Candidates accuracy comparison: Gukesh vs Fabiano vs Hikaru vs Ian

Hi all, I was curious who was the most accurate player of the tournament so I took the analysis of all games of the top 4 players and counted the average centipawn loss for each one of them and I got the following results:

GUKESH:

1. 13 (draw vs Vidit)

2. 18 (win vs Praggnanandhaa)

3. 7 (draw vs Ian)

4. 2 (draw vs Fabiano)

5. 25 (win vs Nijat)

6. 6 (draw vs Hikaru)

7. 32 (loss vs Alireza)

8. 9 (win vs Vidit)

9. 6 (draw vs Praggnanandhaa)

10. 4 (draw vs Ian)

11. 6 (draw vs Fabiano)

12. 12 (win vs Nijat)

13. 10 (win vs Alireza)

14. 5 (draw vs Hikaru)

average: 11.07

===========

HIKARU:

1. 9 (draw vs Fabiano)

2. 31 (loss vs Vidit)

3. 4 (draw vs Nijat)

4. 7 (draw vs Praggnanandhaa)

5. 9 (win vs Alireza)

6. 7 (draw vs Gukesh)

7. 7 (draw vs Ian)

8. 9 (win vs Fabiano)

9. 40 (loss vs Vidit)

10. 11 (win vs Nijat)

11. 12 (win vs Praggnanandhaa)

12. 10 (win vs Alireza)

13. 8 (draw vs Ian)

14. 5 (draw vs Gukesh)

average: 12.07

=============

FABIANO:

1. 10 (draw vs Hikaru)

2. 9 (win vs Nijat)

3. 8 (draw vs Alireza)

4. 2 (draw vs Gukesh)

5. 13 (draw vs Vidit)

6. 3 (draw vs Fabiano)

7. 7 (draw vs Praggnanandhaa)

8. 24 (loss vs Hikaru)

9. 5 (draw vs Nijat)

10. 14 (win vs Alireza)

11. 5 (draw vs Gukesh)

12. 14 (win vs Vidit)

13. 11 (win vs Praggnanandhaa)

14. 15 (draw vs Ian)

average: 10.00

============

IAN:

1. 4 (draw vs Nijat)

2. 11 (win vs Alireza)

3. 6 (draw vs Gukesh)

4. 11 (win vs Vidit)

5. 10 (draw vs Praggnanandhaa)

6. 3 (draw vs Fabiano)

7. 7 (draw vs Hikaru)

8. 4 (draw vs Nijat)

9. 8 (draw vs Alireza)

10. 5 (draw vs Gukesh)

11. 9 (win vs Vidit)

12. 4 (draw vs Praggnanandhaa)

13. 8 (draw vs Hikaru)

14. 15 (draw vs Fabiano)

average: 7.50

So, I got the interesting results as it turned out the winner Gukesh was actually just third in the terms of accuracy. The most accurate player was Ian Nepomniachtchi with average centipawn loss of just 7.5, the second was Fabiano Caruana with 10. Only then comes the winner Gukesh with 11.07 and the last was Hikaru Nakamura with 12.07.

So there you go. Keep in mind that these are just numbers. There is also a factor of risk involved so it would be wrong to say Gukesh played worse than i.e. Ian because he was less accurate on average. It may just mean that he played more risky chess which would imply sharper,more complicated poisitions with more time outside of theory which would inevitably make him play less accurate compared to some deep theoretical lines.

So don't take it as evaluation of chess skill, take it just as interesting stats, nothing more. Cheers!

Interesting analysis, catmate! I did not expect that Gukesh would perform lower in this metric.
Of course, this does not take into account the nature of the position. At one extreme, you have to make a series of precise moves to avoid a large drop in centipawn loss and engines excel in this. At the other extreme are positions where many moves result in the same result - evaluations may vary but to a human whether it is +4 or +2 is still winning. And in fact a human may choose the +2 move over +4 as the winning path is simpler to calculate.
At the midway point, I did another type of analysis looking at the candidates efficiency in converting better positions and saving inferior ones-
lichess.org/forum/general-chess-discussion/midway-at-the-candidates-conversion-and-resistance#1
Here too, Gukesh did not stand out and there were others that were superior at conversion. However, there were two things that did stand out even at the halfway mark-
1. Gukesh was the best at getting an advantage and if he'd converted with 100% efficiency he'd have been sole leader.
2. Gukesh did not need to hold inferior positions like the rest. Nepomniachtchi was excellent at conversion but look how many games he had to save.

Don't forget it's also dependent on number of moves. If Caruana Nepo game ended on move 72 with threefold repetition their centipawn loss would be 20/22.

Fabi was on prep upto move 70 in most of these games

Here is a video that shows the game u don't need to see much
youtube.com/shorts/50tj53xULaY?si=SS0HgdpM29kIhHWF

It would have been cool to see innacuracy, mistake ,blunder and accuracy not just acpl , that would be interesting xx,