Do experts make more errors when they are losing or behind?

Although Stockfish and Komodo have differences in their evaluation scales—happily less pronounced than they were 1 and 2 years ago—they agree that the world’s elite made six times more large errors when on the lower side of equality.

We don’t know how general this phenomenon is, but interestingly it seems to hold much more strongly for top players than for weak players.  That is from chess of course.

Here is much more detail from Ken Regan, along with some suggested hypotheses and resolutions.


This seems to be the opposite of the mistake Trump made when he was winning marginally in Iowa...sitting out the Iowa debate like a coward afraid of a girl. Let's see how that works out for him.

Without looking at the paper (my internet is slow right now, here in the Third World Philippines sometimes the "DSL" internet is actually as slow as dial-up modem back in the days), I think the paper is addressing "Fischer Fear" meaning chess players collapse when faced with a stronger opponent. Kasparov himself vs Deep Blue made this mistake when he could have drawn or won some games--he assumed he could not win against "the infallible machine" and collapsed.

Ken's site is finally loading, I will modify the above remarks if I'm wrong...

Yes, I was right. Damn I'm good. This is analogous to science discovering what "old wives tales" knew all along (smallpox/cowpox virus comes to mind; malaria as swamp 'masma', eat your vegetables, etc etc etc)

Regan's paper: "I have previously reported some phenomena that my student Tamal Biswas and I believe owe primarily to human psychology. This one I believe is different—but I isolated it in full only a week ago so who knows. It is that the proportion of large errors by human players in positions where computers judge them to be a tiny fraction of a pawn ahead is under half the rate in positions where the player is judged ever so slightly behind. "

" I think the paper is addressing “Fischer Fear” meaning chess players collapse . . . "

"Yes, I was right. Damn I’m good"

Never underestimate Ray Lopez's self bias.

"That this training set includes only games in which the players are evenly matched minimizes any variance from overconfidence or “fear factors.”

" I think the paper is addressing “Fischer Fear” meaning chess players collapse . . . "

"Yes, I was right. Damn I’m good"

Never underestimate Ray Lopez's confirmation bias.

"That this training set includes only games in which the players are evenly matched minimizes any variance from overconfidence or “fear factors.”

In chess, often the better your position, the easier it is to improve it. If your material up/have a good attack/space advantage theb you iften have many options and your opponent is under pressure to find specific defenses.

This experience comes largely from other games, but it should apply to chess as well: The farther you think you are behind, either to the player or position, the more risk you are willing to take, so you make moves that you know are capable of being blunders on the hopes that they will be brilliant, or that your opponent will blunder in response, because you know you cannot win playing it safe. When far ahead, your goal is to avoid blunders, so you make less of them.

This is true: often the winning player is supposed to be very conservative, and so of course is not going to get out of line. The losing player is supposed to shake things up, and is willing to do harmful or risky stuff because if they don't work out, well, you were losing anyway. This is assuming that it's binary win/lose -- it could be different if there's a way to lose even worse than you're currently losing.

My response is the same as Oli5679 and a commenter at Ken's site - the worse your position, the narrower your options.

But if I read Regan's paper correctly, you and Oli are making a mistake: the player behind tends to blunder less, statistically. The front runner chokes and blunders (statistically) more often.

My bad, I misread this paper, though another 2014 paper by Regan shows that in general the player ahead tends to 'choke' more often (slightly):

For almost any degree of advantage or disadvantage, a human player has a significant 2–3% lower scoring expecta-tion if it is his/her turn to move, than when the opponent is to move; the effect is nearly absent for computers

People make more mistakes when under stress. Is this a surprise to anyone?

Isn't this the problem Bernie Madoff had? He started out as an honest investment manager, but when things turned sour he started faking it.

And wasn't there something similar in a couple recent scandals at investment houses? One in London and another in Hong Kong, I believe. A trader starts losing money, so he starts making bigger and bigger plays in hopes of making it all back. I'll bet you see this sort of thing at casinos all the time.

Oli, Zvi, and Bruce: the phenomenon shows up sharply when a position goes from infinitesimally better to infinitesimally worse, and most sharply for the best players as Tyler highlighted. The "What Can Explain It?" section expressly throws a challenge against the kind of psychological and risk-rational explanations you are offering (also see the 2014 paper on the latter).

Did you investigate to try to determine if the effect is simply from players getting small advantages/disadvantages due to them playing well or not well? One thought would be to note how many moves into a game it takes for a small advantage to be gained, where the longer it takes the better the player that ends up in a disadvantageous position must have been playing. If players that went longer into games before falling behind had a lower error rate afterwards than those that fell behind earlier I think you would confirm that it is an after effect of playing above or below ones average ability.

All good suggestions, subject to the limitations of human-interpretive examination of 10,000s of data points. The simplest way to enjoy an advantage that computers put in the 0.01--0.40 range for a dozen moves and more is to have the White pieces and be reasonably booked up.

Ah, that explains Kurada's move from yesterday, and why most economists truly suck the big one- they are always behind and don't even know it.

It makes sense, also when you are losing, you're often in time trouble which compounds the problem. Being short on time increases errors for most players

Fascinating piece and thank you Prof Cowan for linking to it. Not being a chess buff, my first thought was "Does this arise in tennis?" Many of the categories can be ported over "Slightly behind" obviously and "major error" (e.g. double fault). Would love to see if the effect is universal.

Tennis had the advantage that being ahead or behind does not dictate a player's options - each point is an independent event. A player is not forced to take risks simply due to lack of other options

It has the disadvantage that minor injuries, a cold, etc. Mean that an individual player's skill level can vary quite a bit from match to match.

Comments for this post are closed