The Philosophy of Poker, Chapter 3: Psychology and the Art of Randomness

Hey all,

Summer classes are in full swing, and I’ve been super busy. Right now I’m taking a class on Virginia Woolf, which I needed for graduation. I don’t know how many of you have actually read her novels… but holy shit, are they tedious. Her short stories and essays I can appreciate more, but something about the post-Victorian stream of consciousness is just… ugh.

Haha, I probably shouldn’t criticize until I’ve gotten through the entire class. I also sent in my DSLR camera into the shop about a month ago. Those of you who have been following my blog may have noticed that after a certain point, there was a weird black spot in all of my photographs. Apparently, during a storm in Venice, a particle of dirt got lodged really deep in my camera’s sensor. I had to send it off to Canon, and after a month and $200 now it’s finally good as new, so I will probably start taking photographs again soon. The mental coaching is going well, and my students are readily improving. Once again, if you’re interested, contact me at haseebcoach at gmail dot com.

Here’s the third chapter of The Philosophy of Poker. This contains a lot of original material that I always wanted to teach someday, that I think no one else in the poker world has really articulated. Let me know what you guys think. As always, the next chapter will be forthcoming next week.

Enjoy!

[video]

I

There are two main aspects to playing good poker. The first we call strategy and the second, psychology, and the two are very different. But what exactly is the difference between strategy and psychology? Intuitively you know there’s a difference, but try to state it in words. It’s trickier than you might think.

You might have come up with something like “strategy is how to play good fundamentals, while psychology is about getting into people’s heads.”

That’s a good start, but we can be more exact than that. Psychology is about getting into people’s heads, but more precisely—we get into people’s heads so we can predict their gameplay. Strategy is what allows us to exploit decisions, but psychology is what allows us to predict them in the first place. Strategy is like the blade of a sword that lets us cut, but psychology is the hilt that allows us to guide the blade to where it’s most effective. They are complementary.  You cannot exploit someone if you do not know what he’s going to do—but simply knowing what he’s going to do doesn’t mean you’re going to be able to maximally exploit it. Strategy works from a static place; that is, it’s algorithmic. You input your opponent’s strategy, and it spits out the proper counter-strategy. But psychology is more artistic. It is what allows you to predict and pre-empt those strategies and how they change, based on observations. Neither is sufficient without the other.

II

We can narrow down the role of psychology as it relates to poker in an interesting way. I discussed in the previous chapter various approaches to setting up bluffing and valuebetting frequencies. You might falsely presume that bluffing and valuebetting frequencies are essentially the same sort of thing, that to manipulate them you just raise one and lower the other as you please. But in reality, it’s not that simple.

If you think about it, to say that your bluffing frequency is 0% is another way of saying that you never choose to bluff when given the opportunity. Remember, almost always in a specific spot, you’ll have some hands you can bluff with. So to say your bluffing frequency is 0% is to say that you intentionally forego making bluffs. But on the other hand, to say that your valuebetting frequency is 0% is to say that you don’t have any value hands in that spot. When you have a value hand, you don’t “make a choice” whether or not to valuebet; you always valuebet when you have a set, or top pair, or whatever—thin value bets excluded.

Let’s look at this more in-depth. Say you want to change your balance in a certain spot. To change your bluffing frequency is very easy, and it can be done immediately, because bluffing frequency is a matter of choice. It is therefore local. You simply “decide” that you want to bluff more in this spot, and then you do it; there are no other external factors you need to consider. But on the other hand, if you want to valuebet more in a certain spot, you cannot simply “decide” to valuebet more. Remember, your stack of wood, your supply of good hands, is finite and limited. So if you don’t have any here with you, you’ll have to take wood from elsewhere and re-allocate it here. Therefore, raising your value-betting frequency is not local, it is systemic, because you have to make changes to other parts of your poker system in order to balance and use the resources you need. That is, you have to remove the hands from somewhere else to bring them here.

We can think about this as the difference between psychology and strategy. Strategy is much more about making systemic, large-scale changes to our game to exploit our opponents. But psychology is more about internal, localized bluffing decisions. In other words, psychology usually centers on bluffing, because bluffing is where actual moment-to-moment choices are made. Valuebetting is tends to be much more a matter of large-scale strategy and range construction.

Of course, this is a simplification, and it’s less than perfect. Technically speaking, there are some times that your range is so strong in a certain spot that you need to start including bluffs, and so on. But despite its simplification, it is a useful way of thinking, and it will give us a place to begin.

III

It’s impossible to talk about poker without talking about psychology. And yet, it certainly seems that most of the language that we use to talk about psychology is incredibly imprecise. If you query random players about why they do what they do, you’ll often hear people talking about their guts, about timing, about game flow. But if you probe them to explain these concepts, they’re unable to do so. And how can you blame them? They’re difficult concepts to put into words, and much of what happens on the psychological field is more intuitive than rational. But that doesn’t mean that psychology is impervious to rational analysis. It simply means that it’s difficult, and we’re going to have to use language and concepts that are more precise than those we’re used to using.

Psychology will take us many chapters to fully explore, but let’s start from the beginning. What is gameflow?

