tongyeong ( 🇰🇷 ) M 41.1 (2018) 32.5 (2022) 8.6
Yokohama ( 🇯🇵 ) M 63.4 (2015) 54.8 (2017) 8.6
Yokohama ( 🇯🇵 ) M 63.4 (2018) 54.8 (2017) 8.7
Edmonton ( 🇨🇦 ) W 70.9 (2016) 62.2 (2013) 8.7
New Plymouth ( 🇳🇿 ) M 30.4 (2016) 21.6 (2019) 8.7
Cannigione, Arzachena ( 🇮🇹 ) M 46.4 (2021) 37.7 (2022) 8.7
Hamburg ( 🇩🇪 ) M 38.8 (2016) 30.1 (2015) 8.7
Hamburg ( 🇩🇪 ) M 42.4 (2011) 33.6 (2010) 8.8
Hamburg ( 🇩🇪 ) M 42.4 (2011) 33.6 (2022) 8.8
Cagliari ( 🇮🇹 ) W 46.3 (2024) 37.5 (2022) 8.8
Karlovy Vary ( 🇨🇿 ) W 33.4 (2018) 24.6 (2019) 8.8
Hamburg ( 🇩🇪 ) M 39.5 (2021) 30.6 (2024) 8.9
Hamburg ( 🇩🇪 ) W 42.2 (2009) 33.4 (2015) 8.9
London ( 🇬🇧 ) W 49.2 (2015) 40.3 (2011) 8.9
New Plymouth ( 🇳🇿 ) M 30.5 (2015) 21.6 (2019) 8.9
Yokohama ( 🇯🇵 ) M 61.6 (2023) 52.7 (2022) 8.9
Yokohama ( 🇯🇵 ) M 63.4 (2015) 54.5 (2024) 8.9
Hamburg ( 🇩🇪 ) W 42.3 (2017) 33.4 (2015) 8.9
London ( 🇬🇧 ) W 41.3 (2012) 32.4 (2010) 9.0
Yokohama ( 🇯🇵 ) M 63.4 (2018) 54.5 (2024) 9.0
Hamburg ( 🇩🇪 ) W 42.4 (2021) 33.4 (2015) 9.0
Hamburg ( 🇩🇪 ) M 39.1 (2017) 30.1 (2015) 9.0
Yokohama ( 🇯🇵 ) M 61.6 (2023) 52.6 (2019) 9.0
Yokohama ( 🇯🇵 ) W 67.1 (2022) 58.1 (2019) 9.0
Yokohama ( 🇯🇵 ) W 69.1 (2015) 60.1 (2021) 9.0
Hamburg ( 🇩🇪 ) M 39.2 (2009) 30.1 (2015) 9.1
London ( 🇬🇧 ) M 39.8 (2012) 30.7 (2010) 9.1
Yokohama ( 🇯🇵 ) W 69.2 (2024) 60.1 (2021) 9.1
Tongyeong ( 🇰🇷 ) M 52.6 (2009) 43.3 (2016) 9.2
Hamburg ( 🇩🇪 ) W 42.6 (2016) 33.4 (2015) 9.2
Hamburg ( 🇩🇪 ) M 38.8 (2016) 29.5 (2013) 9.3
Tongyeong ( 🇰🇷 ) W 56.2 (2009) 46.9 (2015) 9.3
Yokohama ( 🇯🇵 ) M 63.4 (2015) 54.0 (2016) 9.3
Hamburg ( 🇩🇪 ) M 39.5 (2021) 30.1 (2015) 9.4
Yokohama ( 🇯🇵 ) M 63.4 (2018) 54.0 (2016) 9.4
Hamburg ( 🇩🇪 ) M 38.8 (2016) 29.4 (2018) 9.4
Edmonton ( 🇨🇦 ) M 64.0 (2016) 54.6 (2019) 9.4
New Plymouth ( 🇳🇿 ) W 33.6 (2015) 24.2 (2017) 9.5
Yokohama ( 🇯🇵 ) W 69.6 (2018) 60.1 (2021) 9.5
Yokohama ( 🇯🇵 ) M 65.4 (2014) 55.9 (2009) 9.5
Tongyeong ( 🇰🇷 ) W 44.6 (2018) 35.1 (2022) 9.5
Hamburg ( 🇩🇪 ) M 39.1 (2017) 29.5 (2013) 9.5
Hamburg ( 🇩🇪 ) W 42.2 (2009) 32.7 (2024) 9.6
Hamburg ( 🇩🇪 ) M 39.2 (2009) 29.5 (2013) 9.6
Yokohama ( 🇯🇵 ) W 69.7 (2014) 60.1 (2021) 9.6
Hamburg ( 🇩🇪 ) W 42.2 (2009) 32.6 (2013) 9.6
Hamburg ( 🇩🇪 ) M 39.1 (2017) 29.4 (2018) 9.6
Hamburg ( 🇩🇪 ) W 42.3 (2017) 32.7 (2024) 9.7
Hamburg ( 🇩🇪 ) W 42.3 (2017) 32.6 (2013) 9.7
Hamburg ( 🇩🇪 ) M 38.8 (2016) 29.2 (2014) 9.7
Hamburg ( 🇩🇪 ) W 42.4 (2021) 32.7 (2024) 9.7
Hamburg ( 🇩🇪 ) M 39.2 (2009) 29.4 (2018) 9.7
Hamburg ( 🇩🇪 ) W 42.4 (2021) 32.6 (2013) 9.7
Tongyeong ( 🇰🇷 ) W 56.8 (2011) 46.9 (2015) 9.8
Hamburg ( 🇩🇪 ) W 43.3 (2012) 33.4 (2015) 9.9
Edmonton ( 🇨🇦 ) W 70.9 (2016) 61.1 (2019) 9.9
Hamburg ( 🇩🇪 ) W 46.3 (2011) 36.4 (2022) 9.9
Hamburg ( 🇩🇪 ) M 39.5 (2021) 29.5 (2013) 9.9
Hamburg ( 🇩🇪 ) M 39.1 (2017) 29.2 (2014) 9.9
Yokohama ( 🇯🇵 ) M 65.4 (2014) 55.5 (2021) 10.0
Hamburg ( 🇩🇪 ) W 42.6 (2016) 32.7 (2024) 10.0
Tongyeong ( 🇰🇷 ) M 52.6 (2009) 42.6 (2014) 10.0
Hamburg ( 🇩🇪 ) M 39.2 (2009) 29.2 (2014) 10.0
Hamburg ( 🇩🇪 ) W 42.6 (2016) 32.6 (2013) 10.0
Hamburg ( 🇩🇪 ) M 39.5 (2021) 29.4 (2018) 10.0
Yokohama ( 🇯🇵 ) W 69.1 (2015) 59.0 (2017) 10.1
Edmonton ( 🇨🇦 ) M 62.5 (2014) 52.5 (2021) 10.1
Yokohama ( 🇯🇵 ) W 69.1 (2015) 59.0 (2016) 10.1
London ( 🇬🇧 ) M 43.9 (2013) 33.8 (2009) 10.1
Yokohama ( 🇯🇵 ) W 69.2 (2024) 59.0 (2017) 10.2
Yokohama ( 🇯🇵 ) W 69.2 (2024) 59.0 (2016) 10.2
Edmonton ( 🇨🇦 ) W 68.0 (2014) 57.8 (2017) 10.2
Hamburg ( 🇩🇪 ) W 42.2 (2009) 32.0 (2018) 10.3
Hamburg ( 🇩🇪 ) M 39.5 (2021) 29.2 (2014) 10.3
Hamburg ( 🇩🇪 ) W 42.3 (2017) 32.0 (2018) 10.3
New Plymouth ( 🇳🇿 ) W 33.6 (2015) 23.3 (2019) 10.4
Edmonton ( 🇨🇦 ) M 62.5 (2014) 52.1 (2017) 10.4
Hamburg ( 🇩🇪 ) W 42.4 (2021) 32.0 (2018) 10.4
Tongyeong ( 🇰🇷 ) W 57.4 (2016) 46.9 (2015) 10.4
Hamburg ( 🇩🇪 ) W 42.2 (2009) 31.7 (2014) 10.5
Yokohama ( 🇯🇵 ) W 69.6 (2018) 59.0 (2017) 10.5
Yokohama ( 🇯🇵 ) W 69.6 (2018) 59.0 (2016) 10.5
Hamburg ( 🇩🇪 ) W 42.3 (2017) 31.7 (2014) 10.6
Hamburg ( 🇩🇪 ) W 43.3 (2012) 32.7 (2024) 10.6
Hamburg ( 🇩🇪 ) W 43.3 (2012) 32.6 (2013) 10.6
Hamburg ( 🇩🇪 ) W 42.4 (2021) 31.7 (2014) 10.6
London ( 🇬🇧 ) M 44.4 (2015) 33.8 (2009) 10.7
Yokohama ( 🇯🇵 ) M 65.4 (2014) 54.8 (2017) 10.7
Yokohama ( 🇯🇵 ) W 69.7 (2014) 59.0 (2017) 10.7
Edmonton ( 🇨🇦 ) M 65.3 (2015) 54.6 (2019) 10.7
Yokohama ( 🇯🇵 ) M 63.4 (2015) 52.7 (2022) 10.7
Yokohama ( 🇯🇵 ) W 69.7 (2014) 59.0 (2016) 10.7
Hamburg ( 🇩🇪 ) W 42.6 (2016) 32.0 (2018) 10.7
Yokohama ( 🇯🇵 ) M 63.4 (2018) 52.7 (2022) 10.7
Yokohama ( 🇯🇵 ) M 63.4 (2015) 52.6 (2019) 10.8
Yokohama ( 🇯🇵 ) M 63.4 (2018) 52.6 (2019) 10.9
Edmonton ( 🇨🇦 ) W 73.1 (2015) 62.2 (2013) 10.9
Edmonton ( 🇨🇦 ) W 68.8 (2021) 57.8 (2017) 10.9
Hamburg ( 🇩🇪 ) W 42.6 (2016) 31.7 (2014) 10.9
Yokohama ( 🇯🇵 ) M 65.4 (2014) 54.5 (2024) 10.9
Yokohama ( 🇯🇵 ) W 69.1 (2015) 58.1 (2019) 10.9
Yokohama ( 🇯🇵 ) W 69.2 (2024) 58.1 (2019) 11.1
San Diego ( 🇺🇸 ) M 69.1 (2013) 57.9 (2012) 11.3
Hamburg ( 🇩🇪 ) W 43.3 (2012) 32.0 (2018) 11.3
Yokohama ( 🇯🇵 ) M 65.4 (2014) 54.0 (2016) 11.4
Yokohama ( 🇯🇵 ) W 69.6 (2018) 58.1 (2019) 11.4
London ( 🇬🇧 ) M 42.2 (2014) 30.7 (2010) 11.4
Hamburg ( 🇩🇪 ) W 46.3 (2011) 34.8 (2010) 11.5
Auckland ( 🇳🇿 ) M 65.2 (2011) 53.7 (2015) 11.5
Hamburg ( 🇩🇪 ) M 42.1 (2012) 30.6 (2024) 11.5
Edmonton ( 🇨🇦 ) M 64.0 (2016) 52.5 (2021) 11.5
Hamburg ( 🇩🇪 ) W 43.3 (2012) 31.7 (2014) 11.5
San Diego ( 🇺🇸 ) W 75.5 (2013) 64.0 (2012) 11.5
Yokohama ( 🇯🇵 ) W 69.7 (2014) 58.1 (2019) 11.6
Hamburg ( 🇩🇪 ) M 42.4 (2011) 30.6 (2024) 11.8
Edmonton ( 🇨🇦 ) M 64.0 (2016) 52.1 (2017) 11.9
London ( 🇬🇧 ) W 49.2 (2015) 37.3 (2009) 11.9
Hamburg ( 🇩🇪 ) M 42.1 (2012) 30.1 (2015) 12.0
Edmonton ( 🇨🇦 ) W 73.1 (2015) 61.1 (2019) 12.1
Edmonton ( 🇨🇦 ) W 68.0 (2014) 55.8 (2018) 12.2
Edmonton ( 🇨🇦 ) M 62.5 (2014) 50.3 (2018) 12.3
Auckland ( 🇳🇿 ) W 71.3 (2011) 59.0 (2015) 12.3
Hamburg ( 🇩🇪 ) M 42.4 (2011) 30.1 (2015) 12.3
Auckland ( 🇳🇿 ) M 65.2 (2011) 52.7 (2014) 12.5
Hamburg ( 🇩🇪 ) M 42.1 (2012) 29.5 (2013) 12.5
Hamburg ( 🇩🇪 ) M 42.1 (2012) 29.4 (2018) 12.7
Yokohama ( 🇯🇵 ) M 65.4 (2014) 52.7 (2022) 12.7
Edmonton ( 🇨🇦 ) M 65.3 (2015) 52.5 (2021) 12.8
Hamburg ( 🇩🇪 ) M 42.4 (2011) 29.5 (2013) 12.8
Yokohama ( 🇯🇵 ) M 65.4 (2014) 52.6 (2019) 12.8
Hamburg ( 🇩🇪 ) W 46.3 (2011) 33.4 (2015) 12.9
Hamburg ( 🇩🇪 ) M 42.1 (2012) 29.2 (2014) 12.9
Hamburg ( 🇩🇪 ) M 42.4 (2011) 29.4 (2018) 12.9
Edmonton ( 🇨🇦 ) W 68.8 (2021) 55.8 (2018) 13.0
London ( 🇬🇧 ) W 45.4 (2013) 32.4 (2010) 13.0
Edmonton ( 🇨🇦 ) W 70.9 (2016) 57.8 (2017) 13.1
Edmonton ( 🇨🇦 ) M 65.3 (2015) 52.1 (2017) 13.1
London ( 🇬🇧 ) M 43.9 (2013) 30.7 (2010) 13.1
Hamburg ( 🇩🇪 ) M 42.4 (2011) 29.2 (2014) 13.2
Yokohama ( 🇯🇵 ) M 61.6 (2023) 48.2 (2012) 13.4
London ( 🇬🇧 ) W 45.7 (2014) 32.4 (2010) 13.4
Valencia ( 🇪🇸 ) W 53.0 (2020) 39.6 (2022) 13.4
Valencia ( 🇪🇸 ) M 49.9 (2020) 36.3 (2022) 13.5
Hamburg ( 🇩🇪 ) W 46.3 (2011) 32.7 (2024) 13.6
Hamburg ( 🇩🇪 ) W 46.3 (2011) 32.6 (2013) 13.7
London ( 🇬🇧 ) M 44.4 (2015) 30.7 (2010) 13.7
Edmonton ( 🇨🇦 ) M 64.0 (2016) 50.3 (2018) 13.7
Edmonton ( 🇨🇦 ) M 68.7 (2011) 54.6 (2019) 14.1
Hamburg ( 🇩🇪 ) W 46.3 (2011) 32.0 (2018) 14.3
Cagliari ( 🇮🇹 ) W 59.7 (2016) 45.3 (2018) 14.4
Yokohama ( 🇯🇵 ) W 66.7 (2023) 52.3 (2012) 14.4
Hamburg ( 🇩🇪 ) W 46.3 (2011) 31.7 (2014) 14.6
Yokohama ( 🇯🇵 ) W 67.1 (2022) 52.3 (2012) 14.9
Auckland ( 🇳🇿 ) M 68.6 (2012) 53.7 (2015) 14.9
Edmonton ( 🇨🇦 ) M 65.3 (2015) 50.3 (2018) 15.0
Auckland ( 🇳🇿 ) W 71.3 (2011) 56.2 (2014) 15.1
Edmonton ( 🇨🇦 ) W 70.9 (2016) 55.8 (2018) 15.2
Yokohama ( 🇯🇵 ) M 63.4 (2015) 48.2 (2012) 15.2
Yokohama ( 🇯🇵 ) M 63.4 (2018) 48.2 (2012) 15.2
Edmonton ( 🇨🇦 ) W 73.1 (2015) 57.8 (2017) 15.3
Auckland ( 🇳🇿 ) W 74.9 (2012) 59.0 (2015) 15.9
Auckland ( 🇳🇿 ) M 68.6 (2012) 52.7 (2014) 16.0
Cagliari ( 🇮🇹 ) W 59.7 (2016) 43.7 (2017) 16.0

