Magnus Carlsen capped off the FTX Crypto Cup with an epic comeback win over Wesley So, first winning on demand as Black to force an Armageddon game, then winning that to claim tournament victory. It was an inspiring performance that cemented Magnus’ status as the comeback king. Apart from all his other gifts, it’s become common wisdom that Magnus is especially tough with his back against the wall.
But it got me wondering, does Magnus really perform better after a loss? Or does he just perform at his normal Magnus level? Which is to say, incredibly strong, but not particularly stronger than after a win or draw?
To answer this question I got Magnus’ most recent 1000 games from the Chessbase online database. I threw out bullet games but left in classical, rapid, and blitz. During the pandemic faster time controls have taken the spotlight, but bullet still seemed like a bridge too far. I also threw out the first round from each event, because it seems fairly clear that we’re talking about the ability to recover within an event, not between events, and in the first round there’s no previous round to look at. Finally, I compared how Magnus scored after a win, loss, or draw respectively.
I have to be honest here. I was not expecting there to be a big difference. Very often, we see stories in the data that aren’t really there. I was expecting to Magnus to perform about as well regardless of the previous game. But in this case, the data backed up the story. Magnus scored 66% after a win, 60% after a draw, and a whopping 73% after a loss. This is one time where the data is as epic as the story.
It also occurred to me that grouping events together could introduce some bias. For instance, if all you know is that Magnus won his first round, that would make it more likely that he’s participating in an event with weaker competition, and would therefore be more likely to win future rounds as well. The same logic applies to losses: a preceding loss should increase the odds of a future loss, all things being equal. This makes the 73% score following a loss even more impressive.
Nonetheless, I tried to address that potential bias by also looking at performance rating, which inherently takes opposition strength into account. The results were similar. Magnus performed at a 2854 level after a win, 2826 after a draw, and 2924 after a loss.
Finally, I did a similar analysis for several of his closest competitors.
There are some interesting quirks in these data. Fabiano Caruana, who has the reputation of a supremely even-keeled professional, actually suffered the most after a loss. In contrast, Ian Neponmiachtchi (Magnus’ opponent in the upcoming World Championship match) has the reputation of a temperamental player, but performed at a steady level regardless of the result in the previous game.
But Magnus’s performance is the headline. As strong as he is under normal circumstances, the data support the idea that he has the ability to tap into an even higher gear following a loss. That’s a scary thought for anyone looking to dethrone him.