Gameflow is notoriously difficult to define. Just the word “flow” seems to suggest a sort of airy elusiveness to the phrase. And yet, anyone who’s put in significant hours at a poker table seems to know intuitively what it means.

The first way of getting at this word is to look how it’s actually used in practice. What do people say about it? You might hear things like “the gameflow wasn’t going my way,” or “you should either bet or check this spot depending on gameflow.” These are two different referents, and in order to understand this word we must separate them out.

In the first case, what gameflow refers to is the “momentum” of the match. I will address momentum later on in this chapter, but for our purposes we should try to eradicate this use of the word gameflow, since we already have a more exact word for the phenomenon.

The second case is the more typical, and more recalcitrant against definition. “Bet or check depending on gameflow.” How do we parse this? What phrase can we substitute in here for “gameflow”?  We might say “the flow of the match,” but that doesn’t really help at all; it just substitutes another undefined word. We could try “what you think he’s going to do,” but this seems imprecise and vague. Game flow is more specific than that.

How can we define game flow?

Let’s try a little experiment. Historically, poker notation has been designed to notate hands synchronically. That is to say, we’ve come up with ways to denote how a single hand was played, notating every action on every street, which has evolved into modern hand histories. But we don’t have any widely accepted notation on how to notate hands diachronically. The only way that we can show someone an entire session is by cutting and pasting every single hand played over that period of time. But let’s see if we can do better than that.

First, let’s come up with a symbol for each major play. We’ll define valuebetting as V, bluffing as B, folding as F, and calling as C. So if there’s a sequence of hands where, at the same pivotal point in the hand, on the first hand our opponent valuebets, then on the second hand he bluffs, then he valuebets again, then he check/folds, then he calls, we’d denote that as: V · B · V · F · C. That is, value, bluff, value, fold, call.

You might notice that in using this notation, we’re discarding a lot of information. We don’t say how big the pots are, or even similar they are, or even who the raiser was. This can be problematic, so we’ll have to stipulate that we will only use this notation to analyze similar hands. For the purposes of this example, let’s imagine that we’re analyzing preflop 4-bets in a heads up match by one player. That is, this notation only arises once his opponent 3-bets this player, and this player chooses either to 4-bet for value, 4-bet as a bluff, to call his 3-bet, or to fold. So let’s imagine the sequence of 20 hands that looks like this (which is transcribed from an actual match):

F · F · C · F · V · F · B · V · F · C · F · F · F · B · F · F · F · V · F · C

Now, the first thing you might notice over this sample is that this person 4-bets quite a lot by any standard (adding up the Vs and Bs, you get 5/20 or 25%), but this is only a small stretch of hands. Nevertheless, seeing it notated this way brings many things to light.

First, recall how I explained that we don’t really “choose” to valuebet. We simply valuebet when we’re dealt a good hand. The same can be said for calling hands to 3-bets. Although there is a little variation in players’ 3-bet calling ranges, most players are calling primarily the same range (with a little variation along the tails of that range). Along those lines, a call doesn’t really tend to influence the game flow of the 4-betting game, because people are almost always calling a pre-defined range, and don’t play with that range too much. He simply calls when he’s dealt a certain hand that he’s supposed to call with. A call is also transparent—an opponent usually knows exactly what you’re doing and what your range looks like when you call (you might mix this up by flatcalling AA sometimes, but this is so rare that it doesn’t really affect the point).

So in a way, a call is a non-decision event in the gameflow, so we may as well eliminate them and get closer to the psychologically relevant action. By paring out the calls, the sequence then becomes:

F · F · F · V · F · B · V · F · F · F · F · B · F · F · F · V · F

Now, the argument we made about not “choosing” to call can be extended in the same way to 4-betting for value. Although there is a little variation in player’s 4-betting value ranges, most players are 4-betting largely the same range for value (some players will 4-bet/call off AJ, 77+ in a heads up match if they 4-bet more, whereas tighter players may elect not to, but it’s not a dramatic difference from the average). We can simplify and say that in a sense, a player has no control over his Vs either, and if this exact sequence of cards were dealt to each and every one of us, all of our Vs would appear in the exact same spots in the sequence.

However, this doesn’t mean that we can eliminate the Vs, because in fact our Vs are indistinguishable*by our opponent from our *Bs, therefore they are significant events. From his perspective both events look identical—they are both simply 4-bets to him, and he cannot know which one is which. Our Vs therefore influence the psychological game, even though we have no control over them, because every V and every B is evidence to him for us having more Bs. That is to say, every 4-bet that he sees is evidence that we’re 4-bet bluffing more often. Even if we 4-bet ten hands in a row with value hands, he is only likely to see at showdown only one or two of those, and assume that the ones he didn’t see were bluffs.

So let’s see what this looks like when we emphasize the Bs and Vs—that is, the 4-bets.

F · F · F · V · F · B · V · F · F · F · F · B · F · F · F · V · F

By highlighting it this way, we can see how the Fs become like the spaces around the 4-bets. So how can we explain the dynamics behind this sequence? Look at our pattern of Fs here. First we have three continuous folds, then we have a value hand. Then we fold one hand, then we make a bluff. We get dealt a value hand immediately afterward. Knowing that our opponent just saw a condensed sequence of bets, we decide to cool it, and we make four folds. Then one bluff, then three more folds, then a value hand comes by again, and after that a fold.

