Here we show the model's predictions for upcoming fights which are on the vegas books (only those for which both
fighters have fought previously in the UFC).
The predicted winner is shown in gold.
In the final column we show the Kelly bankroll percentages of betting on the fighter or opponent based on the predicted
odds and the DraftKings odds. The number shown is bankroll percentage recommended by the Kelly criterion which maximizes long term profit for binary outcome bets.
In green we show the positive bankroll percentages indicating a likely profitable bet.
Upcoming Fights
Fighter
Opponent
Predicted Fighter Odds
Predicted Opponent Odds
Best Bookie For Bet
Kelly bankroll%
Kelly Criterion Calculator
Here we give the user the capability to input live odds from any bookie (the above odds were grabbed on Wednesday 8pm MST from DraftKings and they have likely changed).
Live odds can be found at fightodds.io
Fighter Predicted Odds
Opponent Predicted Odds
Fighter Vegas Odds
Opponent Vegas Odds
Fighter Kelly Bankroll %
Opponent Kelly Bankroll %
-
-
Past Fights (Recent)
We started running this program to predict fights on April 23, 2022. Below we show the history of predictions
made
before each event.
For reference, the average bookie odds are also shown (this average is formed from the betting odds on
Draftkings,
BetMGM, Caesars, BetRivers, FanDuel, PointsBet, Unibet, Bet365, BetWay, SD, Ref).
After each event, the table is automatically updated to indicate the winners in gold and keep a running total of the prediction accuracy.
Correctly predicted fights are indicated in green and incorrectly predicted
fights are shown in red.
After a large number of fights, the accuracy is expected for theoretical reasons to converge to about 64%.
Fight Prediction History
Fighter vs Opponent
Predicted Odds
Best Bookie
Kelly Criterion Bet
Bankroll After Fight
The following statistics were compiled based on the model's performance
since April 23, 2022
Model Accuracy
Bookie Accuracy
The machine learning model implemented on this page will analyze all known fighter stats (scraped from ufcstats.com)
and predict the most likely outcome if they were to fight exactly as they were on the specified dates.
Details on how these models were constructed can be found here on our Github.
Accuracy of predicting the correct winner is 63%.
All photos are scraped from Google image search acted on by a convolutional neural network (CNN) for artistic
style transfer (8,964 photos total).
Bookie odds are scraped from here.