Risk-taking and attractiveness in chess
In a recent research project on expert chess players, scientists found that
male chess players choose more aggressive chess openings on average when playing
against good-looking female opponents compared to when playing against less
attractive female opponents, although they are equally skilled, experienced
and of similar age. This increased aggressiveness, which corresponds to about
one percentage point, does not improve performance (it rather decreases performance,
although the difference is not significantly different from zero). Female players
do not change their choice of opening strategy depending on the attractiveness
of the opponent. The researchers also found that men, when playing against an
attractive woman, are less interested in an early or arranged draw (defined
as a draw in less than twenty moves). That is, they are not as interested in
a peaceful outcome as when they play against a less attractive woman.
All in all, this indicates that male players behave more aggressively when
confronted by a beautiful woman. The reason for this behaviour was not investigated
in the study, but it certainly directs our thoughts towards some kind of alpha-male
behaviour. And although such behaviour does not pay off on the chess board,
it might generate some added value in what follows after the game has ended.
The measured effects, however, are average effects, implying that not all players
need to follow this behaviour. Nevertheless, similar results have been found
in other studies, for instance that male bank employees take higher risks when
approving loans to attractive women.
ChessBase interviewed Patrik Gränsmark (above), who is a leading
Scandinavian researcher studying this topic. The other researchers are Anna
Dreber, Christer Gerdes and Michael Rosholm.
ChessBase: In your research you find that men
choose more aggressive openings against good-looking women?
Patrik Gränsmark: Yes, actually we find that the more attractive
the female opponent, the more aggressive the opening. And this holds also when
comparing two female opponents with the same Elo rating.
How did you obtain the data to carry out this analysis?
Well, to begin with we actually used the data available in Chessbase 10, where
we had about 30,000 players and about 1.5 million games. All players had an
Elo of at least 2000.
First of all how do you know whether an opening is to be regarded as
aggressive or solid?
We let eight expert chess players with an Elo of between 2000 and 2600 classify
all the 500 Eco codes as being either solid, aggressive or neutral. Six out
of eight had to agree to classify an opening as being solid or aggressive. This
was described in an
article on your news page.
And secondly: how did you measure the attractiveness of the players?
That was the tricky part. We turned to Amazon.com which offers an online labor
market. We then paid 2,000 people in the USA to rate photos of the players as
regards attractiveness. We used about 600 of the photos supplied in the Players
Encyclopedia of Chessbase 10. The players were rated on a scale from zero to
ten, and we then calculated the mean rating for each photo. Each player was
rated about 50 times.
Okay, now the obviously most important question: who was most attractive
Actually, to avoid abuse and personal suffering we replaced the names with
an ID number. We are only interested in the overall effects, not to analyze
a particular player. So we don’t reveal the rating for a specific player.
That is a pity, but understandable. So what was the purpose of your investigation?
First of all, similar behaviour has been found in other situations, but we
wanted to see if expert chess players also behaved this way. Studying chess
players offers a somewhat different setting, as they are considered to be sophisticated
and intelligent. So by studying expert chess players we may obtain answers corresponding
to some upper boundary of human behaviour.
What do you mean by “upper boundary”?
One of the most common assumptions in economics is that of full rationality,
that is, we assume that people behave in a way that maximizes their profits,
given that they know how to do that. However, it is not hard to find cases when
this is not the case. Actually, it seems to be quite common. Dealers on the
stock market may be one example. Stock market actors and other financial actors
are probably more similar to expert chess players than the mean individual of
the whole population.
So by studying chess players we may learn how the most rational people
behave, right? And what can we learn from the findings?
In earlier studies, it has been observed that men applying for bank loans become
more impatient and ask for more money when seeing a beautiful woman. Actually,
their demand for money increases with an amount corresponding to a decrease
in the interest rate by 25%. If these findings are correct, imagine what could
happen at a job interview or wage negotiation between a male recruiter and a
female job applicant.
But why do you study chess players? Why not turn directly to those areas
you want to analyse?
It is very difficult to obtain adequate data on this kind of behaviour. Most
data is based on self-reported behaviour. And if you ask a recruiter or bank
employee if he behaves differently depending on who is sitting in the opposite
chair, he is quite likely to say no. That doesn’t necessarily mean that he doesn’t.
If you want to study how people behave in reality, one way is to turn to lab
experiments. Lab experiments are usually expensive and give quite few observations
and since people tend to change their behaviour somewhat when knowing they are
being observed, the problem isn’t really solved. A chess game is not that different
from a controlled lab experiment, and since the rules are the same all over
the world, it is quite easy to compare the results across countries. As we study
the behaviour ex-post, the behaviour is not affected by the presence of the
How reliable are these results? Couldn’t there be other hidden explanations
for your findings?
I guess that there can always be hidden driving factors behind research results.
However, one great advantage with the chess data is the existence of the Elo
rating. This allows us to account for differences in playing skills. By using
powerful statistical approaches we can also make sure that the effects are not
driven by age differences, cultural differences, style of play as regards the
preferences for aggressive or solid openings, etc. And due to the fact that
we can follow most players over several years, it is possible to account for
differences that are constant over time. This is a great advantage.
A final question, you talk about men’s behaviour. Don’t the female players
change depending on the attractiveness of the opponent.
No, not in this study. And, in general, this kind of behaviour is usually only
found for men. Nevertheless, intuitively I doubt that men would bother to behave
more aggressively or to take higher risks if it didn’t pay off in some sense.
It would be extremely interesting to see if homosexual players react the same
way. If the increased aggressiveness is due to sexual arousal then we would
see no such reaction from opposite-sex opponents. And if we would find that,
for instance, homosexual men play more aggressively against female opponents,
then it has probably more to do with social or cultural norms. However, at the
moment I feel that it would be a little too much to ask people about the sexual
orientation. Perhaps in the future.
We understand that you have prepared a short survey to try to find answers
to some of your research questions, right?
it is a survey for chess players with an Elo of at least 2000, and I hope to
receive enough answers to be able to tackle at least some of the questions.
It only takes a minute to answer and, as a reward, we offers a $500 prize as
a value check at Amazon.com.
Thank you for sharing your research results with us and good luck with
To answer the survey and participate
in the lottery with a
chance to win a US $500 value check at Amazon.com,
click on the yellow banner on the right.
Links to working papers on chess (in PDF):