About
About Fight Prophet
Fight Prophet is a public MMA analytics project built to make fight data easier to inspect. The site combines rankings, fighter cards, model diagnostics, and plain-English model notes so readers can understand what the model sees and where it can be wrong.
Last updated: May 17, 2026
What Fight Prophet Publishes
The public site focuses on original analysis around UFC and MMA data: all-time and division rankings, fighter-card context, prediction logic, historical model audits, and responsible-use notes. The goal is not to copy fight news. The goal is to explain model outputs and make the data behind them easier to question.
Publisher Identity
Fight Prophet is maintained by Uppercut Analytics as an independent analytics and software project. It is not affiliated with, endorsed by, or sponsored by UFC, ESPN, sportsbooks, fighters, or promotions mentioned in the data.
Open Source Roots
The project is developed in public where possible, including the public website and supporting MMA analytics tools. Open-source visibility lets readers inspect the methods instead of treating model claims as a black box.
How The System Works
Fight Prophet starts with structured fight records, fighter profiles, outcomes, method information, weight classes, title context, and manually reviewed overrides for cases such as country normalization, activity status, vacancies, and unusual title situations. Public data can be incomplete, so every output is treated as an estimate that may need correction.
Ranking Logic
Rankings reward wins, opponent quality, finishes, title stakes, defenses, streaks, and durability across time. Losses, inactivity, and weaker context reduce the score. The same engine can be read globally or filtered to one division.
Prediction Logic
Predictions compare model probability with market-implied probability and flag confidence only when the signal clears quality checks. Sparse data, short-notice changes, inactivity, and model disagreement reduce confidence.
Quality Controls
Fight Lab keeps historical picks next to outcomes, Brier score, log loss, calibration, and model disagreement. That audit trail matters because a model can look persuasive on one card and still be poorly calibrated across many cards.
What Makes The Site Useful
- Rankings are computed from fight-level signals instead of editorial ballots.
- Prediction pages explain probability, confidence, and model disagreement.
- Fight Lab keeps historical predictions next to outcomes for auditability.
- Policy pages describe limitations, corrections, privacy, and responsible use.