Clearly, this player is usually choosing to fold three hands or so between his bluffs. After the sequence of V · F · B· V put together, he folds for an extra-long sequence of four hands, probably to compensate for the fact that his opponent assumes he’s getting out of line, and he wants to restore his image. At the end of the sequence there is one last valuebet, followed by a bluff. If I were a betting man, I’d bet that following this pattern, the player in this example folded the next two or three hands after this.

Basically, what we’ve just done is analyzed the gameflow of this twenty-hand sequence. Using this perspective we can define the word more concretely: gameflow is the pattern of decisions made over time, and how that pattern influences subsequent decisions (we might be inclined to call it the pattern of bluffs and valuebets, but it’s not always restricted to that). In trying to analyze this pattern, there are two main elements of gameflow: simulated randomness, and emotional dynamics, each of which I will go over in depth. But I want to start our exploration with one peripheral thesis in the background, which I want you to consider as we start uncovering these topics.

The thesis is this: gameflow is a human phenomenon. That is, if two computers played each other (and both computers knew the other was a computer), gameflow would not exist.

IV

How good do you think you are at being random?

No need to answer, actually. According to the bulk of scientific studies, most of us are pretty bad at it. For example, let’s say that you took a bunch of people into a room, and asked them to simulate random coinflips. That is, you don’t give them any coins, but tell them to imagine that they’re flipping a fair coin 20 times, and to write down the imagined results. What do they do?

Well, they usually end up pretty close to 10 heads and 10 tails, often within one flip. This might be reassuring to you, but it shouldn’t be. People will almost never end up with 7 heads and 13 tails, or 6 heads and 14 tails—on average, the proportionality will tend to look too perfect compared to true randomness.

Secondly, their clusters tend to be too small. That is, if H is heads and T is tails, people would tend to write something like HHHTTHTTTHTTTHHHTHTH. Observe this sequence. The longest cluster here is three letters. That might seem fine and random-looking to you, but what that means is that, predictably, the chance of the pattern discontinuing after three letters is close to 100%, where obviously it should be 50%.

Now here’s an actual randomly generated coinflip sequence: THTHHTTHTHTTTTTTTHTHT. As you can see, this particular string has 6 heads and 14 tails, with one cluster 5 letters long.

Another example of this phenomenon is if you take dots in space. If I ask you to imagine what a random grid full of 100 dots looks like, you’re probably going to imagine something like this:

Wrong. A truly randomly generated grid of dots actually looks more like this: 

What can we conclude from this? Basically, people have a pre-conceived idea in their head of what randomness “looks like.” That image of randomness tends to be quite pristine and orderly compared to reality; it is unnaturally periodic and uniform. True randomness has clusters, and tends to look less pretty.

This is the same sort of thing that is going on in gameflow. A big element of what we’re doing in gameflow is* attempting to simulate randomness*.

Remember the sequence from before:

F · F· F · V · F · B · V · F · F · F · F· B · F · F · F · V · F

This time I’ve bolded the folds and bluffs. Recall how I said that we don’t really “choose” when to valuebet, we simply valuebet when we have good hands. By that logic, what I’ve bolded are the plays we’re in control of. In other words, it is through our pattern of folding and bluffing that we attempt to simulate randomness in the 4-betting game.

Now, let’s pull back for a second, and try to generalize to a higher-level metagame. When it comes to gameflow, we have two metagame choices at any given time. The first is to try to look random, to try to look honest, so to speak. The second is to try to appear intentional and non-random. When would we want to do either of these?

Specifically in the 4-betting game, there are two possible incentives for us. First would be for us to be bluffing and for our opponent to think we’re valuebetting, and the second would be for us to be valuebetting and for him to think we’re bluffing. We have to consider how we can accomplish either of these through our actions.

Remember, the only truly random events in the 4-betting game are getting dealt value hands—everything else is only simulated randomness. So if we can convince him that all bluffs are actually just the distribution of those random value hands, then we accomplish the first incentive—getting him to think we’re valuebetting when we’re bluffing. The second incentive—getting him to think we’re bluffing when we’re valuebetting—is less actionable, because in order to accomplish that we need to make our valuebets look intentional, or non-random. Of course, it’s great when that happens. You’ll notice how when you get big pocket pairs three times in a row, you tend to get paid off on the second and third one. This is why—it doesn’t look random to your opponents. Unfortunately, when we get dealt our value hands is not within our control. The only thing we can control is how well we simulate randomness to our opponent.

So the importance of simulating randomness is getting our opponent to think we’re valuebetting when we’re actually bluffing—in other words, to overrepresent the Vs in our sequence. How can we use all this information to help us improve at gameflow?

V

In essence, game flow is a language. The phrases in this language are the patterns of folds and bets that two players make with one another. But as any language, they are not communicating directly—their phrases only have meaning by way of their referents, or signification. In this case, what they signify are the patterns of simulated randomness.

