top of page

TAKE YOUR PREDICTION GAME TO THE NEXT LEVEL

See the story behind every prediction.

Know in advance which predictions are most reliable.

Maximize your team's competitive edge.

Unlike any other prediction system, relevance-based prediction pinpoints the unique set of data that matters most for each task.

CSA Prediction Vault

Next-level predictive analytics tailored for your needs

 

Scan a high-level overview of predictions, dive into the components of every individual prediction, explore detailed player profiles, and more.

csa_vault_2.png
csa_vault_1.png

Make informed decisions with confidence

​Relevance-Based Prediction finds the nuanced patterns that matter.

 

And it is completely transparent, unlike AI and machine learning, so you can see which prior players and variables drive each individual prediction.

CSA Prediction Engine

Integrate relevance-based prediction in your own data process

Our flexible API gives you instant access to every aspect of relevance, including optimized grid prediction, task-specific fit, variable importance, similarity and informativeness breakdowns, and much more. 

Every aspect of prediction is customizable, computationally efficient, and robust to common challenges such as missing data that confound traditional methods. 

API Highlights

 

A robust, powerful, and versatile prediction and analytics engine that interacts with any dataset of predictive variables and outcomes.

  • Parallel computing for unmatched scalability​

  • Advanced engineering for complexity and high dimensionality

  • Full transparency and interpretability with detailed outputs​

  • Robust treatment of missing data

Basketball Hoop

HOW RELEVANCE-BASED PREDICTION WORKS

Relevance-based prediction (RBP) is an innovative model-free forecasting routine that overcomes the limitations of both classical prediction models and machine learning.

 

RBP forms predictions as weighted averages of past observations, where the weights are determined using fundamental principles of information theory to gauge which observations and predictive variables are most important for each individual prediction task.

 

By considering thousands of combinations of observations and variables for each prediction task, RBP extracts as much information from complex datasets as machine learning models. But unlike machine learning models, which rely on uninterpretable parameters, RBP reveals precisely how each observation and each variable influences the prediction. 

Key advantages

  • RBP addresses complex relationships that are beyond the reach of linear regression.

  • RBP is transparent and adaptive, unlike machine learning models.

  • RBP identifies the optimal blend of observations and predictive variables for each individual prediction task. 

  • RBP anticipates task-specific reliability for each individual prediction. 

  • RBP guards against overfitting and offers protection against data errors.

  • RBP explains variable importance for each task, improving on conventional t-statistics and Shapley values.

  • RBP is theoretically justified by information theory and other fundamental principles. 

  • RBP is robust to missing data and unforeseen circumstances. 

Prediction Revisited

 

The foundational book from Megan Czasonis, Mark Kritzman, and David Turkington introduces the principles of relevance-based prediction and reveals a new perspective on data analytics. 

2022 Harry M. Markowitz Award

For best article in the Journal of Investment Management. Final award selected by a panel of Nobel Prize laureates in economics.​​​

Black Fork Ribbon_edited_edited.png

2023 Roger F. Murray First Prize

For best presentation at the Institute for Quantitative Research in Finance.​​​

Black Fork Ribbon_edited_edited.png

Awards and Recognition

Relevance-based prediction has received prestigious awards for its application in economics and financial markets. 

Get in touch

© 2024 Cambridge Sports Analytics. All rights reserved.

bottom of page