im working on a horse betting model but i was wondering something. I focus on quinella betting and on races with at least 12 horses running and so far i use data from 2021 to 2023 for a total of 1920 races.
The attached image gives the final bankroll based on edge and kelly for each year (starting at 5K), the number of tries (a try being a combination where the model predict a return superior to the bet + edge, based on the average return for similar odds), and finally the number of winning combinations.
Do you think i have sufficient data at this point ? i considered the number of tried combinations enough but the fact that the winning combinations are so low concerns me. Should i had more data or is this enough to prove that the model is working ?