👫 Method 2: using women/men differences

The method leverages the women/men comparison, resembling the approach to estimate the benefit of the wetsuit.

As a riminder, the formula to compute the extra-time for wetsuit is:

extra_time_for_wetsuit = t1_with_wetsuit - t1_no_wetsuit

Let's consider as an example the 2024 Cagliari ( :italy: ) event:

  • This event is very special because women raced with wetsuits, while men did not.
  • Trying to apply the formula for women:
    • t1_with_wetsuit is known.
    • t1_no_wetsuit is unknown.
  • To estimate the missing t1_no_wetsuit for women, we introduce a function, f_from_men_to_women_no_wetsuit:
    • Input: t1_no_wetsuit for men.
    • Output: t1_no_wetsuit for women in the same event.
  • Using this function, the women's extra-time for wetsuit can be computed:
    • extra_time_for_wetsuit_women = t1_with_wetsuit_women - f(t1_no_wetsuit_men)
  • This approach will be repeated for all events where women used wetsuits and men did not: see the table below.

How to derive the f_from_men_to_women_no_wetsuit function?

  • Using all events where both men and women swam without wetsuits.

What model for this f_from_men_to_women_no_wetsuit function?

  • It is a design choice: A linear model? An affine model? A more complex function approximation such as a neural network? How many layers and parameters then?
  • I have decided to model T1 as a composition of two parts:
    • Static part: The athlete stands in front of the bike, drop the swimming equipments (goggles and caps) into the box, and put the helmet on. Reminder: both women and men without wetsuit here.
    • Moving part: The athlete runs from the water to the bike, and then runs pushing the bike until the mounting line at the end of the blue carpet.
  • I have made the following assumptions:
    • The duration of the static part, named helmet_duration, is identical for men and women.
    • Women are assumed to run slower than men by a constant percentage, diff_wm_t1_run.
      • diff_wm_t1_run = (t1_run_women - t1_run_men) / t1_run_men, where:
        • t1_run_women = t1_women - helmet_duration
        • t1_run_men = t1_men - helmet_duration

The resulting model to estimate the missing t1_no_wetsuit for women is:

t1_women_no_wetsuit = f(t1_men_no_wetsuit)
                    = (t1_men_no_wetsuit - helmet_duration) * (1 + diff_wm_t1_run) + helmet_duration`

Parameters:

  • helmet_duration is set to 3s.
    • Testing values of 1 or 5 seconds showed variations of less than 0.15 seconds in the final extra_time_for_wetsuit estimate.
  • diff_wm_t1_run is estimating using the 147 events where both women and men swam without wetsuit.
    • It is found to be 10.0%, i.e. women run 10% slower than men during T1.
    • This value is further discussed in a subsequent paragraph.

The derived f_from_men_to_women_no_wetsuit function is applied to events featuring the women-without / men-with scenario:

YEAR EVENT EXTRA TIME FOR WETSUIT (s) (women) EVENT CATEGORY
2011 Tongyeong ( 🇰🇷 ) 6.5 WORLD-CUP
2020 Cannigione, Arzachena ( 🇮🇹 ) 6.8 WORLD-CUP
2021 Haeundae ( 🇰🇷 ) 7.8 WORLD-CUP
2014 Tongyeong ( 🇰🇷 ) 8.1 WORLD-CUP
2024 Cagliari ( 🇮🇹 ) 8.6 WTCS
2012 Sydney ( 🇦🇺 ) 8.8 WTCS
2022 Yokohama ( 🇯🇵 ) 9.5 WTCS
2024 Yokohama ( 🇯🇵 ) 9.6 WTCS
2015 Stockholm ( 🇸🇪 ) 9.9 WTCS
2016 Tongyeong ( 🇰🇷 ) 10.0 WORLD-CUP
2021 Edmonton ( 🇨🇦 ) 11.3 WTCS

Results:

  • mean = median = 8.8 seconds (from 11 events).
  • Restricting to world-series:
    • The w/m difference for running during T1 stays around 10% (9.8%) (from 59 events).
    • The estimated extra_time_for_wetsuit for women raises by ~1s to 9.7 seconds (from 6 events).

Notes:

  • Using the women/men comparison offers a strong advantage: the exact distance of T1 does not matter, as the course and timing methods are identical for both genders.
  • One limitation: this method relies on rare events where women raced with wetsuits while men did not, or inversely for the men-estimate but the scenario is even rarer.

💡 FINDINGS AND DISCUSSION:

  • Both methods to estimate extra_time_for_wetsuit yield similar results:
    • The time charged to the wetsuit during the T1 transition lays around 9 seconds.
      • For world-series events, it is a bit higher: around 10 seconds.
    • These estimations could serve as training references for coaches and athletes.
  • I would first have expected a lower time.
    • However, T1 involves more than just removing the wetsuit in front of the bike.
    • Athletes must also run while wearing the wetsuit, which is heavy and elastic.
    • Additionally, they need to unzip the wetsuit, begin removing the sleeves while running, and ensure it is properly placed in the designated box.
  • Two additional findings regarding the women/men comparison:
    • Women appear slightly more affected by the wetsuit.
      • But the difference is tiny: < 0.4s, and completely negligible at the world-series level.
    • Women run 10% slower than men during T1.
      • This percentage can be compared to the results of the women/men comparison section, in particular the ~14% for the running leg.
      • Probably all athletes run slower during T1 compared to the subsequent 5k or 10k. For diverse reasons:
      • Probably the decrease in running speed is more important for men, leading to a lower gender difference.
  • Finally, while these estimates focus on time, the wetsuit also provides energy savings (e.g. buoyancy and warmth), as discussed in the section Is the wetsuit worth for 300m?.

🎯 RACE SCENARIO

This section looks at the race dynamics of WCTS and games-related events (no world-cup).

  • Based on the information "does a bunch manage to breakaway on the bike?" 🐦

To obtain this information, the size of the front pack at the end of the bike is estimated as follows:

  • Compute, for each athlete, the cumulative times after the bike:
    • start_to_t2 = swim + t1 + bike.
  • Identify which athlete enters T2 first: min(start_to_t2).
  • Count how many athletes enter T2 pack_duration_s, e.g. 10s, or less after this first athlete.
  • This gives the size of the front pack at the end of the bike.

Note: A breakaway on the bike can happen:

  • Either by attack and escape from the main bike pack (rare).
  • Or directly from the swim (most common).
  • It should be possible to automatically retrieve the breakaway type among the two, using the swim rank of athletes composing the breakaway. (I have not done it).

Two additional pieces of information are retrieved:

  • 🏆 winner_in_front_pack: was the winner already in the leading group at the end of the bike, or did she/he come back on the run?
  • 👟 is_best_runner_in_front_pack.
scenarios.png
Size of the front pack at the end of the bike.

About the number of finishers:

  • ~4.5 more finishers in men's races.
    • The standard deviation is higher for women's race.
  • ~3.5 more** finishers in sprint format races.
    • Because the olympic format is longer, weaker athletes are more likely to be caught and eliminated by being lapped?
  • It would be interesting to know the number of starters as well.

On the olympic format, men are more likely to break away.

  • Because they are stronger on the bike?
    • Georgia Taylor-Brown ( 🇬🇧 ), one of the strongest rider in the field, admits in this video: "I would love to be able to attack and stay away, but I do not have that power."
    • Probably bike is the sport among three where top athletes will progress the most in the future.
  • At the same time, the probability to win for athletes from the front group is much lower for men than for women.

Small front packs, i.e. small breakaways, are more likely on the olympic format.

  • Possibly because a longer swim leads to larger gaps at T1?
  • On the other hand the bike is longer than for the sprint format, which should give more time for the other packs to catch up.

Some very large front groups at T2 (London ( 🇬🇧 ) at the top 💂 ):