Remember, people tend to have an inordinately tidy notion of what randomness looks like, and poker players are no exception to this rule. Furthermore, when poker players are faced with randomness, they face it experientially­– even if they are cognitively aware of the aforementioned randomness bias, they are not facing randomness in poker as one discrete string of letters, like the Ts and Hs we’re presented with in our analysis. Most poker players are therefore subject to the much the same tendencies in this regard.

 So we know what our opponent expects randomness to look like—orderly and uniform. In that case, it doesn’t matter that we know true randomnesstends to cluster and look more irregular. Our goal is in fact not to be truly random, but to make him think that we’re random, which means we must accord to hispreconceived notions of randomness.

You might say therefore, that our final conclusion would be that in creating our gameflow strings, it is best to make them to look regular, and to distribute them as evenly and orderly as possible (with enough variations to look random-ish, and not simply repeating after every third fold). That would be an easy conclusion, and in fact, you probably already intuitively do that. But it is from here that leveling game and complexities of gameflow begin.

You could say in a sense that because almost everybody tries to construct their gameflow in this way, this simple method of depicting randomness has become the norm. Your opponents will automatically assume that this is what you’re doing—in fact, it’s just what people do. We could say in a sense that this is the first level of the randomness-simulating game. If you’re any good at all, your opponent will assume that you know that this is how randomness-simulating is “supposed” to be done.

What we can do next, then, is to depart from this first level, and start to engage in wordplay.

First, remember in this first level of randomness-simulating, the Vs, which are truly random, will sometimes unwittingly cluster after Bs. This is simply a matter of happenstance. A first-level randomness-simulator will almost always immediately cool off after such a cluster, as though to signal to his opponent, “don’t think this reflects on my image! My hands just randomly clustered; it wasn’t my fault!”

The second leveler decides to start playing with this language. The second leveler knows that his opponent has been conditioned by all of the other poker players to see clusters as being unintentional—something that a poker player will apologize for, and avoid. In this way, clusters have effectively taken on a meaning of harmlessness. And because he perceives that, a second leveler realizes that he can use clusters as a tool in his gameflow arsenal.

Instead of always trying to simulate spacious orderliness, he starts utilizing clusters. He’ll bluff twice in a row. Three times in a row. He’ll B · B · F · B. Not too often of course—he cannot afford for his opponent to be absolutely sure how he is playing with this language. The second leveler wants his opponent to believe that he is just a first leveler on a good run. In this way, he develops a second-order metagame of gameflow phrases. For example, we could say 1 = B · F · F · F, his basic spaciously random phrase, and 2 = B · B · F · B (and so on, but we’ll stop here to simplify the example). The second order gameflow metagame might look like:

1 · 1 · 2 · 1 · 1 · 2 · 1 · 2 · 1 · 1

So in order to attain mastery of gameflow, one must learn how to be able to play with clusters. Great players all are aware on some level of their mutual capability to play with these clustered strings, and it is the battleground for many serious mindgames. But in the end it is only through play and experimentation that you discover how your opponent reacts to these clusters, how willing he is to believe them, and what sort of threshold he has before he is impelled to action. Ideally, you want to skirt until the very edge of this threshold.

You want to find and bluff up until the very edge of his patience, for that line is the line of maximal exploitation.

I would be remiss, however, not to admit that I have simplified this example. Generally speaking, one does not simply choose 4-bet bluffing hands on the basis of gameflow—most people are choosing only good hands to 4-bet with that have some playability in the case that your opponent calls, or some card-cancelling effects (i.e., ace-rag). This has the effect of automatically further randomizing your 4-bets, provided that you never 4-bet unless your hand falls within that threshold of hands you’re willing to 4-bet with. In essence, this dampens the gameflow. It still exists, but it is less granular than in other more obvious gameflow sequences.

VI

In the end, gameflow is the stream through which everything in poker flows. It is omnipresent, inescapable. Every hand that is ever played between two humans will be processed through these patterns.

Consider the large scale adjustments we discussed in the previous chapter. We talked about sweeping ideas like balance, constructing frequencies and ranges and so on. Perhaps when it comes to being dealt value hands, which are out of our control, we can concede to that approach to analyzing poker. But when it comes to bluffing, we see now just how many decisions there are to make, how many games within games, how inexhaustible the complexities of the moment. It is one thing to say we want to bluff 66%, a grand holistic number, but the art of bluffing from hand to hand is atomic, reductionistic, one hand after another.

Do these perspectives contradict one another? Which is correct?

Let’s construct an example. Let’s say that you wanted your bluffing frequency in all spots X to be exactly 50%. And let’s say that aggregated over an entire match, you bluffed in spot X fifty times, where spot X came up 100 times over the match. On first glance, we could say that you have accomplished the frequency you set out for. But perhaps, in fact, for X1, X2, X3X50 you never bluffed, and then for X51, X52, X53X100, you always bluffed. Averaged over the total match, you could consider this to be the 50% bluffing frequency you wanted—but obviously that’s absurd on its face. We would naturally call this instead a bluffing frequency of 0% for 50 hands, and then 100% for the next 50 hands.

