How well does Carlsen score in equal Endgames?

Magnus Carlsen is strong in all areas of chess, but he is most famous for his endgame skills, especially his ability to squeeze out wins from drawn endings. But I’ve never seen any data to back this up, so I decided to take a closer look at how he scored in equal endgames to see if this is really true.
I used Lichess’ definition of endgames, namely that positions with 6 or fewer pieces (excluding pawns and kings) are endings. To find equal endgames, I checked if the first position in the endgame is equal. I also looked only at games with a classical time control.

Measuring endgame performance

My biggest issue was coming up with a good way to measure Carlsen's performance in the endgames. Since equal endgames between two top players will end up as draws, it’s difficult to say what score should be considered “good”.
I didn’t want to compare Carlsen with individual players, as there was no obvious player to compare him to.
I ended up using all games where one player was rated over 2800 as a comparison and scored them from the perspective of the 2800 player. By doing the comparison this way, I only get games where the players have a rating similar to Carlsen and are the rating favourites in each game.
To compare Carlsen’s endgame performance to the other 2800 players, I looked at the relative number of wins and losses they got from the equal endgames.

Carlsen wins more equal endgames and loses fewer compared to the other 2800 players. So the data backs up that Carlsen performs very well in these equal endgames.
Overall, he scored 58% from these equal endings, while the other 2800s scored 54%. Also, Carlsen’s performance rating was 2803, which is significantly higher than the performance rating of 2752 from the other 2800s and shows that his better score doesn’t come from playing weaker players.

Comparing Carlsen over the years

I also took a look at Carlsen’s score in these endings over the years since he became number 1 in the world.

The 2022-2024 period looks by far the best for Carlsen’s endgames, but during this time, he played quite a few league games and open tournaments, so his average opponent was lower rated. Still, he had a performance rating of 2807 in this period, which is above average, but from 2010-2012, his performance rating in equal endings was higher with 2821.
It’s interesting to see that Carlsen got the highest number of relative wins from 2010-2012. This also fits the narrative that Carlsen got even more from these equal endings early in his career.

Final thoughts

I found the results very interesting, but one could also do a more granular analysis by looking at the quality of each move played in equal endgames. I didn’t do this for this post, as the analysis of all moves would have taken way too much time.
In the past, I thought about looking at how many mistakes top players make in tablebase endgames (endings with 7 or fewer pieces, including the kings and pawns). This would be much quicker than analysing position with Stockfish, as one would only have to look into the tablebases. I haven’t done this until now, as these endings are much less frequent.
I’d have also liked to make some historical comparisons to players like Capablanca, but there were fewer strong players around back then, and the level of technique was lower. So, Capablanca scored much better in equal endgames, but this is difficult to put into perspective.
Let me know what you think about this post and if a move-by-move analysis would add insight into the endgame play.

If you enjoyed this post, check out my Substack.