PACK_SIZE YEAR WINNER DISTANCE CAT EVENT
53 2011 Helen Jenkins ( 🇬🇧 ) OLYMPIC WTCS 2011 Dextro Energy Triathlon - ITU World Championship Series London ( 🇬🇧 )
46 2012 Erin Densham ( 🇦🇺 ) OLYMPIC WTCS 2012 Dextro Energy World Triathlon Sydney ( 🇦🇺 )
39 2011 Paula Findlay ( 🇨🇦 ) OLYMPIC WTCS 2011 Dextro Energy Triathlon - ITU World Championship Series Sydney ( 🇦🇺 )
39 2011 Andrea Hansen ( 🇳🇿 ) OLYMPIC WTCS 2011 Dextro Energy Triathlon - ITU World Championship Series Yokohama ( 🇯🇵 )
37 2012 Nicola Spirig ( 🇨🇭 ) OLYMPIC WTCS 2012 ITU World Triathlon Madrid ( 🇪🇸 )
... ... ... ... ... ...
52 2014 Mario Mola ( 🇪🇸 ) SPRINT WTCS 2014 ITU World Triathlon London ( 🇬🇧 )
51 2023 Alex Yee ( 🇬🇧 ) OLYMPIC GAMES 2023 World Triathlon Olympic Games Test Event Paris ( 🇫🇷 )
49 2023 Alex Yee ( 🇬🇧 ) SPRINT WTCS 2023 World Triathlon Championship Series Abu Dhabi ( 🇦🇪 )
46 2016 Mario Mola ( 🇪🇸 ) OLYMPIC WTCS 2016 ITU World Triathlon Yokohama ( 🇯🇵 )
45 2011 Brad Kahlefeldt ( 🇦🇺 ) OLYMPIC WTCS 2011 Dextro Energy Triathlon - ITU World Championship Series Hamburg ( 🇩🇪 )
Click to expand - 💪 Some wins via breakaway (with front-pack-size <= 3).
PACK_SIZE YEAR WINNER DISTANCE CAT EVENT
1 2011 Sarah Haskins ( 🇺🇸 ) OLYMPIC WORLD-CUP 2011 Monterrey ITU Triathlon World Cup ( 🇲🇽 )
1 2016 Flora Duffy ( 🇧🇲 ) OLYMPIC WTCS 2016 ITU World Triathlon Stockholm ( 🇸🇪 )
1 2018 Flora Duffy ( 🇧🇲 ) OLYMPIC WTCS 2018 ITU World Triathlon Bermuda ( 🇧🇲 )
1 2021 Taylor Knibb ( 🇺🇸 ) OLYMPIC WTCS 2021 World Triathlon Championship Finals Edmonton ( 🇨🇦 )
2 2011 Nicky Samuels ( 🇳🇿 ) OLYMPIC WORLD-CUP 2011 Mooloolaba ITU Triathlon World Cup ( 🇦🇺 )
2 2017 Flora Duffy ( 🇧🇲 ) OLYMPIC WTCS 2017 ITU World Triathlon Yokohama ( 🇯🇵 )
2 2021 Taylor Knibb ( 🇺🇸 ) OLYMPIC WTCS 2021 World Triathlon Championship Series Yokohama ( 🇯🇵 )
2 2022 Flora Duffy ( 🇧🇲 ) OLYMPIC WTCS 2022 World Triathlon Championship Series Bermuda ( 🇧🇲 )
3 2009 Emma Moffatt ( 🇦🇺 ) OLYMPIC WTCS 2009 Dextro Energy Triathlon - ITU World Championship Series Hamburg ( 🇩🇪 )
3 2014 Jodie Stimpson ( 🇬🇧 ) OLYMPIC WTCS 2014 ITU World Triathlon Auckland ( 🇳🇿 )
3 2016 Helen Jenkins ( 🇬🇧 ) OLYMPIC WTCS 2016 ITU World Triathlon Gold Coast ( 🇦🇺 )
3 2016 Flora Duffy ( 🇧🇲 ) OLYMPIC WTCS 2016 ITU World Triathlon Grand Final Cozumel ( 🇲🇽 )
3 2017 Flora Duffy ( 🇧🇲 ) OLYMPIC WTCS 2017 ITU World Triathlon Grand Final Rotterdam ( 🇳🇱 )
3 2019 Katie Zaferes ( 🇺🇸 ) OLYMPIC WTCS 2019 MS Amlin World Triathlon Bermuda ( 🇧🇲 )
3 2019 Julie Derron ( 🇨🇭 ) OLYMPIC WORLD-CUP 2019 Weihai ITU Triathlon World Cup ( 🇨🇳 )
3 2021 Maya Kingma ( 🇳🇱 ) OLYMPIC WTCS AJ Bell 2021 World Triathlon Championship Series Leeds ( 🇬🇧 )
3 2024 Lena Meißner ( 🇩🇪 ) OLYMPIC WORLD-CUP 2024 World Triathlon Cup Samarkand ( 🇺🇿 )
... ... ... ... ... ...
1 2013 Alistair Brownlee ( 🇬🇧 ) OLYMPIC WTCS 2013 ITU World Triathlon Stockholm ( 🇸🇪 )
1 2018 Casper Stornes ( 🇳🇴 ) OLYMPIC WTCS 2018 ITU World Triathlon Bermuda ( 🇧🇲 )
1 2023 Morgan Pearson ( 🇺🇸 ) OLYMPIC WORLD-CUP 2023 World Triathlon Cup Karlovy Vary ( 🇨🇿 )
2 2011 Kris Gemmell ( 🇳🇿 ) OLYMPIC WORLD-CUP 2011 Auckland ITU Triathlon World Cup ( 🇳🇿 )
2 2016 Rodrigo Gonzalez Lopez ( 🇲🇽 ) OLYMPIC WORLD-CUP 2016 Chengdu ITU Triathlon World Cup ( 🇨🇳 )
3 2009 Jan Frodeno ( 🇩🇪 ) OLYMPIC WTCS 2009 Dextro Energy Triathlon - ITU World Championship Series Yokohama ( 🇯🇵 )
3 2011 Alistair Brownlee ( 🇬🇧 ) OLYMPIC WTCS 2011 Dextro Energy Triathlon - ITU World Championship Series Kitzbuehel ( 🇦🇹 )
3 2014 Alistair Brownlee ( 🇬🇧 ) OLYMPIC WTCS 2014 ITU World Triathlon Grand Final Edmonton ( 🇨🇦 )
3 2015 Jonathan Brownlee ( 🇬🇧 ) OLYMPIC WTCS 2015 ITU World Triathlon Gold Coast ( 🇦🇺 )
3 2020 Vincent Luis ( 🇫🇷 ) OLYMPIC WORLD-CUP 2020 Karlovy Vary ITU Triathlon World Cup ( 🇨🇿 )

Strong bikers and very versatile triathletes! 👏

  • Especially Flora Duffy ( 🇧🇲 ) and Alistair Brownlee ( 🇬🇧 ).

Below is a more complicated figure: the bars show the evolution of the average first-pack size over the years.

scenarios_over_years.png
Size of front bike pack, over the years.

Run-comebacks, i.e. winning after not being in the front group after bike, are rare.

  • Apart from 2013-2016 (the era of Gwen Jorgensen ( 🇺🇸 )), no comeback has happened on women's olympic races and only a few on women's sprint format.
    • At the same time, the size of front-pack had reduced until 2022 and was even very small for some recent years (2017, 2019, 2021, 2022).
    • Now, top-women-swimmer can ride hard and run fast.
  • Helen Jenkins ( 🇬🇧 ) explains in this 2024 video: "Women's races have definitely changed over the past few years. (...) 2021 was that breakaway era. It definitely comes back to that larger front group over the last couple of years."
    • That statement is perfectly consistent with the women's olympic bar plot.
  • The men's olympic races follow a similar same trend: the front group at T2 has, on average, never been as large, as in 2023 and 2024.

🏃 How often does the best runner win? 🏆

best_runner_wins.png
How often does the best runner win?

More than 2/3 races are won by the best runner.

  • This percentage is higher on the sprint than on the olympic format.
    • Probably because the swim is shorter: "good-runner-but-bad-swimmer"s are less likely to miss the front group on the bike.
  • This percentage could be:
    • Higher, if the winner does not slow down and "enjoy" the last 100m run, and instead sprints until the end, as most athletes behind have to. Example: Leonie Periault ( 🇫🇷 ) in 2024 Yokohama ( 🇯🇵 ).
    • Lower, if good runners who, at T2, are already too far for the win, get as motivated as the athletes at the front, and give their best possible run.
  • The percentage drops to 50% when considering world-cup events only. Why?
    • Because athletes are not as complete as on WTCS?
    • There are more "good-runner-but-bad-swimmer"s who miss the front group on the bike, while good swimmers are not top runners?
Click to expand - Same plots for world-cup events only.
scenarios.png
Size of the front pack at the end of the bike.
scenarios_over_years.png
Size of front bike pack, over the years.
best_runner_wins_wc.png
How often does the best runner wins?

Why are there not more breakaways on the bike?

  • The mini areo TT bars have been banned in 2023.
  • Bike courses are mostly flat.
    • Events often take place on flat coasts, near to a sea, or in flat big cities.
    • Joel Filliol ( 🇨🇦 🏴󠁧󠁢󠁳󠁣󠁴󠁿 ) shares insights in this TTS podcast (around 43:00) about bike courses that allow for a break to stay away.
  • Because of U-turns.
  • Athletes may prefer to conserve energy for the run segment.
    • Valid of strong runners.
    • But for others? Maybe it is worth more to conserve energy by drafting and try to make top-20 with a good run, rather than risk a breakaway that might be caught, be burnt and come back home without any prize money or qualification points?
  • Not enough bike power?
    • Possibly, given these athletes need to be strong in swim and run as well, they cannot afford having massive legs.

🚀 SPRINT FINISH

This section studies the time gap between the winner and the second.

sprint_finish.png
Time gap between the winner and the second.
  • In men's races, 17% (sprint format) and 10% (olympic) are won by a sprint finish, occurring 50% more often than in women's races.
  • Women's races offer examples of wins by very large margins.
sprint_finish_over_years.png
Time gap between the winner and the second, over years. And the proportion of events with a contested win finish (less than 3s difference between first and second).

Gaps between the winner and the second are on average:

  • ~Twice as large in olympic formats compared to sprint formats.
  • ~Twice as large for women compared to men.

It has been a long time since a women's olympic race was won by a sprint.

Click to expand - 📸 Some of the most contested finishes on the blue carpet.
WOMEN
YEAR VENUE DIST. RACE CATEGORY FIRST ( 🥇 ) SECOND ( 🥈 )
2009 Madrid ( 🇪🇸 ) olympic WTCS Andrea Hansen ( 🇳🇿 ) Lisa Norden ( 🇸🇪 )
2010 Hamburg ( 🇩🇪 ) olympic WTCS Lisa Norden ( 🇸🇪 ) Emma Moffatt ( 🇦🇺 )
2010 Sydney ( 🇦🇺 ) olympic WTCS Barbara Riveros ( 🇨🇱 ) Andrea Hansen ( 🇳🇿 )
2010 Tiszaujvaros ( 🇭🇺 ) olympic world-cup Yuliya Yelistratova ( 🇺🇦 ) Jodie Swallow ( 🇬🇧 )
2011 Lausanne ( 🇨🇭 ) sprint WTCS Barbara Riveros ( 🇨🇱 ) Emma Jackson ( 🇦🇺 )
2012 Yokohama ( 🇯🇵 ) olympic WTCS Lisa Norden ( 🇸🇪 ) Anne Haug ( 🇩🇪 )
2012 London ( 🇬🇧 ) olympic games Nicola Spirig ( 🇨🇭 ) Lisa Norden ( 🇸🇪 )
2015 Tiszaujvaros ( 🇭🇺 ) sprint world-cup Felicity Sheedy-Ryan ( 🇦🇺 ) Audrey Merle ( 🇫🇷 )
2017 Cape Town ( 🇿🇦 ) sprint world-cup Lucy Buckingham ( 🇬🇧 ) Jessica Learmonth ( 🇬🇧 )
2017 Abu Dhabi ( 🇦🇪 ) olympic WTCS Andrea Hansen ( 🇳🇿 ) Jodie Stimpson ( 🇬🇧 )
2017 New Plymouth ( 🇳🇿 ) sprint world-cup Katie Zaferes ( 🇺🇸 ) Joanna Brown ( 🇨🇦 )
2018 Cagliari ( 🇮🇹 ) sprint world-cup Lisa Perterer ( 🇦🇹 ) Taylor Spivey ( 🇺🇸 )
2018 Karlovy Vary ( 🇨🇿 ) olympic world-cup Vendula Frintova ( 🇨🇿 ) Kaidi Kivioja ( 🇪🇪 )
2019 Madrid ( 🇪🇸 ) sprint world-cup Emilie Morier ( 🇫🇷 ) Sandra Dodet ( 🇫🇷 )
2019 Tiszaujvaros ( 🇭🇺 ) sprint world-cup Emma Jeffcoat ( 🇦🇺 ) Sara Vilic ( 🇦🇹 )
2022 Bergen ( 🇳🇴 ) sprint world-cup Tilda Månsson ( 🇸🇪 ) Jolien Vermeylen ( 🇧🇪 )
2023 Tiszaujvaros ( 🇭🇺 ) sprint world-cup Tilda Månsson ( 🇸🇪 ) Noelia Juan ( 🇪🇸 )
2024 Huatulco ( 🇲🇽 ) sprint world-cup Alberte Kjær Pedersen ( 🇩🇰 ) Rachel Klamer ( 🇳🇱 )
2024 Wollongong ( 🇦🇺 ) sprint world-cup Tilda Månsson ( 🇸🇪 ) Maria Carolina Velasquez Soto ( 🇨🇴 )
MEN
YEAR VENUE DIST. RACE CATEGORY FIRST ( 🥇 ) SECOND ( 🥈 )
2009 Tongyeong ( 🇰🇷 ) olympic WTCS Bevan Docherty ( 🇳🇿 ) Brad Kahlefeldt ( 🇦🇺 )
2009 Des Moines ( 🇺🇸 ) olympic world-cup Simon Whitfield ( 🇨🇦 ) Brad Kahlefeldt ( 🇦🇺 )
2010 Seoul ( 🇰🇷 ) olympic WTCS Jan Frodeno ( 🇩🇪 ) Courtney Atkinson ( 🇦🇺 )
2011 Hamburg ( 🇩🇪 ) olympic WTCS Brad Kahlefeldt ( 🇦🇺 ) William Clarke ( 🇬🇧 )
2012 Mooloolaba ( 🇦🇺 ) olympic world-cup Laurent Vidal ( 🇫🇷 ) Brad Kahlefeldt ( 🇦🇺 )
2013 London ( 🇬🇧 ) olympic WTCS Javier Gomez Noya ( 🇪🇸 ) Jonathan Brownlee ( 🇬🇧 )
2013 Hamburg ( 🇩🇪 ) sprint WTCS Jonathan Brownlee ( 🇬🇧 ) Alistair Brownlee ( 🇬🇧 )
2014 Chengdu ( 🇨🇳 ) olympic world-cup Wian Sullwald ( 🇿🇦 ) Kevin McDowell ( 🇺🇸 )
2014 London ( 🇬🇧 ) sprint WTCS Mario Mola ( 🇪🇸 ) Richard Murray ( 🇿🇦 )
2014 Yokohama ( 🇯🇵 ) olympic WTCS Javier Gomez Noya ( 🇪🇸 ) Mario Mola ( 🇪🇸 )
2014 Tiszaujvaros ( 🇭🇺 ) sprint world-cup Ákos Vanek ( 🇭🇺 ) Rostislav Pevtsov ( 🇦🇿 )
2015 Chengdu ( 🇨🇳 ) olympic world-cup Ryan Fisher ( 🇦🇺 ) Rostislav Pevtsov ( 🇦🇿 )
2016 Miyazaki ( 🇯🇵 ) olympic world-cup Uxio Abuin Ares ( 🇪🇸 ) Joao Silva ( 🇵🇹 )
2016 Salinas ( 🇪🇨 ) sprint world-cup David Castro Fajardo ( 🇪🇸 ) Matthew McElroy ( 🇺🇸 )
2016 Tongyeong ( 🇰🇷 ) sprint world-cup Uxio Abuin Ares ( 🇪🇸 ) Matthew McElroy ( 🇺🇸 )
2018 Tongyeong ( 🇰🇷 ) sprint world-cup Max Studer ( 🇨🇭 ) Felix Duchampt ( 🇷🇴 )
2019 Cagliari ( 🇮🇹 ) sprint world-cup Alistair Brownlee ( 🇬🇧 ) Justus Nieschlag ( 🇩🇪 )
2019 Hamburg ( 🇩🇪 ) sprint WTCS Jacob Birtwhistle ( 🇦🇺 ) Vincent Luis ( 🇫🇷 )
2019 Montreal ( 🇨🇦 ) sprint WTCS Jelle Geens ( 🇧🇪 ) Mario Mola ( 🇪🇸 )
2019 Madrid ( 🇪🇸 ) sprint world-cup Justus Nieschlag ( 🇩🇪 ) Lasse Lührs ( 🇩🇪 )
2019 Tiszaujvaros ( 🇭🇺 ) sprint world-cup Eli Hemming ( 🇺🇸 ) Ryan Fisher ( 🇦🇺 )
2020 Mooloolaba ( 🇦🇺 ) sprint world-cup Ryan Sissons ( 🇳🇿 ) Hayden Wilde ( 🇳🇿 )
2021 Edmonton ( 🇨🇦 ) olympic WTCS Kristian Blummenfelt ( 🇳🇴 ) Marten Van Riel ( 🇧🇪 )
2021 Hamburg ( 🇩🇪 ) sprint WTCS Tim Hellwig ( 🇩🇪 ) Paul Georgenthum ( 🇫🇷 )
2022 Bergen ( 🇳🇴 ) sprint world-cup Dorian Coninx ( 🇫🇷 ) Kristian Blummenfelt ( 🇳🇴 )
2022 Huatulco ( 🇲🇽 ) sprint world-cup Genis Grau ( 🇪🇸 ) Tyler Mislawchuk ( 🇨🇦 )
2023 Pontevedra ( 🇪🇸 ) olympic WTCS Dorian Coninx ( 🇫🇷 ) Tim Hellwig ( 🇩🇪 )
2023 Sunderland ( 🇬🇧 ) sprint WTCS Pierre Le Corre ( 🇫🇷 ) Léo Bergere ( 🇫🇷 )
2023 Huatulco ( 🇲🇽 ) sprint world-cup David Castro Fajardo ( 🇪🇸 ) Tyler Mislawchuk ( 🇨🇦 )
2023 Valencia ( 🇪🇸 ) olympic world-cup David Cantero Del Campo ( 🇪🇸 ) Lasse Nygaard Priester ( 🇩🇪 )

