Applications of Prospect Theory to Poker
Introduction
In this article, we explore a theory from behavioural economics about decision making under risk called ‘prospect theory’. The first conclusion of this theory is that people tend to be risk averse when offered positive options and risk seeking when offered negative ones. We also explore the idea that these prospects may be affected by recent changes in the person’s situation to which they have not yet adapted; recent losses can make all options seem negative and hence provoke risk seeking and vice versa. Finally, we attempt to apply some of these principles to poker, drawing on examples of common mistakes made at the tables.
Prospect Theory
In 1979, Daniel Kahneman and Amos Tversky published “Prospect Theory: An Analysis of Decision under Risk” [1]. This paper presented a critique of ‘expected utility theory’ - the prevailing theory of decision making under risk at the time - before proposing an alternative model, ‘prospect theory’. Based on the responses of a student body to hypothetical choices, Kahneman and Tversky demonstrated that several core principles of expected utility theory were systematically violated in human behaviour. Prospect theory was an attempt to integrate some of these deviations into the decision making model and its success was cited in Kahneman’s 2002 Nobel Memorial Prize in Economics.
The first deviation observed in the paper was that “people overweight outcomes that are considered certain, relative to outcomes which are merely probable - a phenomenon which we label the certainty effect”. In other words, given two options with similar expected values, where one is guaranteed and the other involves chance, the former option will be artificially emphasised by it’s certainty. To demonstrate this, we examine the following problem which was presented to 95 participants.
Problem 1
A: 80% of the time receive 4,000
B: Always receive 3,000
In this study, 80% of the participants chose option B over option A despite the lower expected value. These same participants were then given the following problem.
Problem 2
A: 20% of the time receive 4,000
B: 25% of the time receive 3,000
Now, 65% of the participants chose option A with the higher expected value. Note that problem 2 is simply problem 1 with both probabilities divided by 4. Since the value of the prizes did not change, these results suggest that “reducing the probability of winning from 1.0 to 0.25 has a greater effect than the reduction from 0.8 to 0.2”. So far, this phenomenon could be described by the principle of risk aversion, but this solution promptly fails when considering the next deviation raised in the paper.
What would you expect to happen when the gains in problem 1 were replaced by losses? Option A would remain the riskier option and would also now have a lower expected value, so surely an actor whose decision was governed by either of these factors would always choose option B.
Problem 3
A: 80% of the time lose 4,000
B: Always lose 3,000
Despite this logic, 92% of participants now chose option A - the riskier option with lower expected value - directly refuting the solution of universal risk aversion. This pattern was named the reflection effect and can be summarised as “the reflection of prospects around 0 reverses the preference order”. This pattern is obvious in a simpler context: Would you rather receive 200 or 100 units? Would you rather lose 200 or 100 units? It is clear here that your choice would swap when the problem changes from gains to one of losses, i.e. when the prospects are reflected around 0. The interesting point raised in the paper is that this pattern continues when considering more nuanced decision making.
While we witnessed risk aversion in problem 1, we now see risk seeking behaviour in problem 3, despite the lower expected value. The implication of this is that “risk aversion in the positive domain is accompanied by risk seeking in the negative domain”. In problems 1 and 3, this is a direct consequence of the certainty effect. In each case the guaranteed option is overweighted, reducing the appeal of the risky option for positive outcomes (risk aversion) and increasing the appeal of the risky option for the negative outcomes (risk seeking).
We now have a model that predicts an inflection of preference around 0, but this reveals an important ambiguity. Until now, we have assumed that the gains and losses offered in the problems were being compared to some reference point that was treated as 0. In many situations, this is not an accurate reflection of the decision making process. In fact, “a discrepancy between the reference point and the current asset position may arise because of recent changes in wealth to which one has not yet adapted”. In particular, recent losses may lead to the perception of all immediate outcomes as negative when compared to the reference point of just before the loss. This negative translation of these prospects would lead to risk seeking behaviour.
Applications to Poker
In order to apply these conclusions to poker, we must first establish the relative risk of various actions. I propose that, in general, the more aggressive actions in poker tend to be more risky - primarily since they involve putting a larger sum of money into the pot. Therefore, betting and raising are more risky than calling or checking, which are in turn more risky than folding.
With this in mind, consider the first implication of prospect theory - positive prospects produce risk aversion while negative prospects induce risk seeking. One scenario with almost entirely positive prospects in poker is being in position on the river with a value hand facing a check. In this scenario, the expectation value of both betting or checking would surely be positive and hence we may expect to see risk averse behaviour. Indeed, I think that a common mistake made by poker players in this spot is to be reluctant to go for thin value - justifying their actions with phrases such as “I’m happy with the pot”. In other words, when faced with a more profitable option which has higher associated risk, people will tend to choose the option in which they are certain to see a showdown.
In theory, it is impossible to construct a similar situation in poker with purely negative prospects since folding always has an expected value of 0 relative to the money already contributed. Despite this, the second implication of prospect theory is that decision making does not always occur with an appropriate reference point and recent changes in wealth can shift new prospects both negatively and positively. As mentioned at the end of the previous section, if one has just incurred significant losses, following decisions may be compared to the reference point of just prior to the loss. This would artificially make all options seem negative and hence lead to risk seeking behaviour.
I think that this is a valuable lens through which to view ‘tilt’ in poker. Trying to “win back” losses is a common phenomenon across gambling, with poker being no exception. The behaviour that this encourages comes in the form of playing more hands preflop, folding less draws or bluff-catchers postflop, and generally seeking greater risks despite lower expected value - just as prospect theory predicts.
This effect can also be seen after a recent positive change when many players are tempted to “book a win”. This mentality leads to playing tighter preflop, avoiding difficult decisions, and overall playing more passively - in summary, playing more risk averse! We can see that regardless of whether you have recently won or lost significant amounts of money, both of these can have a serious detrimental effect on the quality of your decision making.
Conclusions
This article has covered some of the key concepts of prospect theory - namely the certainty effect, the reflection effect, and shifts of reference. We have looked at how these were first recognised from empirical data and considered their implications in the wider context of decision making under risk. Finally, we applied these lessons to poker and suggested some real life tendencies which may be manifestations of these ideas.
I hope that this has been an enjoyable exploration of a concept from behavioural economics and that it has offered insights into a few common errors made in poker. Thank you for reading and good luck at the tables!
[1] Kahneman, Daniel, and Amos Tversky. “Prospect Theory: An Analysis of Decision under Risk.” Econometrica, vol. 47, no. 2, 1979, pp. 263–291. JSTOR,
www.jstor.org/stable/1914185.