Let’s extend that same argument to a more realistic sequence. Perhaps you think you’re being balanced over a 10 hand stretch, but your pattern of bluffs is F · B · F · B · F · B · F · B · F · B. You might reasonably call this a balanced sequence where your bluffing frequency is 50%. But we could also say that for each individual spot in the sequence, your bluffing frequency is either 0% or 100% depending on whether the instance is odd or even. You see, even if we make the pattern more complex and not perfectly repeating: B · F · B · F · B · B · F · F · B · F, you can extend the very same argument, albeit with lower accuracy (instead of 100%, it might become 80% likelihood on every even number).

You can see where this argument is going. Perhaps it’s possible that in our humanness, in our bounded abilities to simulate randomness, we will never be able to truly randomly re-create any large-scale statistical percentages; perhaps over an entire match when we tell ourselves we’re going to bluff 33%, really in every given spot we’re predictably (perhaps we don’t know the algorithm that predicts it and perhaps our current opponent doesn’t either, but predictably nevertheless) bluffing a lot higher or a lot lower than 33%, and never at any given time are we close to the actual frequency we’re aiming for. Are we foolish to even plan something like that to begin with?

Let’s say that hypothetically we equipped ourselves with a random number generator, into which we could enter in a frequency (e.g., 50%), and it would output either B or F perfectly randomly. That is, we’d forego trying to simulate randomness—we’d use real randomness. Imagine the consequences.

It’s tantalizing, isn’t it? Essentially, if we did this, we would throw off our humanness. We would become one step closer to machines, and escape the stream of gameflow. Would that not be a molting out of our human shells; would that not be a victory of our ingenuity?

In fact, I know of some people who have attempted to play poker this way. Such programs are actually very easy to make. And yet, you look around and nobody does this. So why isn’t this widespread? Why didn’t this method catch on?

In a way, doing something like this—randomizing your gameflow—is like trying to be perfectly balanced. It doesn’t attempt to exploit your opponents, but instead, absconds from the exploitative game altogether. The field of battle becomes smaller. Whatever potential mistakes your opponent could be induced into making in gameflow are now dismissed when you randomize in this way. If in every spot X you always bluff exactly 50% of the time completely randomly, then it doesn’t matter how your opponent patterns his guesses over time, the only thing that matters is his overall proportion of guesses. Once you do this, the river of gameflow begins to dry up.

Consider the following: by necessity, in all matches, if one is not artificially perfectly randomizing one’s gameflow, then someone is always going to be better than the other at the randomness game. That is to say, either you will make your opponent guess wrong better-than-chance, or your opponent will be able to guess right better-than-chance. It’s always one or the other, in every match. So if we were to choose when to use the randomizing machine, the optimal strategy would be to only use it in cases where we believed our opponent was better than us at the randomizing game. In any case where he isn’t, we should choose instead to play the game of gameflow, to dissect the stream into its individual atoms, to look him in the eyes and try to outwit him.

But in the end, the reason why people haven’t started using randomizing programs (other than, of course, because people tend to overestimate their own skill) is because reading and creating gameflow are skills that can only be honed through undergoing repeated practice and stress. Using a program of that nature to escape gameflow only retards your progress as a poker player. After all, the very players who can out-read you are the ones against whom you want to be honing and practicing your gameflow, in order to get to the point that you can overtake them. It is only through failure and loss that any of us have the opportunity to grow.

After all, gameflow is a characteristically human phenomenon. But as poker is a game played between humans, it becomes a vital terrain to master.

VII

Now, in our reductionism we have been overly simplistic. We have assumed we can dissect a poker match into similarly related hands, and analyze their individual sequence in isolation. But we all know that in reality, a poker match is not that cut and dry. The way we presented the 4-betting game was very discrete, but in reality, most bluffing spots affect one another, even if the spots are dissimilar. It’s clearly relevant to a 4-bet in our pattern whether or not somebody just lost a huge all-in the previous hand.

This is due to the emotional dynamic. The emotional dynamic can be defined simply as the way that emotions and the perceptions of emotions affect gameplay. Note the emphasis on “perceptions.” It’s obvious that two very emotional players will be subject to emotional dynamics, but it is also the case that two especially stoic players, who do not emotionally respond to something like losing a big pot, are still responding to emotional dynamics. That is, each player will assume that the other is responding to emotional dynamics, and will also assume that his opponent assumes that he himself is going to be emotional even if he isn’t, and so on.

For example, let’s say you never get angry after losing big pots (perhaps because you just got married today, and so you’re too elated to be phased). The moment after you lose a big pot, the emotional dynamics that govern that match aren’t dismissed merely because you don’t feel any anger—you have to respond to the fact that your opponent expects you to feel angry. The fact that you’re “supposed” to feel angry implies that the next hand, your opponent will expect you to get out of line, and so you have added incentive to do the opposite. Emotional dynamics are just as much about perceptions and stereotypes as they are about reality.

Emotional dynamics is like the undercurrent of feelings that underlies a match. It is the tide of emotions as it ebbs and flows over time. As human beings, we are acutely aware of emotions before we think in terms of strategy. Long before we were assigning preflop raising standards and postflop tendencies, we were saying “this guy must be pissed off,” or “that guy must be thrilled.” It is the most apparent way of making sense of what’s going on at a poker table, but even at the highest levels, it must be taken into account.

