Frequently Asked Questions

This site shows how each contestant's chances of winning change throughout a Jeopardy game. Games are updated overnight and cover 20+ years of Jeopardy games.

All regular season games dating back to 2002. Postseason is not included for now.

The probabilities are determined by a machine learning model that prices in the returning champion's stats, current round and questions remaining, daily doubles remaining on the board, which player has control of the board, current scores, and in-game player statistics.

No. The probabilities are calculated before Final Jeopardy wagers are revealed. This is why you'll sometimes see a player with a high probability who ultimately loses due to a surprising wager.

Use the search bar at the top of the main page. Type any contestant's name (like "Ken Jennings" or "Amy Schneider") to see all their games. Click any game to view its chart.

The Daily Double Analysis page shows whether contestants made optimal wagers on Daily Doubles. Optimal wagers are determined by the changes in win probability if the player gets their question right or wrong for a given wager size, and calculates their expected change in win probability for a given get rate of the question. The process is roughly similar to the 4th down conversion calculators in football.

Visit the DD Analysis page and select any game using the season and date dropdowns. Choose a specific Daily Double from the list, and you'll see a chart showing the optimal wager percentage for different get rates. The player's actual wager is shown as a green line, and historical average get rates are shown in red. This reveals whether they wagered too much, too little, or just right.

Power ratings measure each player's relative Coryat scoring strength using a Bradley-Terry model. Every game is broken into head-to-head matchups between pairs of players, and the model estimates how much stronger or weaker each player is compared to the field. A rating of 0.00 represents an average player. Ratings are updated daily as new games air.

The numbers represent relative strength on a continuous scale. As a rough guide: a player rated 1.55 (like Ken Jennings) is expected to win about 88% of games against average competition, while a player rated 0.50 wins about 58%. The tier labels (All-Time Great, Elite, Strong, Solid, Competitive, Field) provide quick context for where a player falls among rated players.

Players with fewer than 4 games don't have enough data for the model to produce a reliable rating. The 4-game threshold filters out one-and-done contestants where the rating would be heavily influenced by a single game's variance.

The Coryat score measures a player's raw knowledge independent of wagering. It counts the value of every clue a player answers correctly (and subtracts incorrect responses) while ignoring Daily Double wagers and Final Jeopardy. This isolates fundamental buzzer-and-knowledge ability from strategic wagering decisions.

The forecaster uses a multinomial logistic regression model trained on historical tournament games. It takes the BUTTREY ratings of three players and predicts each player's probability of winning. The model was trained on over 400 tournament matchups from 2003 to present.

The model correctly predicts the winner about 46% of the time on held-out test data. In a 3-player game, random chance would be 33%, so the model provides a meaningful edge. Jeopardy outcomes have inherent variance — even the strongest players lose regularly — so no model can achieve very high accuracy.

Yes. After generating a prediction, click the "Copy Shareable Link" button below the results. This creates a URL with the three players encoded in it, so anyone who opens the link will see the same matchup and prediction.

The Excitement Index is a numerical rating indicating how exciting a game of Jeopardy is to watch. It is comprised of 6 components: volatility of win probability (with higher weight on late-game volatility), lead changes, the strength of an upset of the defending champion, the aggressiveness of Daily Double wagering, the suspense going into Final Jeopardy, and the total scoring of the game.

The default weights for the Excitement Index weight all components equally. If you want to set your own custom weights and/or see the components of each game, you can do so on the Custom Rankings page.

All data — win probability charts, Daily Double analysis, player ratings, and excitement scores — is updated automatically overnight after new games air. There is no manual intervention required.

This site is a project of Colin Davy, a data scientist in Chicago and Jeopardy champion. You can read more on the About page.