I’ve made huge changes in the design of the model. From here on out, we’re including the odds and we’re training on Google Vertex AI, not manually. Second, we added some new features. Third, we’re now doing a parlay strat of 2-3 legs on somewhat random assortment of the picks. So far, in the past […]
MMA-AI.net v3.3 – Google vertex AI Read More »
Been hard at work rewriting my model from scratch. The basis of the rewrite is brand new scrapers that run faster and more accurately, removing all my years of accumulated technical debt, and test driven development from the beginning to so I can be sure I’m not making small mistakes anywhere in the data or
MMA-AI v3.0: Coming soon! Read More »
Did I announce MMA-AI v2.0 months ago? Yes. Did I announce that too early? Sure did. Originally, my intention was to release MMA-AI 2.0 with a model that was trained on profitability rather than accuracy. I discussed this idea with Chris from wolftickets.ai and he got a stellar implementation working almost immediately which is available
MMA-AI 2.0: We’re back with dramatic profitability improvement Read More »
TPOT was the original library I used for algorithm and hyperparameter tuning but I was advised there’re probably better ways of doing it by my friend Mike Conover who suggested weights and biases or Optuna. I ended up using Optuna since it was fast so I did 2,000 runs. import optunaimport xgboost as xgbfrom sklearn.metrics
version 1.3: Optuna tuned hyperparameters Read More »