Not many highly contested finishes on women's world-series for a while!

By the way, World Triathlon rules that the win must be contested:

  • Triathletes should not "finish in a contrived tie situation where no effort to separate the finish times has been made".
  • At Tokyo ( 🇯🇵 ) 2019, Jessica Learmonth ( 🇬🇧 ) and Georgia Taylor-Brown ( 🇬🇧 ) were disqualified after crossing line hand-in-hand.

🗓️ LEVEL OVER YEARS

The same outlier removal is applied as described in the ⏱️ PACES section.

repeated_events_standard_w.png
Olympic format. Women.
repeated_events_standard_m.png
Olympic format. Men.
Click to expand - 🚀 Same plots for the Sprint format.
repeated_events_sprint_w.png
Sprint format. Women.
repeated_events_sprint_m.png
Sprint format. Men.
Click to expand - 🏅 Same plots for the Top-3.
repeated_events_standard_w_top3.png
Olympic format. Top-3 women in each leg.
repeated_events_standard_w_top3.png
Olympic format. Top-3 men in each leg.
repeated_events_sprint_w_top3.png
Sprint format. Top-3 women in each leg.
repeated_events_sprint_w_top3.png
Sprint format. Top-3 men in each leg.
Click to expand - 🏁 Same plots for the Top-10.
repeated_events_standard_w_top10.png
Olympic format. Top-10 women in each leg.
repeated_events_standard_w_top10.png
Olympic format. Top-10 men in each leg.
repeated_events_sprint_w_top10.png
Sprint format. Top-10 women in each leg.
repeated_events_sprint_w_top10.png
Sprint format. Top-10 men in each leg.

In this early-2024 video, Vincent Luis ( 🇫🇷 ) explains:

"My running times are very similar to what I was doing in 2019-2020, when I was world champion. [...] It's just that the level has gone up [...]. I ran at 20km/h in Yokohama (2024). This is a pace that 5-6 years ago was enough to win some World Series."

The above plots seem consistent with this statement. 📈

  • The figures for women's and men's olympic-format show recent improvements of the 10k run pace (green bars) on WTCS races. For 2019 -> 2021 -> 2023, paces were:
    • 3:33 -> 3:28 -> 3:23 for women.
    • 3:06 -> 3:04 -> 2:59 for men.
  • The 3:00 /km pace mention by Luis was, in 2023, hardly enough to finish between 5-th and 9-th on WTCS men's races.
  • The same 3:00 /km was, in 2019, probably enough to win a WTCS, since the 5-th to 9-th places were about 3:06 /km at that time.

In this other 2024 video:

  • Alex Yee ( 🇬🇧 ) mentions that the run of 2012 London ( 🇬🇧 ) Olympics is a reference: "The run was held as the best run that has ever been done in triathlon".
  • He explains: "In the last season (2023), we had a few races which came very close to that."
  • This statement seems also to be correct: on the run subplot of the men's olympic format figure, London 2012 was the fastest until 2023.
  • This article from triathlon.org, released just before Paris ( 🇫🇷 ) 2024, confirms: "Brownlee’s times (London ( 🇬🇧 ) 2012) will likely come under threat. Indeed, it seems highly likely that we could see the first ever sub-29 and sub-33 minute 10km times in an Olympic triathlon this summer."
    • The 5-9th men ran at the 2024 Paris Olympics much slower that the year before for the test event.
    • Also, Alex Yee ( 🇬🇧 ) won in 2024 with a 29:49 run, compared to 29:00 in 2023.
    • Because of the heat (the men's race started at 10:45 am instead of 8:00 am)?
    • Or the fatigue caused by the very long swim?
    • Or was the run course longer in 2024? In this case the women's run time improvement would be even more impressive: Beth Potter ( 🇬🇧 ) did 32:57 in 2023, compared to 32:42 for Cassandre Beaugrand ( 🇫🇷 ) in 2024, who run faster than 30% of men finishers (15 / 50)!

Yokohama 🇯🇵

  • The event took place 13 times between 2009 and 2024 in an olympic format.
    • I do not know if the courses have changed.
    • But if not, the comparison should be very relevant.
  • Run times have decreased in the last five editions. 👟
    • Leading to the best ever times in 2024.
  • It is interesting to note that runs were particularly good on olympics years (2012, 2016, 2021, 2024), especially for women.
    • Can it be that this race (usually happening early in the season) was used as qualification criteria or as a rehearsal, and therefore attracting very fit and motivated athletes?

Hamburg 🇩🇪

  • The event took place 12 times between 2009 and 2024 in a sprint format.
  • Run times have been constant until 2024, where a clear improvement can be seen.
    • The 2024 race was used by many athletes as a final repetition before the Olympics. But still, the improvement is huge.
    • Can it be that the 2024 run course was shorter? Indeed, the 2024 result page indicates Run 4.905km (2 laps) for 2024, compared to Run 5km (2 laps) for 2022.

Impact of carbon plate running shoes? 👟

  • The Nike Vaporfly came out in 2017.
  • As for Yokohama ( 🇯🇵 ), the running paces have been constantly improving on WTCS since 2017-2018.
    • Running times were already good before 2014.
    • But since 2021, paces have never been so low.
  • Carbon plate technology could be one of the main factors explaining this improvement.
    • But how to explain that running performances on world-cups have not followed the same trend?

The bike(s) 🚲

  • On olympic format, bike times have been drastically improving (ignoring covid 2020 year) over the past 6 years, especially for the men.
  • Can it be due to tech innovations?
  • Maybe the level has gone up: it is no longer just about being a good swimmer and an excellent runner?
  • Maybe some athletes tend to take more risks on the bike, sometimes reckless as reports Vincent Luis ( 🇫🇷 ) in this interview.

World-cup vs WTCS 🏆

  • Apart for 2018, the running level is consistently higher in WTCS than in world-cups.
  • Which could be expected since WTCS are so much selective.

The swim of 2024 Paris ( 🇫🇷 ) Olympics was very though because of the current on the way back:

  • 20:26 (01:22 /100m) for men.
  • 22:33 (01:30 /100m) for women.

Care is needed when comparing swim times:

  • In the WTCS (green bars), women's times appear to have reached historic lows, while 2024 men's times are the third slowest since 2009.
  • This is likely because, in the two 2024 olympic WTCS events considered (Yokohama ( 🇯🇵 ) and Cagliari ( 🇮🇹 )), women swam with wetsuits while men did not.

Criticisms ⚠️

  • Not all events have identical distances and conditions.
    • However, averaging many events (the number below the year on the x-axis), with multiple venues repeating every year, helps to mitigate this variability.
  • Swim 🏊
    • Distances can vary based on buoy positions.
    • Differences in water conditions (e.g. rough sea, current) can also significantly affect swim times, making comparisons challenging.
    • Last but not least, the wetsuit may be allowed or not.
  • Bike 🚴
    • Weather conditions, particularly wind and rain, can influence bike times.
    • Variations in course profiles (hilly vs. flat) can make direct comparisons of bike times unfair.
  • Run 🏃
    • Run times are generally more comparable as World Triathlon run courses tend to be predominantly flat, reducing variability.

🌡️ TEMPERATURES

This section examines the recorded water and air temperatures and, inspired by the recent work by Gibson (2024), investigates their impact on swimming and running performance, respectively.

temperatures.png
Recorded water and air temperatures.

The temperature ranges are broad:

  • 🌊 80% of the recorded water temperatures are between 16.3 and 26.6 °C.
    • (Mean: 21.4 °C, SD: 3.8 °C, Min: 13.7 °C, Max: 31.8 °C)
  • ⛱️ 80% of the recorded air temperatures are between 17.0 and 29.6 °C.
    • (Mean: 23.2 °C, SD: 5.2 °C, Min: 7.6 °C, Max: 35.5 °C)

Several limitations should be considered:

  • Incomplete data: Many events lack temperature data, limiting the analysis's comprehensiveness.
  • Pre-race measurements: Measurements are typically taken before the event.
    • Depending on the race timetable, the actual temperature experienced during the race can be lower or higher than the reported temperature.
    • "Water temperature must be taken one hour prior to the start of the event on competition day. It must be taken at the middle of the course and in two other areas on the swim course, at a depth of 60 cm."
    • Air temperatures are likely recorded before the race as well, potentially leading to significant temperature increases by the time athletes begin the run.
    • For instance, the air temperature for the men's race (10:45 am) at Paris 2024 ( 🇫🇷 ) is reported at 23.9 °C, while it was closer to 28 °C during the run.
  • Humidity Impact: While humidity could significantly affect running performance, its values are not reported.
  • Last but not least, swim and run courses vary in distance, and water conditions (e.g., waves, salinity, current) also differ between events, making comparisons challenging.

⚠️ Given these limitations, especially the variation in course distances, conclusions should be drawn with caution.

🌊 Water temperatures and swim times

temperatures_water.png
Water temperature and swim times.

