AI Insights; New Model (win_chance v0.0.1) Released

Here at Betacus, we are always striving to uncover hidden value for our users and identify value bets. Traditional handicapping has limits, but Machine Learning (ML) allows us to process large datasets to uncover patterns humans miss. Rather than relying just on intuition, we leverage robust architectures like gradient boosting and neural networks to drive predictive engines.


Unlike Large Language Models (LLMs) that dominate headlines, our models are built for precision. LLMs hallucinate, but our ML models deterministically map inputs to testable outputs. This is far more useful in the world of Race Analysis than an LLM.


Last year, we released a trained model that predicted the outcome of a particular bet strategy, the 1 over 7 exacta. The model was somewhat useful - it was very pessimistic about most bets, but, when it was optimistic, it was a great indicator for that bet strategy.


Today, we are releasing win_chance v0.0.1 and retiring the 1 over 7 predictor for now. (It may be re-released with improvements in the future). This model works differently. It determines a signal for every horse in the race, using raw data and data adjusted for the relative strength of the horse within that race. Notably, it excludes all odds and only relies on underlying figures.


win_chance v0.0.1 signals are normalized across a race to determine a % chance the model thinks each horse will win. This value is compared to the oddsmaker's % chance, and we call the difference the advantage. In this way we can pinpoint where the market is underestimating a horse.


Consider win_chance v0.0.1's insights on your next big bet!