Human Wagering Behavior Depends on Opponents’ Faces

Since many of the same questions/comments are arising, I have decided to post a FAQ page regarding this publication:

  1. 1.Are you claiming that your experiment generalizes to ‘real’ poker?

No! I use poker as a research tool for investigating how people make decisions under conditions of risk. Classic behavioral economic experiments  have involved ‘lottery’ tasks that are very novel/unusual to those participating in the experiment. Therefore, I wanted to find a task that is more natural to subjects, but can still be experimentally controlled.

Since I use poker as a research tool, I purposefully ‘strip-away’ many of the factors that contribute to decisions in a ‘real’ poker game. In this study, it meant I removed such things as: position, re-raising, checking, stack-size, learning opponents via outcomes/show-downs, etc. Basically, my experiment involved a heads-up, pre-flop scenario, with an unfamiliar opponent. However, this is *exactly* what I wanted, as I was interested in studying how rapid-impressions of opponents influence behavior, independent of these other factors. Think of it as a ‘first-impression’ of your opponent.

Follow-up question: Then, why did you claim in your abstract that the best poker face may be a trustworthy face?

I meant that in the context of my experiment, and did not mean to generalize across all situations in poker! It may, indeed, be the case that it does hold in several situations when dealing with an unfamiliar opponent, but this is a topic for future investigation.

2. Your subjects were novices, and therefore, your results aren’t accurate/informative.

The fact that my subjects were novices doesn’t impact my results, whatsoever. As I stated above, I’m using poker as a tool for understanding how people make decisions under risk, and not studying ‘actual’ poker. Moreover, I’m looking for *relative* changes in betting behavior as a function of face information. Therefore, it doesn’t matter that subjects may not have an optimal understanding how hands map to EV; I only care how each person changes *their own* wagering patters, and am not interested in their absolute performance. My main concern was assuring that subjects were able to understand the poker scenario, and thus, the pre-experimental questionnaire.

The fact that almost all subjects folded more against trustworthy appearing opponents was surprising! I expected people to use face information, but not so consistently. However, it would, indeed, be interesting to see if experts demonstrate this same pattern, or if they somehow ‘unlearn’ this natural tendency as a result of their vast experience.

Side note: If you’re interested in poker expertise, I’m currently writing-up a study that investigates how brain activity (measured using fMRI) of poker experts is different than poker novices when making wagering decisions.

3. You used computer generated faces, and therefore, your results do not hold.

The faces used in this study were borrowed from a group from Princeton, and were quantitatively derived to optimally predict people’s subjective impressions of trust. Moreover, these faces have been shown to activate emotional areas of our brain (i.e., Amygdala), so they are ‘emotionally relevant’. Finally, people were led to believe that different opponents may have different wagering styles. Therefore, subjects are not being naive if they’re using a conscious strategy, since the faces may be relevant to their success in the simplified poker game.

4. Books like Caro’s and/or my own experience tell me that a happy player is bluffing, so these results aren’t valid.

As I stated above, these results may not extend into expert populations. Caro states that players who act strong, may in fact, be holding a weak hand. His book is aimed at helping novices improve their game, so perhaps my results may be the effect he is trying to ‘train-out’ of players!

An interesting side-note is that I did encounter several people that claim to have *not* used face information, whatsoever. But, in analyzing their data, I would find this not to be true: they still folded at greater rates against trustworthy opponents. So, this (consistent) effect could be the result of: 1. a conscious strategy; 2. an implicit (i.e., unconscious) reaction; or 3. both.

Therefore, research is important as it allows for implicit (unconscious) effects to be uncovered, whereas poker authors who are exclusively relying on personal experience need to be consciously aware of effects, in order to write about them. Researchers in psychophysics have long known that conscious awareness does not guarantee the reliability of information, nor does it preclude that other (implicit) factors influence behavior. The take-home message here, is that you cannot exclusively rely on your own (conscious) experience to make determinations about the causes of phenomena!

5. Your subjects were not monetarily motivated, like ‘real’ poker.

In fact, we did pay subjects based on their decisions, so they had financial incentive (see Methods in paper).

If you have any other questions or comments, please don’t hesitate to contact me; I’m always willing to talk poker research!