Description
Problem: Right now, it looks like the lrn_weight is a single number for all learning cards. However, the probability of getting a card right is partially dependent on which learning step the learn card is at.
Solution: There should be a separate lrn_weight for each learning step. Instead of grabbing the total number of learning cards and multiplying by the weight, it should grab the number of learning cards at each step and have a separate weight for each.
Further, this line of thinking can be applied to review cards as well. We can differentiate young weights (interval <21 days) from mature weights (interval >=21 days <100 days) from supermature weights (interval >=100 days). Doing this with review cards might require a separate larger no_days variable