Tilt is an enormous topic in poker, so I will be addressing only the smaller part of it in this chapter: the tilt of other players. As far as the tilt that you yourself experience and grapple with—that a much larger and more expansive subject, which we will explore later. For now, let’s look at what we can say about other people’s tilt.

Tilt can be defined most simply as allowing one’s emotions to negatively affect one’s gameplay (notably, we tend to exclude fear as an emotion from this definition). Generally speaking, you will see tilt activated by three major events: a failed play, a lost pot, or an emotionally compromising state (i.e., he’s had a bad day, he’s tired, he’s drunk, etc.). This much is obvious if you have any experience at a poker table, and there is little need for us to expound on it. But there are a few important aspects of how tilt arises that I want us to consider.

First, tilt in its nature can be likened to boiling a pot of water. There is a threshold of negative events that a player can take and maintain his constitution, just as how a pot of water can take a certain amount of heat before it reacts. Once it gets near the boiling temperature, only then it begins to bubble—and once you push it far above that boiling temperature, it reacts more and more violently. This is essentially how tilt works. Any opponent will be able to take a certain amount of beating, until he hits his tilt threshold and starts to react. The further you push him past this threshold, the more and more he will react. But not all opponents will tilt in the same way.

There are many different styles of tilt that you’ll see among players, but generally speaking there are two main species of tilt that you should be looking out for. The first is hot tilt and the second is cold tilt. Hot tilt is the more familiar one—it is generally a tilt of anger (whether from entitlement, frustration, victimization, or whatever). The player will be aggressively trying to get even, and so he will play more aggressively, make more wild calls, and more crazy plays. The emblematic play of a hot tilter is a preflop open shove. Of course, once you start playing serious stakes, you’re almost never going to see this (especially not in the year 2012), but it’s nevertheless the spirit behind hot tilt. Every bad play of hot tilt is in a sense just a preflop shove with less money behind it.

The second kind of tilt is cold tilt. Cold tilt is more passive—it is generally a tilt of resignation, lethargy, and disheartenment. A cold tilter will start playing more passively, making less plays at pots, folding more to big bets, and acting as though he’s incapable of wrestling for pots. You’ll find a player in cold tilt when he is running poorly, you’ve got strong momentum on him, and he can’t seem to muster anything together—but you haven’t outright destroyed him in the match. He’ll start to feel a little helpless and feel like all of his moves are getting squashed, but he won’t feel like you’re completely beating him down, so he’ll still have a bit of faith in the core of his game. Generally what this will do is mollify his aggression, lock away his fancy moves back into the closet, and leave him playing only the most basic part of his game. A player on cold tilt is a player who’s lost his confidence, but isn’t convinced yet that he should quit. Where hot tilt tries to overpower and strongarm his opponent to get even, cold tilt is essentially hoping for a good run of cards to get him even.

Not all players tilt the same, and you should be reacting to each of these two types of tilters differently. Against someone who’s going through a stretch of cold tilt, aggression is key. Once your opponent has put down all of his aggressive weapons, you should pull out your entire arsenal and snatch away every dollar he doesn’t fight for. Against someone on hot tilt on the other hand, it is often enough to just play solidly and adapt to his wider postflop ranges, incorporating some heavier aggression to attack his weakened ranges, but not moving too far outside your normal tightness. Simply playing reasonably will often inherently exploit a hot tilter fairly well—just be sure you’re not giving up too many pots simply because of his aggression.

There is another vital point to be made here on how to play against a tilter. Say, okay, we know our opponent is tilting. What do we do against him now? Do we just keep playing our normal game and let him eat his own screw-ups?

First, you must consider an important question. Why is your opponent tilting? This is important. If you can identify the emotion or incentive for which your opponent is tilting, then you can better take advantage of the situation. Remember that even in tilt, your opponent is not acting irrationally. On the contrary, he is still acting in accordance with his values, but his values have now shifted. The value of playing soundly and taking the most +EV line has been superseded by the value of trying to increase variance, or maximize his winnings, or vent his anger, or whatever it is he’s trying to do. You need to consider—what does my opponent value right now? Because once you figure out what that is, you need to give it to him.

What do I mean by this?

It’s the most common thing in the world. You play against somebody who’s tilting, he’s starting to really spew, and soon after, he quits you. You probably don’t even notice or think this is a big deal, but it is. It’s a huge opportunity lost. When your opponent is tilting, you want to keep him tilting. And in order to do that, you must understand how and why he’s tilting, and how to keep him in that headspace.

Say, for example, that you can tell your opponent wants to maximize variance—he’s 3-betting 50% now, and seems to be potting or c/fing on flops. So your opponent’s incentive is maximizing variance. Now, an adjustment a lot of people will make to this will be to start minraising, since if he keeps 3-betting the minraises, it’ll create much deeper SPRs (stack-to-pot ratios) postflop, which will give you the advantage. Well, turns out this is a terrible mistake, because your opponent will leave pretty fast once you start doing this. And of course, because you usually don’t know for sure that he left because of what you were doing, it’s easy to dismiss this event and not recognize it as a mistake.

