About the author
Marco Wen
Marco Wen is the data engineer and algorithm engineer at Mysports.AI, responsible for developing the NBA and MLB prediction models. His work spans data cleaning, feature engineering, model training, validation and monitoring. He builds the models on probabilistic and statistical methods, with a focus on model stability, data quality and prediction error, and on identifying where the model's estimates diverge from market data. Model output is provided for information only and does not constitute betting advice or a promise of profit.
Areas of expertise
NBA and MLB model data pipelines
Builds the data pipelines for both leagues, from collection and cleaning through to storage, so that model training and daily predictions run on consistent, traceable data.
Feature engineering and model training
Converts pace, offensive and defensive efficiency, injuries and schedule information into features the model can use, and owns model training, version control and updates.
Validation, monitoring and error checks
Checks model performance through backtesting and day-by-day live records, tracking prediction error and data quality. Model versions and results are published on the model performance page.
How the work is done
How the models are built
The architecture, features and validation of the NBA model are documented on the NBA AI Prediction Model and AI Algorithm pages and are not repeated here.
Updating on market information
The model recalculates as lines and odds move and folds pre-match information such as injury news into the update. When a key player is confirmed out, for example, the risk on the affected games is reassessed.
Cross-team collaboration and content responsibility
Marco Wen works with the analysts and the content team to confirm that the model output shown on each page matches the actual calculation. His remit is model development and data engineering; he does not give betting advice, and the page copy is written by the Mysports.AI data science team.
Responsibility and disclosure
The author on this page is responsible for data analysis and model-related work. All content is provided for information and entertainment only and does not constitute betting advice or a promise of profit. Every model pick and its result is published in the prediction record, with hits and misses kept together.