FIGHTintelligence
Machine-learning MMA predictions by Vincent Lugat
i
Betting strategy
  • We bet on fighters with edge > 0% and win probability above the threshold set below
  • Stake is 10% of bankroll per bet, capped at the max bet cap set below
1000
2500
5000
10000
25000
50000
100000
i
Max bet cap
  • Default is 5,000€, the real liquidity limit for a single bet on Pinnacle before it starts moving the market
  • You can raise it to explore hypothetical scenarios, but real execution above this size gets harder
500
1000
2500
5000
10000
20000
i
Min win probability
  • Floor is 50%. The underlying data only includes bets already filtered to proba > 50% at the source, so a lower threshold has no effect
%
50%
55%
60%
65%
70%
75%
80%
85%
90%
Walk-forward backtest: retrained monthly, predicting the following month iteratively. Train 2012-2021, Test 2022-01, then Train 2012 to 2022-01, Test 2022-02, and so on (expanding window, one month at a time).

Bankroll Record

Monthly Performance

Model probabilities are evaluated against Pinnacle's margin-adjusted closing probabilities (The Odds API).

Model Calibration

Upcoming Event
Past Events