Instead, a more experienced player knows to give the tilter what he wants. The moment that you are playing against a tilter, he has established a symbiotic relationship with you. You are now exchanging value. You are giving him what he wants (variance, a chance to get even, a way to vent) and he is giving you what you want—his EV. In the end, he doesn’t have to tilt against you. He can go off to some other table for someone else to give him that, or he can go play roulette, or punch his wall, or whatever. If that’s what your opponent wants, then it’s your job to be all those things for him—be his roulette table, his wall to punch, his whatever. Be what he needs, and he will keep playing you.

So what does that mean in practice? It means if your opponent wants to maximize variance by 3-betting you 50%, then let the variance increase. Tighten up your opens, play bigger pots, and play faster. Rather than minraising 100% on the button, 3x 70% instead, widen your 4-betting range, and try to play faster to make the game speed up. This will make him feel like the game is moving and that he can get all the variance he wants in a shorter amount of time. With the game speeding up, your folding is not as likely to make the game (or variance) feel like it’s being slowed down. Chances are, your opponent will stay with you if you keep it up, and you can rest assured that if he leaves, you did all you could to keep him.

Remember, if your opponent is going to be choosing a suboptimal strategy, there is generally more than one way for you to exploit it. You want to weigh each of these options based on how they coincide with what is motivating his tilt, and choose the better option. But generally speaking, you don’t want to just submit to a suboptimal strategy if you can avoid it—i.e., if somebody’s 3-betting 50%, and you don’t feel comfortable calling a wide range against him, you usually don’t want to continue 3xing 100% on the button if you can avoid it. But again the bottom line is: if a tilter wants something, give it to him. The customer is always right.

VIII

Earlier on I spoke of momentum, and vaguely equated it to “how the match is going.” Again, this is one of those intuitive words that we all seem to use, but eludes definition. What is momentum exactly? More specifically—we’ve already come to the conclusion that gameflow is a purely human phenomenon. Is momentum the same way? Is momentum simply “imagined into existence,” a psychological phantasm, or is it a real thing, rooted in the cards themselves? Would a computer playing poker feel momentum?

To begin, let’s look at real instances of momentum. We’ll take two examples—one of positive momentum, and one of negative. In the positive, let’s say that our opponent is down two buyins; we make a big three barrel bluff where a draw misses, and he tanks and folds the river. We feel pretty good about this, and we can imagine he feels pretty wobbly. Now, in the negative, let’s say that we are down two buyins; we’re facing a three barrel when the draw misses, we tank for a long time and then fold. He thinks and doesn’t show. We’re now down three buyins and we feel like we just got body-slammed.

Where does the momentum actually come from in each of these examples? Certainly, a big part of it comes from the emotional reaction to the event. Even the best, most equanimous poker players in the world would have some kind of emotional response here, however small. But let’s try a little thought experiment. Imagine for a second that we don’t have an emotional reaction to this. Totally cut out the emotional element. Imagine maybe that you won the lottery today, so you don’t really care about being down three buyins; you’ve just become a multi-millionaire, so you’re on way too big of a high to care. Pretend you’re totally Zen’d out. Really imagine it!

Close your eyes. He shoves, you tank fold, and he doesn’t show. Now what? Is the negative momentum still there? Do you still feel it?

I’m guessing you do. Even if you don’t feel anything physiologically negative or painful, you feel the feeling of playing from behind. You feel the momentum.

So momentum is more than just the emotion. Well, you could make the argument, as we did with emotional dynamics, that your opponent’s expectations affect the match just as much as the emotions themselves. So if your opponent expects you to feel behind, you have to react to that expectation, and that’s really what momentum is about.

Although this is true, this doesn’t seem satisfactory either. This phenomenon is, after all, a result of imperfect information—our opponent has to make assumptions about us because he doesn’t know us. But let’s say we’re playing someone who knows that we’ve just won the lottery, and so he knows three buyins isn’t going to affect us emotionally. Is the momentum still there? Imagine it.

Yup, still there. So what is momentum then exactly?

It seems almost mystical. Being a poker player, you know from the bottom of your heart that it’s impossible for what you’re dealt in one hand to affect what you’re dealt in the next hand. All hands are independently random events, aren’t they? So why should we have more expectation of losing when the momentum is negative, even when we take all of the emotions out of it? It doesn’t seem to make sense.

And yet, no matter who you are in the world, or how well you control your emotions, we want to advise the person with positive momentum “keep playing!” and the one with negative, “quit now, tomorrow’s another day!”

You could try to explain this all away as an example of the gambler’s fallacy (the tendency to see nonexistent patterns in chaotic information), but I think you miss something if you try to explain it all away. There is another answer is rather more interesting. If you think about it, if you tank-fold a river when the draw misses and don’t get to see what he has, what has taken place is not merely an emotional spike. You have also created an informational asymmetry. In other words, your opponent now has acquired more information than you have. By the fact that you tank-folded, he knows that you were considering a hero-call, and decided to fold it instead. He knows that the weight of evidence you evaluated made you decide it was a fold, so he can reconstruct your mindset and assumptions about him. But you don’t know whether or not he was bluffing. While you have gained no information, your opponent gains a great deal. This is informational asymmetry, and it’s a significant part of the reason why when you have negative momentum, you expect to continue to play from behind.

