Researchers have developed a new Isotonic Bradley-Terry model to improve the analysis of paired comparison data, such as predicting match outcomes and ranking participants. This novel model addresses potential misspecification issues in existing Bradley-Terry and Thurstone-Mosteller models by learning both the rate parameters and the inverse link function through an alternating gradient and isotonic regression approach. The proposed method ensures monotonic improvement in training error and is designed to handle insufficient data by allowing for exact ties. Numerical experiments on synthetic data and real-world datasets from football, baseball, and tennis leagues demonstrated enhanced win probability prediction and ranking performance. AI
IMPACT Introduces a novel statistical modeling technique that could improve prediction and ranking in various domains using comparative data.
RANK_REASON The cluster contains a research paper detailing a new statistical model for paired comparison data, submitted to arXiv. [lever_c_demoted from research: ic=1 ai=1.0]
- alphaXiv
- arXiv
- ATP Tour
- Bradley--Terry model
- CatalyzeX
- DagsHub
- Gotit.pub
- Hugging Face
- IArxiv
- Major League Baseball
- Premier League
- ScienceCast
- Thurstone-Mosteller
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