The women's race at Haeundae ( 🇰🇷 ) (2021) is excluded due to an inconsistency with the 20°C wetsuit rule:

  • The race report notes: "Water temperature 21.3ºC. Air temperature 15.4º C. Wetsuits allowed."
  • The 2024 rule book states in section 4.4.b. that "when the water temperature is at or below 22ºC and the air temperature is at or below 15ºC, then the value of the water temperature will be adjusted."
    • For instance: Air at 15°C and water at 22°C -> The water temperature is adjusted at 18.5°C -> Wetsuit allowed.
    • It is possible that the 15°C threshold was higher in 2021, leading to the discrepancy.

Swim appears slightly faster in water temperatures below 20°C.

  • Likely because wetsuits are allowed at these lower temperatures.
  • Further research could analyse the impact of temperature on swim performance (see already Gay et al, 2021), and try to determine the optimal water temperature range for races, with and without wetsuits.
Click to expand - ❌ Anomalies ignored for this analysis.
SWIM CONDITIONS EVENT WETSUIT WATER TEMPERATURE (°C) 🌊 AIR TEMPERATURE (°C) 🌡️
🔥 wetsuit & water>=20°C 2021 Haeundae (w) ( 🇰🇷 ) True 21.3 15.4 ( ❄️ )
❄️ no-wetsuit & water<20°C 2014 Chicago (w) ( 🇺🇸 ) False 19.2 33.5 ( 🥵 )
❄️ no-wetsuit & water<20°C 2014 Chicago (m) ( 🇺🇸 ) False 19.2 33.5 ( 🥵 )
❄️ no-wetsuit & water<20°C 2013 Hamburg (w) ( 🇩🇪 ) False 17.4 23.8
❄️ no-wetsuit & water<20°C 2013 Hamburg (m) ( 🇩🇪 ) False 18.3 29.6 ( 🥵 )
❄️ no-wetsuit & water<20°C 2012 Sydney (m) ( 🇦🇺 ) False 19.5 22.0
❄️ no-wetsuit & water<20°C 2011 London (w) ( 🇬🇧 ) False 17 21.8
❄️ no-wetsuit & water<20°C 2011 London (m) ( 🇬🇧 ) False 19.5 21.3
🔥 wetsuit & water>=20°C 2011 Kitzbuhel (m) ( 🇦🇹 ) True 20.5 16.0 ( ❄️ )

As already explained above for Haeundae ( 🇰🇷 ) (2021), when the air temperature is extreme, the water temperature is adjusted.

But what is the reason for the other inconsistencies with the 20°C rule?

  • Probably errors in the data reported?
  • Unless the 20°C threshold was different at the time.

🏊 Some cold and hot swims:

YEAR EVENT WATER TEMPERATURE DISTANCE EVENT CATEGORY
2023 Vina del Mar ( 🇨🇱 ) 13.7 🥶 sprint WORLD-CUP
2017 Cape Town ( 🇿🇦 ) 13.8 🥶 sprint WORLD-CUP
2022 Vina del Mar ( 🇨🇱 ) 14.3 🥶 sprint WORLD-CUP
2023 Sunderland ( 🇬🇧 ) 14.4 🥶 sprint WTCS
2012 Auckland ( 🇳🇿 ) 14.6 🥶 olympic WTCS
2014 Stockholm ( 🇸🇪 ) 15.0 🥶 sprint WTCS
2023 Tongyeong ( 🇰🇷 ) 15.0 🥶 sprint WORLD-CUP
2013 Stockholm ( 🇸🇪 ) 15.0 🥶 olympic WTCS
2022 Bergen ( 🇳🇴 ) 15.0 🥶 sprint WORLD-CUP
2018 Cape Town ( 🇿🇦 ) 15.0 🥶 sprint WORLD-CUP
... ... ... ... ...
2023 Valencia ( 🇪🇸 ) 29.0 🥵 olympic WORLD-CUP
2017 Mérida, Yucatán, Puerto Progreso ( 🇲🇽 ) 29.0 🥵 sprint WORLD-CUP
2018 Mooloolaba ( 🇦🇺 ) 29.1 🥵 sprint WORLD-CUP
2019 Huatulco ( 🇲🇽 ) 29.5 🥵 sprint WORLD-CUP
2018 Huatulco ( 🇲🇽 ) 30.0 🥵 sprint WORLD-CUP
2024 Huatulco ( 🇲🇽 ) 30.0 🥵 sprint WORLD-CUP
2021 Huatulco ( 🇲🇽 ) 30.0 🥵 sprint WORLD-CUP
2021 Abu Dhabi ( 🇦🇪 ) 31.0 🥵 sprint WTCS
2022 Huatulco ( 🇲🇽 ) 31.0 🥵 sprint WORLD-CUP
2023 Huatulco ( 🇲🇽 ) 31.8 🥵 sprint WORLD-CUP

⛱️ Air temperatures and run times

temperatures_air.png
Air temperature and run times.

A 2nd degree fit is applied on the scatter plot using seaborn.regplot.

  • It suggests a trend where higher temperatures correlate with slower running paces.
  • While no definitive "optimal" temperature can be reliably determined, this trend aligns with both research findings and personal experiences ( 🥵 ).
  • It also underscores the importance of hydration and cooling strategies, such as the use of cooling headbands.

🏃 Some cold and hot runs:

YEAR EVENT AIR TEMPERATURE DISTANCE EVENT CATEGORY
2015 Edmonton ( 🇨🇦 ) 7.6 🥶 sprint WTCS
2023 Tongyeong ( 🇰🇷 ) 10.0 🥶 sprint WORLD-CUP
2023 Vina del Mar ( 🇨🇱 ) 11.0 🥶 sprint WORLD-CUP
2022 Vina del Mar ( 🇨🇱 ) 11.0 🥶 sprint WORLD-CUP
2016 Edmonton ( 🇨🇦 ) 12.7 🥶 sprint WTCS
2021 Tongyeong ( 🇰🇷 ) 13.2 🥶 sprint WORLD-CUP
2011 Kitzbuhel ( 🇦🇹 ) 13.5 🥶 olympic WTCS
2018 New Plymouth ( 🇳🇿 ) 13.7 🥶 sprint WORLD-CUP
2023 Miyazaki ( 🇯🇵 ) 14.3 🥶 olympic WORLD-CUP
2017 Rotterdam ( 🇳🇱 ) 14.4 🥶 olympic WTCS
... ... ... ... ...
2015 Rio de Janeiro ( 🇧🇷 ) 32.1 🥵 olympic GAMES
2018 Huatulco ( 🇲🇽 ) 32.1 🥵 sprint WORLD-CUP
2014 Tiszaujvaros ( 🇭🇺 ) 32.4 🥵 sprint WORLD-CUP
2014 Chicago ( 🇺🇸 ) 33.5 🥵 olympic WTCS
2024 Huatulco ( 🇲🇽 ) 33.9 🥵 sprint WORLD-CUP
2016 Cozumel ( 🇲🇽 ) 34.0 🥵 olympic WTCS
2010 Madrid ( 🇪🇸 ) 34.0 🥵 olympic WTCS
2021 Abu Dhabi ( 🇦🇪 ) 34.0 🥵 sprint WTCS
2023 Huatulco ( 🇲🇽 ) 35.0 🥵 sprint WORLD-CUP
2015 Tiszaujvaros ( 🇭🇺 ) 35.5 🥵 sprint WORLD-CUP

🌍 HOST COUNTRIES

Countries having hosted more than one world-series, world-cup or games-related event since 2009:

COUNTRY COUNT VENUES
JPN ( 🇯🇵 ) 21 Yokohama (13), Miyazaki (6), Ishigaki (1), Tokyo (1)
AUS ( 🇦🇺 ) 18 Mooloolaba (9), Gold Coast (5), Sydney (3), Wollongong (1)
ESP ( 🇪🇸 ) 17 Madrid (7), Valencia (3), Banyoles (2), Pontevedra (2), Palamos (1), Alicante (1), Huelva (1)
MEX ( 🇲🇽 ) 16 Huatulco (8), Cozumel (4), Monterrey (2), Cancun (1), Mérida (1)
KOR ( 🇰🇷 ) 14 Tongyeong (9), Tongyeong (2), Seoul (1), Haeundae (1), Yeongdo (1)
GER ( 🇩🇪 ) 14 Hamburg (14)
GBR ( 🇬🇧 ) 14 London (7), Leeds (5), Birmingham (1), Sunderland (1)
CAN ( 🇨🇦 ) 13 Edmonton (9), Montreal (4)
NZL ( 🇳🇿 ) 11 New Plymouth (6), Auckland (4), Napier (1)
ITA ( 🇮🇹 ) 11 Cagliari (7), Arzachena (3), Rome (1)
HUN ( 🇭🇺 ) 10 Tiszaujvaros (9)
CHN ( 🇨🇳 ) 10 Chengdu (5), Weihai (2), Beijing (1), Jiayuguan (1), Weihai (1)
UAE ( 🇦🇪 ) 8 Abu Dhabi (8)
CZE ( 🇨🇿 ) 6 Karlovy Vary (6)
USA ( 🇺🇸 ) 6 Des Moines (2), San Diego (2), Chicago (2)
SWE ( 🇸🇪 ) 5 Stockholm (5)
AUT ( 🇦🇹 ) 4 Kitzbuhel (4)
RSA ( 🇿🇦 ) 4 Cape Town (4)
BER ( 🇧🇲 ) 3 Bermuda (3)
BRA ( 🇧🇷 ) 3 Rio de Janeiro (2)
ECU ( 🇪🇨 ) 3 Salinas (3)
SUI ( 🇨🇭 ) 3 Lausanne (3)
CHI ( 🇨🇱 ) 2 Vina del Mar (2)
TUR ( 🇹🇷 ) 2 Alanya (2)

It can be that some events are missing: these entries come from the data used for this report, after filtering and cleaning.

Click to expand - Top host countries for world-series.
COUNTRY COUNT VENUES
GER ( 🇩🇪 ) 14 Hamburg (14)
JPN ( 🇯🇵 ) 13 Yokohama (13)
GBR ( 🇬🇧 ) 12 London (6), Leeds (5), Sunderland (1)
CAN ( 🇨🇦 ) 10 Edmonton (7), Montreal (3)
AUS ( 🇦🇺 ) 8 Gold Coast (5), Sydney (3)
UAE ( 🇦🇪 ) 8 Abu Dhabi (8)
ESP ( 🇪🇸 ) 6 Madrid (5), Pontevedra (1)
SWE ( 🇸🇪 ) 5 Stockholm (5)
USA ( 🇺🇸 ) 4 San Diego (2), Chicago (2)
AUT ( 🇦🇹 ) 4 Kitzbuhel (4)
Click to expand - Top host countries for world-cups.
COUNTRY COUNT VENUES
MEX ( 🇲🇽 ) 15 Huatulco (8), Cozumel (3), Monterrey (2), Cancun (1), Mérida (1)
KOR ( 🇰🇷 ) 12 Tongyeong (8), Tongyeong (2), Haeundae (1), Yeongdo (1)
ESP ( 🇪🇸 ) 11 Valencia (3), Banyoles (2), Madrid (2), Palamos (1), Alicante (1), Huelva (1), Pontevedra (1)
AUS ( 🇦🇺 ) 10 Mooloolaba (9), Wollongong (1)
CHN ( 🇨🇳 ) 9 Jintang (3), Weihai (2), Chengdu (2), Gansu (1), Weihai, Shandong (1)
HUN ( 🇭🇺 ) 9 Tiszaujvaros (7), Tiszaujvaros (2)
ITA ( 🇮🇹 ) 8 Cagliari (4), Arzachena (3), Rome (1)
NZL ( 🇳🇿 ) 8 New Plymouth (6), Auckland (1), Napier (1)
JPN ( 🇯🇵 ) 7 Miyazaki (6), Ishigaki (1)
CZE ( 🇨🇿 ) 6 Karlovy Vary (6)

📆 SEASON DURATION

This section examines the duration of the competition season and the number of races athletes participate in.

season_duration.png
Duration of World Cup and World Series seasons, as well as the duration of the seasons of the top 50 athletes in the WTCS ranking.

World-Series

2020, 2021 and to some extent 2022 have been affected by the covid pandemic. Their rows are written in italic.

