What happened to Ding's play?

Taking a closer look at the games of Ding with engines

Ding struggled a lot since he won the world championship last year. I don't want to speculate about the reasons behind this or talk about his chances in the upcoming world championship match. Instead, I thought that it might be interesting to look at the games he had played see if engines can point out what has changed in his play.

The games

I decided to split the games into three time periods: from 2018 to the first COVID break in 2020, from 2020 to the 2023 world championship match and the games since the match. I decided to look at the games after the COVID lockdowns separately since I thought that this has affected his game already.

I also only looked at classical games. The different time periods have different numbers of games (144 games for 2018-2020, 82 for 2020-2023 and 57 for 2023-2024) so I normalised all the stats I show to make them comparable. I also excluded the world championship match itself.

Analysing the games

The first thing I wanted to look at was Ding's score over the different time periods. I calculated his total score and the scores with White and Black. There is a big difference over the years:

Ding’s score in the different time periods

From 2018 to 2020, Ding scored nearly 60% in his games and his score got even better (especially with Black) from 2020 leading up to the world championship match. But after the match his score dropped significantly and he only scored about 40% in his classical games since the match.

I looked at the average rating of his opponents to see if this might explain the big difference in his score but his opponents had exactly the same average rating in the time periods 2018-2020 and 2023-2024. I also calculated the linear performance rating for each time period.

To dig deeper into why Ding got such a poor score, I analysed his games with engines and looked at some different factors that might have contributed to it.

Conversion and resilience

One stat that I find always very insightful is how often a player wins better positions and manages to hold worse positions. I chose the cutoff for a better position to be +1 according to Stockfish and for a worse position to be at -1. Then I counted the number of games where Ding was better/worse and how many of them he won/lost. I divided this number by the total number of games for each time period to make them comparable.

Relative number of better and won games

This graph clearly shows that Ding got fewer good positions last year and when he got them, his conversion was significantly worse than before. When comparing 2018-2020 with 2020-2023 it’s important to keep in mind that Ding’s average opponent rating was lower in the latter period, so this might explain why he got more good positions and his conversion was better.

The worse and lost games show exactly the same thing.

Relative number of worse and lost games

Again, Ding had more games with bad positions and the relative number of these games he lost was much higher than in the other two time periods. Both these factors contribute to his poor score in that time period.

Note how few worse positions he lost in the 2018-2020 time period. This shows how well he played back then and his resilience seemed even better than from 2020-2023, where his opponents were rated worse on average.

Inaccuracies, mistakes, blunders

Another thing that might have changed in his play is the number of inaccuracies, mistakes and blunders. Ding’s worse play may be due to more big mistakes or a lot more inaccuracies. So I also took a look at these numbers.

To make the numbers comparable for different time periods, I calculated the relative number of inaccuracies, mistakes and blunders per 40 moves. As cutoffs I used a drop in win percentage by 10%, 15% and 20% after a move was played for inaccuracies, mistakes and blunders respectively.

Inaccuracies, mistakes, blunders per 40 moves

Interestingly, the relative number of inaccuracies he made was lower in the last year compared to 2018-2020. The worse results stem from a much greater number of mistakes and blunders.

Sharpness

Finally, I wanted to look at the average sharpness change for each move for the different time periods. The idea behind this score is that aggressive moves should increase the sharpness of the position, leading to a positive sharpness change. And more passive moves should lead to a negative sharpness change.

Average sharpness change per move

The sharpness change indicates that Ding played more aggressively before the world championship match and last year has play was more passive. This might be another reason for his poor play.

Note however that due to Ding’s worse form his opponent’s might have played more aggressively against him, so maybe this is (part of) the reason for the more passive play.

Conclusion

It’s clear that Ding’s play was much worse after winning the world championship match than before. Especially his conversion of better positions and resilience in worse positions weren’t as good as they used to be. And while he didn’t play more inaccuracies, he made more big mistakes than before.

I really hope that he’ll find back to his old form for the upcoming world championship match. Ding is a great player and I hope that he can show this against Gukesh to produce an exciting match.

Let me know what you think about the data I showed here. I also want to take a closer look at Gukesh’s games before the match, so I’m happy to hear any suggestions for that.