Title matches

Reading the match

A lesson in probability, told with chess. What does it mean to give a player 30%? How much is one game worth? How well did the crowd read Game 14 in 2024? This page is about understanding chances, not about betting: ChessFlex takes no bets and links to no one who does.

What a chance really means

The coach gives each side a chance from its evaluation of the position, and the app shows it by each player's name. Here is what title-match positions at each chance for White actually ended as, 76,761 positions in all (light: White won, amber: drawn, dark: Black won), in the middlegame (moves 25–45) and in the ending (after move 45).

The same edge is worth less in an ending. At 60–69%, White went on to win 46% of middlegames but only 31% of endings; 68% of those endings were drawn. Endings are where draws live, and champions defend.

Game 14, 2024: one move, two readings

Ding and Gukesh were level at 6½ each, in a rook-and-bishop ending most people expected to be drawn. On move 55 Ding played Rf2, and the title went with it.

The coach: Gukesh's chances (Black), move by move

Gukesh wins game 14 Draw Ding wins

Polymarket's traders, by the clock in Singapore (their public trade feed): before the blunder, a draw at about 97% and Gukesh near 2–3%; minutes later, Gukesh above 99%.

Two readings of one position. The coach saw Black better, about 71% around move 50, and in title-match endings at that kind of edge the better side went on to win 53% of the time. The crowd knew this particular ending as a well-known draw, and it held for nearly four hours. Neither number was wrong; they answer different questions. One move turned the position into a lost one, and both readings changed at once.

A model of the match

A simple model, in the open: each game from the two ratings, a White edge and a draw rate, fitted on 456 rated title-match games since 1972; while a game is on, from what positions like it actually ended as (above). The surprise in the fit: a rating gap counts for only about 35% of its size in a title match.

Gukesh v Ding 2024

Model: Gukesh's chance of winning the match Polymarket: Gukesh to win (after each game)

Polymarket's price after each game (its daily price history) against the model, which never saw it. Before the match the crowd was surer of Gukesh than the model, 78% to 65%. After Ding won game 1, 55% and 43%. After Gukesh won game 11, 82% and 85%; after Ding hit back in game 12, 50% and 56%. Going into the last game, close to a coin toss for both.

Every modern match, read the same way

2026

The model gives Gukesh about 31% to keep the title, from the ratings alone. On Kalshi, a regulated US prediction market, Gukesh traded at about 28% and Sindarov 72% on 19 October 2026 (bid–ask midpoint). Two very different methods, nearly the same answer. We'll update both after every game.

Prices are quoted for information, with their source; nothing here is advice. Real-money betting and prediction markets are not legal everywhere, India included.