US 12,293,263 B2
System, method, and model structure for using machine learning to predict future sport outcomes based on match state transitions
Kyle Engel, New York, NY (US); Chris Flynn, New York, NY (US); and Henry Sorsky, New York, NY (US)
Assigned to SIMPLEBET LLC, Boston, MA (US)
Filed by SIMPLEBET LLC, Boston, MA (US)
Filed on Feb. 24, 2021, as Appl. No. 17/184,132.
Claims priority of provisional application 63/116,573, filed on Nov. 20, 2020.
Prior Publication US 2022/0164702 A1, May 26, 2022
Int. Cl. G06N 7/01 (2023.01); G06F 17/18 (2006.01); G06N 20/00 (2019.01)
CPC G06N 20/00 (2019.01) [G06F 17/18 (2013.01); G06N 7/01 (2023.01)] 20 Claims
OG exemplary drawing
 
1. A computer-implemented system for predicting future outcomes in a sporting match of a sport of interest configured to process match state transitions using a computing device, the system comprising:
a transition machine learning (ML) model operatively configured to analyze historical match data from a specified sport and generate predicted probability distributions for various match state transitions;
a state updater comprising a meta-ML model configured to refine match state transition predictions by integrating empirical match data, and to output one or more future states based on one or more observed outcomes or trends within the sport;
a final outcome machine learning (ML) model trained on historical match data and configured to receive the one or more future states from the state updater and generate a plurality of probability distributions for specific future match outcomes based on the future states; and
a total probability predictor configured to aggregate the match state transition predictions from the transition ML model, state updater, and final outcome ML model utilizing a Bayesian network framework to synthesize and output refined probability distributions for desired match outcomes;
wherein the system is configured to execute real-time data analysis and model adjustments, dynamically responding to live match data to update predictive models, adjusting the match state transition predictions based on current match conditions and player performances,
wherein the meta-ML model is configured to process empirical data for future state extrapolation, integrating the empirical data into the Bayesian network to increase precision of predictive outcomes, and
wherein the computing device is configured to execute a series of steps, the steps comprising:
inputting an initial match state So of the sporting match into the transition machine learning model;
generating, using the transition machine learning model, predicted probability distributions on a plurality of transition outcomes PT0-PTi where i is an integer;
inputting the plurality of transition outcomes PTi into the state updater;
generating, using the state updater, a plurality of predicted probability distributions on future states, S1-Si, where i is an integer, conditioned on each possible transition outcome, PTi;
inputting the plurality of predicted probability distributions on future states Si into the final outcome machine learning model;
generating, using the final outcome machine learning model, predicted probability distributions on a desired final outcome, PF;
inputting PTi, Si, and PF into the total probability predictor; and
generating, using the total probability predictor, parameters for a Bayesian network to produce a probability distribution of a desired outcome occurring in a future match state, the generating comprising synthesizing the aggregated match state transition predictions based on association of various match state transitions, enabling the computing device to integrate probabilistic relationships among the predicted probability distributions of transition outcomes, future states, and final outcomes, to output a prediction of match developments.