This is the 39th installment of my fortnightly poker column in the Economic Times, Range Rover.
Being human sucks in many ways, but one of its great advantages is that little thing called the imagination. We can imagine away our frailties and pretend to rise above our cognitive limitations. We are all Walter Mitty and Mungerilal, so this following thought experiment should appeal to you. Imagine that you are not a human being, but a computer designed to play poker perfectly and take the money of puny humans. Now tell me: what would change in the way you play the game? (Pause and think about this before you go to the next para, please.)
If you were God, you would know what cards your opponents held and the rundowns of all future boards. But as a computer, you wouldn’t need that information. You would play game-theory optimal (GTO) poker, with a strategy guaranteed not to lose in the long run regardless of the hands others might have or what they might do with them. Most of us humans, on the other hand, play exploitive poker, for which the hands and tendencies of others do matter. Let me illustrate the difference.
You are heads up in a hand, and on the river make a pot-size bet. Your opponent is getting 2 to 1 to call, and needs to be right one in three times to break even. Now, the aim of GTO poker is to make your opponent indifferent to calling or folding. You will do this by having what is known as a ‘balanced’ range jn this spot. Because you are offering him 2 to 1, a balanced range here would have 1/3 bluffs and 2/3 value hands. (Note that the composition of a balanced range depends on bet sizing, or the odds you give the opponent. If you bet half-pot, giving him 3 to 1, a balanced range would have 75% value hands.) Being balanced in any spot means that your opponent has to play perfectly to break even—and if he calls too much or folds too much, you make money. Basically, you cannot lose, and are thus likely to win.
Unless you’re playing high stakes online cash games, you’re unlikely to ever actually need to play GTO. The cash-game poker I play is exploitive poker. I try to identify mistakes my opponents tend to make and exploit them. In the above example, if my opponent tends to give up too often on the river, I will increase the number of bluffs in my range. If he is a calling machine and never folds, I will have 100% value bets in my range. While this is exploitive, this is also exploitable. By deviating from GTO to exploit his mistakes, I offer him (or someone else) a chance to exploit me. If i start bluffing more because he folds too much, he, or another player, could increase their calling frequency against me.
A computer would aim to play GTO poker, and it would do this by building balanced ranges for every spot, starting from preflop, across streets and board textures. This is incredibly complicated, and humans can just come to an approximation of this. This is useful, for understanding balanced ranges help us understand our own mistakes, and those of others, even if we don’t actually intend to play GTO poker. But my question at the start of this piece was not supposed to turn into a lecture on game theory. Indeed, my own answer to that question has nothing to do with game theory or exploiting others.
In any game I play, I tend to assume, correctly so far, that I can acquire the technical knowledge to beat the game. My big leaks are temperamental ones. If i was a computer, I would not feel any emotion, and would thus avoid all the pitfalls of being human at a poker table. We lose money in poker not because we think too little but because we feel too much. I shall elaborate on this in my next column.