Year Num. events Season duration Start End First event Last event
2009 7 127 days (~ 4.2 m) 05-02 09-09 Tongyeong ( 🇰🇷 ) Gold Coast ( 🇦🇺 )
2010 7 147 days (~ 4.9 m) 04-11 09-08 Sydney ( 🇦🇺 ) Budapest ( 🇭🇺 )
2011 8 160 days (~ 5.3 m) 04-09 09-19 Sydney ( 🇦🇺 ) Yokohama ( 🇯🇵 )
2012 8 186 days (~ 6.2 m) 04-14 10-20 Sydney ( 🇦🇺 ) Auckland ( 🇳🇿 )
2013 7 155 days (~ 5.2 m) 04-06 09-11 Auckland ( 🇳🇿 ) London ( 🇬🇧 )
2014 7 143 days (~ 4.8 m) 04-06 08-29 Auckland ( 🇳🇿 ) Edmonton ( 🇨🇦 )
2015 9 189 days (~ 6.3 m) 03-06 09-15 Abu Dhabi ( 🇦🇪 ) Chicago ( 🇺🇸 )
2016 9 186 days (~ 6.2 m) 03-05 09-11 Abu Dhabi ( 🇦🇪 ) Cozumel ( 🇲🇽 )
2017 9 191 days (~ 6.4 m) 03-03 09-14 Abu Dhabi ( 🇦🇪 ) Rotterdam ( 🇳🇱 )
2018 8 190 days (~ 6.3 m) 03-02 09-12 Abu Dhabi ( 🇦🇪 ) Gold Coast ( 🇦🇺 )
2019 8 171 days (~ 5.7 m) 03-08 08-29 Abu Dhabi ( 🇦🇪 ) Lausanne ( 🇨🇭 )
2020 1 0 days ( 😷 🤒 ) 09-05 09-05 Hamburg ( 🇩🇪 ) Hamburg ( 🇩🇪 )
2021 5 170 days (~ 5.7 m) 05-15 11-05 Yokohama ( 🇯🇵 ) Abu Dhabi ( 🇦🇪 )
2022 6 190 days (~ 6.3 m) 05-14 11-24 Yokohama ( 🇯🇵 ) Abu Dhabi ( 🇦🇪 )
2023 6 199 days (~ 6.6 m) 03-03 09-22 Abu Dhabi ( 🇦🇪 ) Pontevedra ( 🇪🇸 )
2024 2 14 days (~ 0.5 m) 05-11 05-25 Yokohama ( 🇯🇵 ) Cagliari ( 🇮🇹 )

World-Cups

Year Num. events Season duration Start End First event Last event
2009 3 131 days (~ 4.4 m) 06-27 11-08 Des Moines ( 🇺🇸 ) Huatulco ( 🇲🇽 )
2010 5 193 days (~ 6.4 m) 03-27 10-10 Mooloolaba ( 🇦🇺 ) Huatulco ( 🇲🇽 )
2011 8 234 days (~ 7.8 m) 03-26 11-20 Mooloolaba ( 🇦🇺 ) Auckland ( 🇳🇿 )
2012 7 193 days (~ 6.4 m) 03-24 10-07 Mooloolaba ( 🇦🇺 ) Cancun ( 🇲🇽 )
2013 10 221 days (~ 7.4 m) 03-16 10-27 Mooloolaba ( 🇦🇺 ) Guatape ( 🇨🇴 )
2014 9 213 days (~ 7.1 m) 03-15 10-18 Mooloolaba ( 🇦🇺 ) Tongyeong ( 🇰🇷 )
2015 7 220 days (~ 7.3 m) 03-14 10-24 Mooloolaba ( 🇦🇺 ) Tongyeong ( 🇰🇷 )
2016 10 227 days (~ 7.6 m) 03-12 10-29 Mooloolaba ( 🇦🇺 ) Miyazaki ( 🇯🇵 )
2017 12 263 days (~ 8.8 m) 02-11 11-04 Cape Town ( 🇿🇦 ) Miyazaki ( 🇯🇵 )
2018 11 269 days (~ 9.0 m) 02-11 11-10 Cape Town ( 🇿🇦 ) Miyazaki ( 🇯🇵 )
2019 13 257 days (~ 8.6 m) 02-09 10-26 Cape Town ( 🇿🇦 ) Miyazaki ( 🇯🇵 )
2020 4 233 days (~ 7.8 m) 03-14 11-07 Mooloolaba ( 🇦🇺 ) Valencia ( 🇪🇸 )
2021 6 158 days (~ 5.3 m) 05-22 10-30 Lisbon ( 🇵🇹 ) Tongyeong ( 🇰🇷 )
2022 9 165 days (~ 5.5 m) 05-28 11-13 Arzachena ( 🇮🇹 ) Vina del Mar ( 🇨🇱 )
2023 14 226 days (~ 7.5 m) 03-25 11-11 New Plymouth ( 🇳🇿 ) Vina del Mar ( 🇨🇱 )
2024 6 84 days (~ 2.8 m) 02-24 05-18 Napier ( 🇳🇿 ) Samarkand ( 🇺🇿 )

Athlete seasons

The following plots represent the seasons (columns) of 50 athletes (rows).

athlete_season_duration_m.png
Seasons of the top-50 athletes. Men.
athlete_season_duration_w.png
Seasons of the top-50 athletes. Women.

Top athletes seem to prefer world series over world cups (more red than blue in the top rows).

  • Probably because WTCS events offer more points and prize money.
  • Some athletes do not race in any world cup events (light blue background).

Regarding the competition season: from 2009 to 2019, it became longer.

  • The 2009 WTCS season started in May, but from 2015-2019, it began in early March.
  • In 2023, the WTCS season was 6.5 months long, compared to 4.2 months in 2009.
  • The number of world cup events increased significantly during this period: from 3 in 2009 to 13 in 2019.
    • The 2018 world cup season lasted 9 months.
  • The COVID-19 pandemic halted this trend, although 2023 seems similar to 2019.
    • The ranking process became complicated: e.g., some 2021 races were considered for the 2022 ranking.

Regarding the athletes' season: it follows the competition season.

  • Their World Triathlon season extended from 130 days in 2009 to 200 days in 2023.
  • On average, athletes raced 10 times in 2019 and 2023, compared to 6 times in 2009.

The results found (number of races and season duration) are lower bounds:

  • Athletes participate in other formats besides World Triathlon olympic and sprint races.
    • For instance, the French Grand Prix was very popular in the 2010s.
    • Some athletes are racing the supertri and Ironman 70.3 formats as well.
  • Their competition season is probably longer, including indoor races in the winter.

🌍 ATHLETE NATIONS

This table examines the nationalities of athletes in the top-50s of the WTCS Ranking from 2009 to 2023:

RANGE (%) NATIONS (W) NATIONS (M)
11-12 🇺🇸 (11.0%)
10-11 🇬🇧 (10.9%) 🇦🇺 (10.1%)
9-10 🇫🇷 (9.6%)
8-9 🇯🇵 (8.0%) 🇬🇧 (8.4%) 🇩🇪 (8.1%)
7-8 🇩🇪 (7.6%) 🇦🇺 (7.7%) 🇪🇸 (7.0%)
6-7 🇺🇸 (6.1%)
5-6 🇳🇿 (5.7%) 🇫🇷 (5.0%)
4-5 🇳🇿 (4.6%) 🇨🇭 (4.6%) 🇨🇦 (4.4%) 🇵🇹 (4.3%)
3-4 🇨🇦 (3.7%) 🇪🇸 (3.7%) 🇮🇹 (3.7%) 🇳🇱 (3.6%) 🇨🇭 (3.3%) 🇷🇺 (3.6%) 🇿🇦 (3.4%)
2-3 🇦🇹 (2.6%) 🇲🇽 (2.1%) 🇧🇷 (2.0%) 🇯🇵 (2.9%) 🇧🇪 (2.9%) 🇭🇺 (2.6%) 🇮🇹 (2.4%) 🇳🇴 (2.4%) 🇲🇽 (2.3%)
1-2 🇧🇪 (1.9%) 🇨🇿 (1.7%) 🇿🇦 (1.7%) 🇷🇺 (1.6%) 🇵🇱 (1.3%) 🇭🇺 (1.3%) 🇧🇲 (1.3%) 🇨🇱 (1.1%) 🇮🇪 (1.0%) 🇧🇷 (1.7%) 🏳️ (1.4%) 🇦🇹 (1.1%) 🇸🇰 (1.1%) 🇮🇪 (1.0%) 🇮🇱 (1.0%)
0-1 🇵🇹 (0.9%) 🇸🇪 (0.7%) 🇩🇰 (0.7%) 🇱🇺 (0.4%) 🇺🇦 (0.4%) 🇸🇮 (0.4%) 🇳🇴 (0.4%) 🇪🇪 (0.1%) 🇰🇿 (0.7%) 🇦🇿 (0.7%) 🇩🇰 (0.7%) 🇨🇿 (0.6%) 🇺🇦 (0.3%) 🇨🇷 (0.3%) 🇵🇷 (0.3%) 🇧🇧 (0.3%) 🇳🇱 (0.3%) 🇦🇷 (0.3%) 🇱🇺 (0.3%) 🇵🇱 (0.1%) 🇨🇴 (0.1%) 🇲🇦 (0.1%) 🇨🇱 (0.1%)

🎫 Olympics qualification

The Olympics competition has 55 spots, with a limit of 3 athletes per nations.

Considering nations with a high percentage of athletes in top-50, let's estimate how many athletes miss the Olympics because of the 'max-3-rule':

  • Women:
    • 🇺🇸 : ~11.0% => ~3.0 top-50 athletes rejected. 😞
    • 🇬🇧 : ~10.9% => ~3.0 top-50 athletes rejected. 😞
    • 🇦🇺 : ~10.1% => ~2.6 top-50 athletes rejected. 😞
    • 🇯🇵 : ~8.0% => ~1.4 top-50 athletes rejected. 😞
    • 🇩🇪 : ~7.6% => ~1.2 top-50 athletes rejected. 😞
    • 🇳🇿 : ~5.7% => ~0.1 top-50 athlete rejected. 😞
  • Men:
    • 🇫🇷 : ~9.6% => ~2.3 top-50 athletes rejected. 😞
    • 🇬🇧 : ~8.4% => ~1.6 top-50 athletes rejected. 😞
    • 🇩🇪 : ~8.1% => ~1.5 top-50 athletes rejected. 😞
    • 🇦🇺 : ~7.7% => ~1.2 top-50 athletes rejected. 😞
    • 🇪🇸 : ~7.0% => ~0.9 top-50 athlete rejected. 😞
    • 🇺🇸 : ~6.1% => ~0.4 top-50 athlete rejected. 😞

The estimated numbers of rejections are probably lower bounds:

  • World Triathlon limits the number of athletes per nation for its races too.
    • As a result, some strong athletes, such as US women, cannot participate in important World Triathlon races, and thus do not gain points for the ranking.

The percentages are averages since 2009. Some years, they can be much higher:

  • For instance, 8 women in the top-50 (16.0%) for:
    • 🇺🇸 2015 (a pre-olympics year), with 6 top-30 and the 4th athlete was ranked 16th.
    • 🇺🇸 2016, with 6 top-20.
    • 🇦🇺 2017.
    • 🇬🇧 2021, with 8 top-30.
      • The same year, 🇺🇸 had 5 athletes in the top-12.

It is no wonder that some athletes change nationality to try to qualify for the Olympics.

👶 AGE

ages.png
Age of athletes ranked 5th-9th over years.

Athletes finishing 5th-9th are, on average, 26.3 to 27.7 years old.

  • Ages are similar for women and men.
  • Athletes are slightly older in the olympic format compared to the sprint format: about 1 year difference.
    • I would have expected a larger difference.
  • There are some small variations, but no significant trends over the years.
  • Do Olympics punctuate athletes' careers? The age of women on the olympic format reaches local peaks at the years of the Olympic Games in Rio ( :brazil : ), Tokyo ( :jp : ) and Paris ( :fr : ).

🏁 AGE OF LAST RACE

ages_of_last_race.png
Age of last world-cup, world-series or major games.

The average age of the last race is similar for women and men: around 31 years.

The distribution is broad (~4y std), because there are various reasons for ending a World Triathlon sprint- and olympic-distance top career, such as:

  • Age limit for elite sport.
  • Transitioning to longer-distance triathlons
  • Injury.
  • Personal reasons, such as pregnancy or changing careers.

