Tennis match prediction ML model — live odds and results analysi

I designed and trained a machine learning model to predict tennis match outcomes based on historical data and live bookmaker odds.

First, I prepared a structured dataset: historical matches, pre-match and in-play odds, score evolution, tournament, players and final result. Using an AI/AutoML approach, I experimented with different regression and neural-network models and selected the configuration that provided the best prediction quality (win probabilities and their dynamics during the match).

In addition, I developed a separate live parser that, in near real time, fetched odds and intermediate results for ongoing tennis matches, passed these features into the model and received probability estimates for possible outcomes. The whole pipeline ran in a cloud environment: data ingestion, preprocessing, model inference and logging of predictions.

I was responsible for the full cycle:
– data schema and collection logic design;
– ML model training and tuning (AutoML approach, regression and neural networks);
– development of the live data parser for odds and results;
– integrating everything in the cloud and delivering sufficiently accurate predictions for analytics and strategy testing.

Tech stack: ML/AutoML libraries for regression and neural-network models, cloud environment for training and inference, live data parser (tennis matches, odds, results), structured datasets (CSV/table format, database).