Poker

Research on Poker

A large body of research has focused on the probabilities of winning and the timely use of bluffing behavior. Players typically are not rational thinkers, despite previous experiences that should inform their current decisions. More specifically, contrary to learning theory, people tend to bet more money when they should stay and quit the hand more often when they should continue. Furthermore, based on mathematical models of risk, informed poker players do not raise as frequently as they should, and uninformed players call too frequently. To be successful, players instead should strike a balance between aggressive tactics and restraint, and the optimal strategy is to call a potential bluff half of the time.

Which half, however, is up for debate. Knowing when to bluff, mathematically, is only part of the equation—knowing how to communicate competently with the other players may be equally important. Taking these factors into consideration, computer models have attempted to produce a program that can win against the best human players. To date, however, no advanced model of human intelligence has defeated human players, perhaps because such programs cannot account for idiosyncrasies in verbal and nonverbal communication.

Another body of research is dedicated toward detecting “tells” or leakage behaviors indicating that a player is a lying. Several factors may account for why it is so challenging to determine when someone is bluffing. Poker players may lack certain visible signs of deception because they feel no guilt when lying, due to the social contract permitting them to do so. Alternatively, it may depend on the stakes. Players betting their remaining money have higher stakes and may leak signs of deception; whereas players who are betting lower or inconsequential amounts may not fear losing. Other factors may include the level of expertise of the player, the amount of practice the player has at bluffing, or the initial reaction to the cards dealt.

Some research has examined physiological changes when bluffing in a poker context and has found that measures of stress predicted bluffing behavior 71 percent of the time. Specifically, using variation in skin conductance and voice pitch, the researchers were capable of identifying high-stress situations, which, in turn, predicted bluffing. Methodological idiosyncrasies, such as comparing skin reactivity to a data point after the poker hand is completed and the short duration of speaking (for example, “call” or “hit”), however, preclude this research from more real-time analysis of lie detection in poker.

...