Some extreme values:

  • <25 years:
    • Hollie Avil ( 🇬🇧 ): 21y, 13 races.
    • Kirsten Nuyes ( 🇳🇱 ): 22y, 21 races.
    • Ellen Pennock ( 🇨🇦 ): 23y, 17 races.
    • Hanna Philippin ( 🇩🇪 ): 24y, 32 races.
    • Marc Austin ( 🇬🇧 ): 24y, 25 races.
    • Sophia Saller ( 🇩🇪 ): 24y, 26 races.
    • Oliver Freeman ( 🇬🇧 ): 24y, 20 races.
    • Raphael Montoya ( 🇫🇷 ): 24y, 21 races.
  • >38 years:
    • Magali Di Marco Messmer ( 🇨🇭 ): 39y, 52 races.
    • Greg Bennett ( 🇦🇺 ): 39y, 70 races.
    • Hunter Kemper ( 🇺🇸 ): 39y, 65 races.
    • Kate Allen ( 🇦🇹 ): 39y, 24 races.
    • Samantha Warriner ( 🇳🇿 ): 42y, 48 races.
    • Kiyomi Niwata ( 🇯🇵 ): 43y, 96 races.
Click to expand - 📜 Full list.
ATHLETE COUNTRY AGE OF LAST RACE NUMBER OF RACES
Hollie Avil 🇬🇧 21 13
Kirsten Nuyes 🇳🇱 22 21
Ellen Pennock 🇨🇦 23 17
Hanna Philippin 🇩🇪 24 32
Marc Austin 🇬🇧 24 25
Sophia Saller 🇩🇪 24 26
Oliver Freeman 🇬🇧 24 20
Raphael Montoya 🇫🇷 24 21
Akane Tsuchihashi 🇯🇵 25 27
Sarissa De Vries 🇳🇱 25 26
Ron Darmon 🇮🇱 25 44
Daniela Ryf 🇨🇭 25 42
Artem Parienko 🇷🇺 26 23
Jose Miguel Perez 🇪🇸 26 30
James Seear 🇦🇺 26 29
Gareth Halverson 🇦🇺 26 17
Lucy Buckingham 🇬🇧 26 44
David McNamee 🇬🇧 26 34
Wian Sullwald 🇿🇦 26 70
Natalie Milne 🇬🇧 27 14
Aaron Harris 🇬🇧 27 27
Franz Löschke 🇩🇪 27 36
Andrey Bryukhankov 🇷🇺 27 36
Paul Tichelaar 🇨🇦 27 33
Felicity Abram 🇦🇺 27 46
Mariya Shorets 🇷🇺 27 47
Sebastian Rank 🇩🇪 27 28
Matthew Sharp 🇬🇧 27 22
Jenna Parker 🇺🇸 28 21
Denis Vasiliev 🇷🇺 28 29
Benjamin Shaw 🇮🇪 28 50
Sarah-Anne Brault 🇨🇦 28 32
Agnieszka Jerzyk 🇵🇱 28 52
Andrew Yorke 🇨🇦 28 42
Maaike Caelers 🇳🇱 28 56
Tamara Gomez Garrido 🇪🇸 28 32
Ivan Tutukin 🇰🇿 28 34
Rebecca Robisch 🇩🇪 28 46
Kathrin Muller 🇩🇪 28 39
Kirsten Sweetland 🇨🇦 28 49
Paula Findlay 🇨🇦 28 41
Peter Croes 🇧🇪 28 58
Kaitlin Donner 🇺🇸 28 37
Jason Wilson 🇧🇧 28 45
William Clarke 🇬🇧 28 44
Jodie Swallow 🇬🇧 29 32
Vanessa Raw 🇬🇧 29 25
Emmie Charayron 🇫🇷 29 43
Katie Hewison 🇬🇧 29 16
Jillian Elliott 🇺🇸 29 31
Andreas Giglmayr 🇦🇹 29 51
Radka Kahlefeldt 🇨🇿 29 31
Danne Boterenbrood 🇳🇱 29 20
Charlotte Bonin 🇮🇹 29 62
Svenja Bazlen 🇩🇪 29 28
Elizabeth May 🇱🇺 29 55
Annabel Luxford 🇦🇺 29 52
Anastasia Abrosimova 🇷🇺 29 44
Cameron Good 🇦🇺 30 33
Mari Rabie 🇿🇦 30 50
Brendan Sexton 🇦🇺 30 57
Kathy Tremblay 🇨🇦 30 54
Jodie Stimpson 🇬🇧 30 71
Laurent Vidal 🇫🇷 30 61
Yulian Malyshev 🇷🇺 30 33
Mark Buckingham 🇬🇧 30 22
Gregor Buchholz 🇩🇪 30 59
Debbie Tanner 🇳🇿 30 55
Pamella Oliveira 🇧🇷 30 62
India Lee 🇬🇧 31 18
Valentin Mechsheryakov 🇰🇿 31 51
Lauren Groves 🇨🇦 31 57
Christian Prochnow 🇩🇪 31 41
Matt Chrabot 🇺🇸 31 43
Melanie Hauss 🇨🇭 31 45
Clark Ellice 🇳🇿 31 53
Gavin Noble 🇮🇪 31 38
Tony Dodds 🇳🇿 31 58
Emma Snowsill 🇦🇺 31 44
Miguel Arraiolos 🇵🇹 31 72
Ricarda Lisk 🇩🇪 31 63
Line Jensen 🇩🇰 31 20
Ruedi Wild 🇨🇭 31 55
Kate Roberts 🇿🇦 31 56
Helle Frederiksen 🇩🇰 31 28
Joe Maloy 🇺🇸 31 37
Aurelien Raphael 🇫🇷 31 59
Dan Wilson 🇦🇺 31 59
Annamaria Mazzetti 🇮🇹 31 77
Rebecca Spence 🇳🇿 31 39
Simon De Cuyper 🇧🇪 32 55
David Hauss 🇫🇷 32 54
Erin Densham 🇦🇺 32 64
Premysl Svarc 🇨🇿 32 74
Yurie Kato 🇯🇵 32 52
Helen Jenkins 🇬🇧 32 55
Emma Moffatt 🇦🇺 32 76
Jan Frodeno 🇩🇪 32 50
Leonardo Chacon 🇨🇷 32 86
Kyle Jones 🇨🇦 32 90
Alexander Bryukhankov 🇷🇺 32 86
Brent McMahon 🇨🇦 32 71
Manuel Huerta 🇵🇷 32 60
Liz Blatchford 🇬🇧 32 62
Mark Fretta 🇺🇸 32 70
Ivan Vasiliev 🇷🇺 32 81
Nicky Samuels 🇳🇿 33 76
Brad Kahlefeldt 🇦🇺 33 78
Bruno Pais 🇵🇹 33 64
Lisa Norden 🇸🇪 33 67
Kate Mcilroy 🇳🇿 33 39
Jonathan Zipf 🇩🇪 33 46
Mariko Adachi 🇯🇵 33 61
Tomoko Sonoda 🇯🇵 33 34
Irina Abysova 🇷🇺 33 48
Anne Haug 🇩🇪 33 45
Gonzalo Raul Tellechea 🇦🇷 33 51
Frederic Belaubre 🇫🇷 33 44
Mary Beth Ellis 🇺🇸 33 24
Lindsey Jerdonek 🇺🇸 33 38
Margit Vanek 🇭🇺 33 59
Misato Takagi 🇯🇵 33 31
Carole Peon 🇫🇷 34 55
Claude Eksteen 🇿🇦 34 25
Marina Damlaimcourt 🇪🇸 34 53
Thomas Springer 🇦🇹 34 41
Aileen Reid 🇮🇪 34 60
Danylo Sapunov 🇺🇦 34 78
Jarrod Shoemaker 🇺🇸 34 93
Tim Don 🇬🇧 34 73
Christiane Pilz 🇩🇪 34 40
Grégory Rouault 🇺🇸 34 19
Katrien Verstuyft 🇧🇪 34 53
Zuriñe Rodriguez Sanchez 🇪🇸 34 63
Sarah Haskins 🇺🇸 34 41
Maik Petzold 🇩🇪 34 61
Felicity Sheedy-Ryan 🇦🇺 34 50
Hirokatsu Tayama 🇯🇵 35 87
Lisa Mensink 🇨🇦 35 42
Kris Gemmell 🇳🇿 35 84
Sven Riederer 🇨🇭 35 87
Bevan Docherty 🇳🇿 35 78
Andy Potts 🇺🇸 35 32
Zita Szabó 🇭🇺 35 44
Adam Bowden 🇬🇧 35 49
Jessica Harrison 🇫🇷 36 81
Cedric Fleureton 🇫🇷 36 36
Mateja Šimic 🇸🇮 36 47
Bryan Keane 🇮🇪 36 45
Marek Jaskolka 🇵🇱 36 49
Sarah True 🇺🇸 36 85
Kerry Lang 🇬🇧 36 43
Vladimir Turbayevskiy 🇷🇺 36 67
Tony Moulai 🇫🇷 37 58
Diogo Sclebin 🇧🇷 37 81
Laura Bennett 🇺🇸 37 68
Courtney Atkinson 🇦🇺 37 68
Reto Hug 🇨🇭 37 67
Simon Whitfield 🇨🇦 37 78
Ryosuke Yamamoto 🇯🇵 37 82
Ainhoa Murua Zubizarreta 🇪🇸 38 100
Magali Di Marco Messmer 🇨🇭 39 52
Greg Bennett 🇦🇺 39 70
Hunter Kemper 🇺🇸 39 65
Kate Allen 🇦🇹 39 24
Samantha Warriner 🇳🇿 42 48
Kiyomi Niwata 🇯🇵 43 96

📅 MONTH OF BIRTH

For this section and the next one about Body mass index, a larger dataset is used:

  • All athletes part of a non-para ranking, and aged between 15 and 45 years, are considered.
    • The ranking categories defined by World Triathlon can be found here.
  • Leading to a total of 3,439 unique athletes, including 1,382 women and 2,057 men.

The month-of-birth distribution of the athletes is compared to two reference distributions:

  • Reference #1: Uniform month-of-birth distribution.
    • It could be expected that each month accounts for 1/12 = 8.3% of the births.
  • Reference #2: Birth data collected by the United Nation: data.un.org.
    • Birth data of people aged between 20 and 30 years are considered, leading to over 230 million entries.
    • It could be expected that World Triathlon (formerly ITU) and UN month-of-birth distributions match.
Month-of-birth of World Triathlon (formerly ITU) athletes.

The results are even more striking when considering the year quarters.
Year-quarter-of-birth of World Triathlon (formerly ITU) athletes.
Click to expand - 👫 Same plot with gender comparison.
Year-quarter-of-birth of World Triathlon (formerly ITU) athletes.
Click to expand - 👶 Same plot for the Junior categories.

The average age is 19.5 years.
Year-quarter-of-birth of World Triathlon (formerly ITU) athletes in the Junior categories.
Click to expand - 📊 Reference distribution from the United Nations.
Month-of-birth distribution of the population aged 20-30 recorded by UN. Data source: data.un.org

Can these discrepancies be due to differences between the two datasets (ITU and UN), such as the geographical origin of the births?

Click to expand - 🌍 Continent analysis.

The continent distributions differ between the World Triathlon and the UN datasets:

  • ITU dataset: Europe is predominant (~59%). Asia (~13%) and North America (~11%) follow.
  • UN dataset: Asia (~34%), Europe (29%) and North America (24%) form a more uniform top-3.
Continents distribution of the World Triathlon (formerly ITU) dataset
Continents distribution of the UN dataset (used as reference).

Note: The UN dataset has probably missed data from African and Asian countries such as China, India and Nigeria.
Another visualization of the reference month-of-birth distribution.

The month-of-birth and quarter-of-birth distributions for each continent can be visualized:
Month-of-birth distribution, by continent (normalized).
Quarter-of-birth distribution, by continent (normalized).

Conclusion:

  • The continents mainly represented in the two datasets (ITU and UN) share very similar month-of-birth and quarter-of-birth distributions.
  • Therefore, the difference in continent distributions does not explain the discrepancy in month-of-birth and quarter-of-birth distributions between the ITU and UN datasets.
Click to expand - 🎂 Age distribution of World Triathlon athletes used for this analysis.

The UN and ITU datasets share the same average age (25), but the ITU age distribution is not uniform, unlike the UN one.

  • Importance sampling could be applied to the UN dataset to make the two distributions match, but I would be very surprised if that had an impact on the overall conclusion.
age.png
Age of athletes considered for the month-of-birth analysis.

Is the difference between ITU and UN month- and quarter-of-birth distributions statistically significant?

Click to expand - 🧮 Statistical test.

3,439 birth entries have been collected ("observed") from the World Triathlon (formerly ITU) data.

  • A priori there is no link between the quarter-of-birth and the fact of being a high-level triathlete.
  • Therefore, it can be assumed that the ITU observations follow the UN distribution.

From the UN distribution, the expected number of births for each quarter is computed:

OBSERVED (ITU) EXPECTED (UN)
Q1 987 854.443
Q2 859 845.248
Q3 820 893.717
Q4 773 845.592

The number of births in the two columns look very different: more observations than expected for Q1, fewer for Q3 and Q4.

  • The situation is similar to the case of a 6-sided dice, which is suspected to be biased. 🎲
    • After rolling the die many times and counting each outcome, one can ask:
    • Is just due to randomness or is it because the dice is unfair?

Back to our problem:

  • How to quantify the deviation between the two columns?
OBSERVED (ITU) EXPECTED (UN) DIFF DIFF^2 DIFF^2 / EXPECTED
Q1 987 854.443 132.557 17571.3 20.5646
Q2 859 845.248 13.7516 189.106 0.223728
Q3 820 893.717 -73.7166 5434.14 6.08039
Q4 773 845.592 -72.5916 5269.54 6.23178

One can compute the differences between columns, square them to ensure they are positive, normalize them and sum the results:

  • SUM OF [DIFF^2 / EXPECTED] = 20.5646 + 0.2237 + 6.0804 + 6.2318 = 33.10
  • It 33.10 large? What does it mean? What can be concluded?

There is a mathematical formula that, given this computed number (33.10), answers the following question:

  • What is the probability of observing such a discrepancy (33.10) or an even larger one, assuming that the ITU data should follow the UN quarter-of-birth distribution?
  • In other words: How likely is it that the observed deviation is due to random chance?

For 33.10, the formula gives p = 0.0000003, i.e. 0.00003%.

