NBA Player Proposition Markets: A Two-Tier Machine Learning System with Five-Month Out-of-Sample Validation

Read full text Summarise with VIRAT
Authors: Edward Glush
Cited by: 0

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

Working paper and open data from Bet Better (betbetter.world), a sports analytics site in Australia. The paper describes a two-tier machine learning system for NBA player proposition markets with five months of out-of-sample validation, graded in public. Companion open datasets (CC BY 4.0): market calibration across 18,238 settled two-outcome markets (https://betbetter.world/studies/market-calibration), the daily bookmaker margin index (https://betbetter.world/studies/margin-index), and predicted scores for every upcoming game in 11 sports as CSV and JSON (https://betbetter.world/predicted-scores). Public record of every graded prediction: https://betbetter.world/results. API spec: https://betbetter.world/api/openapi.json. Open-source clients: https://gitlab.com/betbetterworld. Author: Edward Glush, ORCID 0009-0008-4074-369X.

Fields of Study & Subjects

Sports Analytics and Performance Forecasting Techniques and Applications Gambling Behavior and Treatments

Publication details

Journal / source
OSF Preprints (OSF Preprints)
Publication date
25 September 2026
Language
EN
Document type
Preprint