Congrats to Ding Liren!

The 2023 FIDE World Chess Championship is now complete. It took place in the city of Astana, Kazakhstan, where Ian Nepomniachtchi and Ding Liren fought in a back and forth battle to take over the reigns from Magnus Carlsen.

You can see the progress of my predictions below. Eventually, the match kept going back and forth and I was too engrossed in the games so I stopped making updates. The page below reflects the tournament after round 6.

Predictions and Probabilities

My pre-match predictions said that Nepomniachtchi held a slight edge, with a 53% chance of winning. The probability of the match proceeding to tiebreaks was at 17%. Now, Nepo has a 51% chance of winning, and a 17.5% chance of going to tiebreaks.

The Method Behind the Predictions

To calculate these predictions, I primarily employed my usual methodology, but given the significance of this event, I went the extra mile. I developed a model specifically tailored to predicting outcomes in world championship matches. For those interested in the technical aspects, I fine-tuned my customary LightGBM model to better predict world championship games. The fine-tuning process revealed that such games are more likely to end in a draw.

Individual Game Model Predictions

When Nepomniachtchi plays white, he has a 24.1% chance of winning, a 60.4% chance of a draw, and a 15.5% chance of Ding Liren emerging victorious. Conversely, when Ding Liren plays white, he has a 21.5% chance of winning, a 62.8% chance of a draw, and a 15.7% chance of Nepomniachtchi winning.

Fine-Tuned Model Differences

The primary difference between the fine-tuned model and the standard model is the slightly elevated draw rate. For instance, when Nepo is white, the draw rate increases from 57.2% to 60.4% in the fine-tuned model. Nepo’s win rate decreases from 25.9% to 24.1%, while Ding Liren’s win rate drops from 16.8% to 15.5%.

To stay updated with the latest predictions and insights, follow me on Twitter!