Conclusion:

  • The extremely low probability (0.00003%) indicates that the observed differences in quarters-of-birth among World Triathlon athletes are highly unlikely to be due to random chance.
  • Therefore, the observed differences are statistically significant.
  • This suggests a systematic deviation from the expected UN distribution.
  • In other words, the quarters-of-birth of high-level triathletes do not align with the general population as represented by the UN distribution.

For more details about the derivation:

⚠️ These findings do NOT indicate that "World Triathlon athletes born in January are more performant than others born in later months"!

  • This birth-of-month analysis rather shows that someone born earlier in one year is MORE LIKELY TO BECOME PRO TRIATHLETE than one born later in the same year.

Here is one possible explanation:

  • Kids born on January, 1st and December 31st of the same year compete in the same age category.
  • For example, at the age of 12, a 12-month difference represents 10% of their lifetime.
  • A 12-year kid born in January could, in theory, be ~10% physically more developed (in terms of strength and stamina) than one born in December.
  • This edge could give an advantage to kids and teenagers born in the first months of a year during their early races in youth categories.
  • They are more likely to perform well, stand out, gain experience, and get selected for international competitions, eventually becoming professional. 🏆
    • Similar to a snowball effect. 🌀
  • This phenomenon, known as relative age effect, was also observed on young Spanish triathletes by Ferriz et al. in 2020.

🏋️ BODY MASS INDEX

"The BMI is defined as the body mass divided by the square of the body height, and is expressed in units of kg/m^2"

To maximize the amount of data, all athletes registered with World Triathlon are considered (not just those participating in world-cups and world-series).

  • In total, 504 athletes aged between 15 and 53 (average is 27) are included.
bmi.png
Body Mass Index.
weight_height.png
Weights and Heights.

First, is the BMI a relevant metric here? No!

  • It is useful for public health, providing a quick and simple measure to identify trends in obesity and underweight conditions.
  • But it does not measure fitness levels, physical endurance, strength, flexibility, or other aspects of physical health.
  • Many athletes may have a high BMI due to increased muscle mass, yet they are fit and healthy.

Second, there is no enough data.

  • Only 14% of athletes registered with World Triathlon have valid weight and height information (504 out of 3560).
  • Most athletes did not input their dimensions.
  • Sometimes for privacy reasons, e.g. the entry for Kristian Blummenfelt's ( 🇳🇴 ) weight: "None of your business".

Third, some data may be outdated.

  • The body weight that can vary during a career (many athletes enter the database as juniors) or even during a season.

Alternative metrics such as Body Fat Percentage or Muscle Mass Percentage would be more appropriate for fitness assessment.

If not much can be concluded, comparing oneself to the distributions can still be interesting.

Click to expand - 📏 Non-standardized data format!

Here are examples of height data. 📏

 '6 Foot',
 "194, 6'4",
 '1.76 mts',
 '62 kg ',
 '1.80 meters',
 '174m',
 '5\'2"',
 '165 cms',
 '1,81 CM',
 '6ft. 2in.',
 '6ft 2in',
 '5’7”',
 '6 ft 0 in',
 '170cm/5.58ft',
 '57kg',
 '5.5ft',
 '1,77mts',
 'MT. 175',
 '164,4cm',
 '1.66 MTS',
 '5\'8"',
 '183 cm',
 "6'4",
 '1.9m',
 '6 Ft',
 "1'72",
 '186cm',
 '182',
 '180 ',
 '182CM'

And some weight data. 🏋️

 'None of your business ',
 '160',
 '70 kg “in season “ ;-)',
 '1:68 m ',
 '140 lb',
 '51kgs ',
 '138 LBs',
 '1,78',
 '75 KG',
 '135ibs',
 '54kg/119lb',
 '42 kilo',
 '135IIbs',
 '187 lbs',
 'KG. 62',
 '49,4 kg',
 '67KG',
 '62 Kg',
 '175',
 '145lb',
 '52kg',
 '67',
 '140 Ibs',
 '65',
 '75 ',
 '66KG'

Interesting to see the different units and formats.

Some processing and filtering are required! 🧹

📜 CONCLUSION

💡 IDEAS

Here are some ideas for data to explore:

🔍 1) MORE DETAILED RACES ANALYSES

As mentioned by Alex Yee in this video:

"You are doing a full gas effort at the start of a race which is 2 hours long. If you told somebody to do that on a marathon, they would laugh in your face."

World Triathlon data provides a single time for each leg: swim, t1, bike, t2, run. There is no detail about the pace evolution during each segment. For instance:

  • Fast swim start to reach the first buoy.
  • Fast bike start to break away or catch a pack.
  • Fast run start - I have never really understood the benefit compared to a steady effort.
  • Fast run finish.

Some events, such as Paris Olympics, offer detailed results with lap times.

  • It is for instance very instructive to compare the runs of Alex Yee ( 🇬🇧 ), Hayden Wilde ( 🇳🇿 ) and Léo Bergère ( 🇫🇷 ).
  • The first two completed the first 1050m in 2:42.
  • Alex Yee then "slowed down" to ~3:00 for the following three laps, while Léo Bergère was a bit more regular: 2:49, 2:56, 3:01, 2:54.
  • Such data should be very interesting to analyse.

In addition to paces, it would be interesting to access data such as:

  • Swim stroke rate.
  • Bike power and cadence.
  • Run cadence.
  • Run maximum speed during the sprint.
  • Heart rate.
  • ...
  • For comparison, heart rate information from some athletes is shown in UCI mountain-bike world-cups videos.

Activity trackers would be needed for these recordings, but athletes rarely wear them while racing, making swim and run data recording difficult.

Strava could be used to retrieve activities:

  • Many athletes publish their activities there, either their races (often limited to the bike section) or training sessions.
  • However, as mentioned, almost no swim or run information is recorded during races.

🛍️ 2) EQUIPMENT CHOICE

What is the best wetsuit? What are the fastest running shoes?

  • There are already some tests and reports conducted by researchers.
  • However, I believe that examining athletes' preferences would provide more reliable and valuable results.

Not all athletes are sponsored by swimming or running brands.

  • Therefore, many have the freedom to experiment, compare, and choose the equipment they believe will enhance their performance.
  • As an example, one could track how many athletes wore Asics, Adidas, Nike, New Balance, etc., at major events, noting the models used, and exclude those provided by sponsors.

I recall reading about a similar project but can no longer find the reference.

💰 3) MONEY

It would be interesting to investigate the financial aspects of the competitions, such as:

  • The prize money for the different World Triathlon race categories.
  • The event registration costs.
  • Eventually, to estimate from which rank an athlete can make a descent living. (Of course, sponsoring and federation support also play a role).

🍝 4) MISCELLANEOUS

  • Analyse the correlation between transition ranking and finish ranking.
    • Especially for T2.
  • Compare WTCS and world-cups more thoroughly.
  • The arbitrary decision to focus on the top 5-9 was made to capture a stable and consistent representation of the general competitive field.
    • However, examining the top performance (e.g. top-1 or top-3) or using a broader range could also yield valuable insights.
  • Conduct advanced analyses of cycling performances would be interesting.
    • Additional data may be required: drawing conclusions based solely on bike split times is challenging, due to the influence of drafting and pack dynamics.
  • Investigate the impact of swim conditions on swim performance and race dynamics.
    • Including water temperature, presence of waves, and salinity.
  • Take a closer look at the critical start of the bike segment.
    • For example: Given a lag at T1, how likely is it to catch the first group?
  • Understand the system of penalties and their impact on race dynamics.
    • The 2024 Paris Olympics ( 🇫🇷 ) saw a surprising number of penalties: 6 women out of 51 and 10 men out of 50 received a 15s penalty.
    • Before focusing on gaining seconds, some athletes may need to prioritize avoiding penalties, such as by correctly timing their dive and making sure to place their helmet inside the transition box.
  • Analyze the trajectory of successful elite athletes.
    • How did they perform as juniors, and how did they progress from junior to U23 to elite?
    • One could also ask: One junior athlete excels in swimming, another one in running. Which is more likely to become an Elite and perform in this category?
  • Try to estimate the level of a race.
    • Propose a formula based on metrics such as the ranking of participating athletes and, if available, their results in this race.
      • Extensions include evaluation of individual race segments (swim, bike, run), and athlete scoring.
    • Potential pre-race and post-race applications:
      • What performance objectives should the coach set, considering the expected race difficulty?
      • Should a federation select an athlete for this event based on the competitiveness of the start list?
      • How strong are the best swimmers for this race? Who would be the optimal neighbours on the start pontoon?
      • How strong was this top-10 finish, given the competition level? Did an athlete overperform or underperform in a specific leg?
      • What are the most hotly contested races? With a formula like this, I could classify the races by level and keep only the e.g. 100 most disputed for my analyses.
  • ...

💻 CODE

The python code to fetch the data, set the parameters and generate plots is available in this GitHub repository. There are two ways to use it:

  • Either locally, if you are familiar with python. 🐍
    • Clone the repo and install the required python packages.
    • Create a key for the World Triathlon API and add it to a api_key.txt file to save in tri_stats/.
    • Run the different scripts of tri_stats/scripts.
  • Or in the cloud, using free tools: Google Colab and Google Drive. ✌️

🥡 TAKEAWAYS

⚠️ Most summary numbers given in this section are AVERAGES.

  • For instance race-averages are computed from the 5th to 9th best times of each leg, and sometimes averaged over multiples years.

Here are some simplified key takeaways:

  • ⏱️ The three sports account for 16.4% ( 🏊 ), 53.1% ( 🚴 ), 28.9% ( 🏃 ) of the overall time. Transitions for 1.1% and 0.5%. [link]
  • 👟 While swim and bike paces are similar between sprint and olympic formats, the 10k run requires 7 s/km more than the 5k. [link]
  • 🏊 Women swim at 1:18 / 100m, men at 1:12 / 100m. [link]
  • 🚴 Women ride 4 km/h slower, at 37.4 km/h, compared to men at 41.4 km/h. [link]
  • 🏃 Women run the 10k at 3:33 min/km (3:26 for 5k), men at 3:07 min/km (3:00 for 5k). [link]
  • 🚿 The time charged to the wetsuit during the T1 transition is ~9s. [link]
  • 👫 Women swim 8.8% slower than men with the same equipment. They also ride 10.6% and run 14.2% slower. [link]
  • 📉 The women/men difference has not significantly reduced on the years, except for the run leg of the sprint-format races (-0.13 % / year) and for the swim of WTCS (-11 % / year). [link]
  • 🐧 There is no evidence that wetsuits reduce swim gaps between top and less competitive swimmers. [link]
  • 🩱 Swim is 4-5% faster with wetsuit. [link]
  • 🇫🇷 The swim of 2024 Paris Olympics was unusually long (more than 2:30 longer), probably because of the current in La Seine. In particular, the 5-9th women swam more than 1:30 / 100m. [link]
  • ⚡ Winning by a run comeback, i.e. after not ending the bike in the front group, is entertaining but rare in the olympic format (28% for men and 7% for women) and is getting even rarer. [link]
  • 🚴 The size of the front group after bike averages around 15. It decreases to 4 or fewer (small breakaway) in about 1/4 of women's and 1/3 of men's olympic-races. [link]
  • 👟 Over 2/3 of races are won by the best runner. [link]
  • 📸 In men's races, 17% (sprint format) and 10% (olympic) are won by a sprint finish, occurring 50% more often than in women's races. [link]
  • 🤸‍♀️ Women's races occasionally feature wins by very large margins. [link]
  • 📏 The gaps between the winner and the second are, on average, twice as large in olympic formats compared to sprint formats, and twice as large for women compared to men. [link]
  • 🚀 Bike and run times in WTCS olympic races have reached all-time lows. [link]
  • 🐧 The wetsuit is allowed in ~1/3 of races. More often for women (37%) than for men (32%). [link]
  • 🌡️ Athletes must compete in a variety of conditions: Notably, 80% of the recorded air temperatures (lower estimates) fall between 17°C and 30°C and almost uniformly so. [link]
  • 🥵 Heat tends to slow down running pace. [link]
  • 📆 On average, athletes raced 10 times (world cups and WTCS) in 2019 and 2023, compared to 6 times in 2009. [link]
  • 📆 Their World Triathlon sprint- and olympic-distance season has extended from 130 days in 2009 to 200 days in 2023. [link]
  • 🎫 The limit of 3 athletes per nation for the Olympics creates challenges for the highly represented nations such as 🇺🇸, 🇬🇧, 🇦🇺, 🇩🇪 and 🇫🇷. [link]
  • 🎂 Athletes finishing 5th-9th are, on average, between 26 and 28 years old. [link]
  • 🏁 Women and men typically race their last world-cup or WTCS at an average age of 31 years, thought there are significant variations. [link]
  • 📊 Someone born earlier in the year is more likely to become professional triathletes compared to those born later in the same year. [link]

Thank you for reading until the end!

  • If you have any questions, suggestions, corrections, or comments, please feel free to contact me at simon.chauvin.contact[at]gmail.com.
  • Cheers,
  • Simon 😃