There are certain situations that produce negative momentum which don’t necessarily have informational asymmetry, such as a big hero call. Say we bluff on a board where the draw misses, and our opponent snapcalls our river shove with bottom pair. He got the information that we are shoving that river with air, but we also got the information that he’s calling with anything and clearly doesn’t believe us. So although we lose a buyin here, we can actually make a useful adjustment by tightening up and not bluffing anymore in spots like this. But again, in real matches, negative momentum will still persist, not only from emotional reactions (who wouldn’t feel a punch in the gut to get a bluff shove snapcalled by bottom pair?), but also from the fear that your opponent has a better and more accurate read than you do.

So momentum is real, in many more senses than one. And what this tells us is that one should never be ashamed of quitting a match when one’s momentum is negative. There are always more matches to play, and more fish in the sea.

IX

So we’ve looked now at the art of randomness and emotional dynamics, but that leaves us with the final and largest element of psychology unexplored—opponent modeling. That is, our ability to observe our opponent and integrate all of the information we collect into a coherent model of behavior. Again, we will begin our exploration in the external, and move toward the internal—that is, we’ll begin with thinking about how our opponents construct their perceptions of us, and then we’ll move on in the next chapter as to how we will construct our models of them.

We refer generally to the phenomenon of our opponents perceiving our behavior as our image. Image is what we imagine our opponents see, and how we think they are judging our behavior. Unfortunately, giving it a name doesn’t help us get any closer to figuring out how it works.

The first and most pre-eminent rule of thumb that dictates how image is formed is the strength of first impressions. The strength of first impressions is a cognitive bias that is widespread and universal, no matter who you’re dealing with. It can be simply defined as the resistance of initially constructed models to be altered by later information.

For example, if in the first couple of hands since you sit down, you 3-bet twice (perhaps the second time for value), your opponent will immediately form a model of you as an over-aggressive 3-betty player. Even if after twenty hands your frequency becomes depressed enough as to appear normal, it will take a much longer time for him to extinguish the attributes that he’s ascribed to you—and even if he finally changes his model, he will be quick to return to the initial one if he sees you ever do a run of 3-bets again.

You should take advantage of this bias by always being sensitive to what your initial plays are, and how they are likely to appear to your opponent. Rather than thinking about your actual frequencies (whether you’re really a tight or a loose 3-bettor), learn to focus instead on what he has seen in his limited experience (i.e., don’t commit the fallacy of omniscience).

The second rule of thumb is the projection bias. The projection bias refers to a subconscious presumption that other people think like we do—which by extension means that they perceive like we do, and that they will act as we do in most situations.

One can extrapolate many conclusions from this bias, but the most important is this: players will expect everyone else to think and act the way that they do. This implies that, all things equal, conservative players will expect other people to play more conservative, and loose players will expect other players to play looser. Generally speaking, if you’re playing against a nit, he probably assumes you’re nittier than you really are, and if you’re playing against a maniac, he probably expects you to be more maniacal than you really are. This is not to say that they think you will do exactly what they would do, but instead that their perception of your behavior will consistently be swayed in their own direction. Remembering this bias will serve you well in reconstructing other player’s perceptions, and expectations.

So we’ve looked closely at some big psychological concepts, but to get further in our analysis of psychology, we must begin to turn inward, and make a deep and substantial analysis of opponent modeling. By first learning how to construct models of our opponents, how to predict their behavior, and studying the nature of our interactions with them, only then will we be able to effectively play the adjustment game against them—all of which we will explore in the following chapter.

I hope you found all that interesting. Leave a comment and let me know what you thought. Or if you have any questions, corrections, or suggestions, drop a comment. Till next week!

Note: this is from the rough draft of my first book, How to Be a Poker Player: The Philosophy of Poker. If you like what you read, consider buying the completed book—it's tightly edited and contains new material! Hope you enjoyed. :)
Haseeb
June 15 2012
Like this post? Subscribe.
I'll send my new posts (and secrets) to your inbox.
or subscribe via RSS.

Managing Partner at Dragonfly. Effective Altruist. Airbnb, Earn.com (acquired by Coinbase) alum. Writer. Former poker pro. Donate 33% of my income to charity.
more

San Francisco
Load comments 

Hey guys, Sorry for missing my update last week. Finals caught me off guard, and it’s been tough to keep up writing consistently. I’m taking a full summer term as you guys know, so there’s a lot of schoolwork to keep up with—5 days a week, with lots of homework. Once this semester is over, I’ll have a four day break and then onto my next term of classes, haha....
read more

Jun 2012
next

Hey guys, My summer classes have just begun. I’m taking a full load in the summer since my goal is to graduate by the end of the year, so my schedule has been really intense. Basically, I’m waking up at 7am every morning, writing in the library in the morning for a few hours on my book, going to class from 11-5PM, then coming home to eat dinner, work out,...
read more

Jun 2012
prev
Popular
Years