Move 23, and you are sure the bishop is free to take. Two moves later it is mate. That game keeps replaying in your head for days: the move where it all went wrong, and the strange feeling that you had actually seen the threat.
Yesterday's fork puzzle, the one you solved on the second try, is already gone. You would expect defeat to teach better: it hurts, so it sticks. Yet a week later you make a similar mistake in a similar position.
How does that happen? The psychology of learning gives a surprisingly layered answer.
A mistake in practice versus a mistake in a game
Start with a difference that is easy to miss. In practice, a mistake is small and local: you get one puzzle wrong and immediately see the right move. There is almost nothing at stake, and the answer is right there.
A lost game works differently. Forty moves produce a single signal — "you lost". It arrives late, does not say where the error was, and stings your pride. Psychologists would call it feedback (information about the result of an action) of a very different quality from what practice gives you.
So the question "do mistakes teach us?" is really two questions. Are mistakes useful at all? And can we make use of our own defeats?
Mistakes can teach — on one condition
On the first question, research is fairly confident: yes. In a major 2017 review, Janet Metcalfe of Columbia University summed up the laboratory evidence. When people make an error and then get corrected, they often remember the material better than if they had never erred.
There is a curious detail known as the hypercorrection effect. Errors made with high confidence are corrected more readily than errors made on a guess. One explanation, proposed by Metcalfe herself, is surprise: "but I was sure!" makes you look harder at the right answer.
The key word is "corrected". Metcalfe stresses that the benefit comes not from the error itself but from corrective feedback, ideally including a look at the reasoning that led to it.
A meta-analysis by Tanmay Sinha and Manu Kapur of ETH Zurich (2021) points the same way. They pooled 53 studies and 166 comparisons. When students first tried to solve a problem on their own, made mistakes, and only then received instruction, they understood the topic better on average and transferred it better to new problems.
The effect was moderate (g = 0.36, a statistical measure of effect size). It was stronger when lessons closely followed the principles of "productive failure", in particular when instruction was built on the students' own mistakes. For younger pupils (grades 2–5), explaining first worked better.
But our own defeats teach less than we think
The second question is harder. In 2019, Lauren Eskreis-Winkler and Ayelet Fishbach of the University of Chicago ran five experiments with more than 1,600 participants. People answered two-option questions and got a short signal: "correct" or "incorrect". With only two options, "incorrect" also revealed the right answer.
Yet on a retest, people who had received failure feedback did worse than those who had received success feedback. They were even less likely to remember what they had answered. The effect held when the task was made easier and when correct answers were paid.
The most interesting part came when participants watched other people's answers. They learned as much from others' failures as from others' successes.
The authors explained this as a threat to self-esteem: failure feels bad, so attention slides away from it. In a 2022 review they also described a second, cognitive barrier. The information in failure is less direct: knowing something is wrong does not tell you what is right.
In chess this is especially visible. In a two-option test, "wrong" automatically points to the other option. In a position with thirty possible moves, knowing yours was bad says almost nothing about the good one. That is our interpretation, not a finding of the study, but it helps explain why a defeat on its own offers so few clues.
Where this explanation may not hold
"People learn poorly from their own failures" sounds convincing. That makes it all the more important to say that it is being challenged.
In 2025, Stephen Spiller published a commentary on the 2019 study in the same journal, Psychological Science. His argument is technical but important. If someone learns nothing and simply repeats the same answer, that counts as a success after "correct" and as an error after "incorrect". So part of the difference may come not from weaker learning but from a habit of answering consistently.
In support, Spiller reproduced a similar effect even in an experiment where participants received no feedback at all. The same effect appeared when he swapped the condition labels in the original studies' data.
Eskreis-Winkler and Fishbach replied in the same issue of the journal (November 2025). They agree that a tendency to answer consistently can explain the difficulty in some tasks, but argue that it does not explain the main results of their study. Both sides agree on one point: consistent answering is another barrier to take into account when measuring learning from failure.
So the question is still open. The careful version is this: there is evidence that we learn less from our own failures than we expect, but how much less, and whether always, is still being worked out.
There are other limits too. Most of these experiments are short quizzes, not chess games or sports matches. And people differ: for some, a defeat brings anger and energy; for others, it knocks them off balance for a long time.
What happens during the game itself
Defeat affects not only learning after the game but the game itself. Economists Akash Adhikari, Stanislav Anatolyev and Dmitry Dagaev analysed 885 games from the Tata Steel tournament in Wijk aan Zee, 2011–2020. That is more than 64,000 moves by elite players rated between 2603 and 2872. Every move was evaluated by the chess engine Stockfish.
The result was mixed. After small inaccuracies, the probability of further mistakes rose slightly — something like what players call "tilt". But after a blunder, players became more careful, not less.
Importantly, the study did not measure emotions directly, and it covered only the elite of one tournament. The model explains only a small share of the differences in move quality (about 3%). So this is an interesting detail rather than a law, and it should be applied to amateurs with care.
Practice gives answers; games ask questions
So should you stop playing and only practise? The data on chess players say it is not that simple.
Neil Charness and colleagues (2005) surveyed two large groups of tournament players about how they spent their chess time. The activity most strongly linked to rating was serious study alone: analysing games, solving puzzles, studying theory. Tournament play and formal instruction were also linked to strength, but more weakly.
Together, the different activities accounted for about 40% of the differences in skill. In their first decade of serious play, grandmasters had logged around 5,000 hours of solo study on average — roughly five times as much as intermediate players.
These data are correlational and rely on players' recollections, so they cannot prove that study is what makes people stronger. But they fit well with everything above. A game supplies the material: painful, concrete questions. The answers come during analysis, not at the moment of defeat.
How to use this (no promises)
None of this guarantees a result. The ideas below follow logically from the research, but they have not been tested separately on chess players.
- Separate the defeat from the analysis. If failure pushes attention away, right after a loss is not the best moment to analyse. Coming back an hour later or the next day is perfectly fine.
- Look at the game as if it were someone else's. In Eskreis-Winkler and Fishbach's study, participants learned as much from others' failures as from others' successes. Whether this works with your own game has not been tested, but trying costs nothing.
- Look for the moves you were sure about. According to the hypercorrection research, confident mistakes are the ones most readily corrected once you see the right answer.
- Write down not "what went wrong" but "what I should have played". That gets past the cognitive barrier by turning a vague "bad" into a concrete alternative.
A small experiment: after your next lost game, find just one moment where you felt most certain, and work out what was stronger. Not a full analysis — one move. Then notice whether you still remember it a week later.
Questions for discussion
- Think of a defeat that genuinely taught you something. What happened between the loss itself and the moment you understood the lesson?
- Do you find it easier to analyse other people's games than your own? Why do you think that is?
- Sometimes a defeat motivates and sometimes it takes away the wish to play. In your experience, what makes the difference?
Log in or Sign up to join the discussion.