 Poker is the most challenging game I've ever played, not because of the strategy, but because of the philosophy behind it. Poker has an infuriating relationship between theory and data, between knowledge and execution, between understanding and demonstration. Here's what I mean. You can play perfect poker and lose all of your money. You can clearly understand the game of poker and be beaten by an unknowledgeable fool. The fool, by contrast, can play objectively horrible moves and yet walk away a winner. This isn't the case with games like chess. In chess, if you play correctly and understand what you're doing, you will win every game against foolish players. It's very easy to demonstrate chess competence. If you know what you're doing, you'll win. If you win, it's a demonstration that you know what you're doing. You can explain this away by saying, well, poker has an element of luck, but chess has no luck. That's the difference. But this misses a much larger philosophical issue, the relationship between theory and data, or more broadly, the relationship between rationalism and empiricism. The question is this. How do you know when you understand the game of poker? Is it when you start winning, when you get an empirical validation that you're doing something right? Or is understanding poker entirely cerebral, meaning you can understand the game without empirical data? These questions parallel debates in epistemology. How can you understand knowledge about the world? Can you merely sit in an armchair and think about it, or do you need to go out into the world and gather data? Furthermore, what happens when the data isn't in accordance with our theory? Do we revise the theory, or do we dismiss the data? Traditionally, rationalists, roughly speaking, say this. Knowledge is ultimately based on rational analysis. You can know certain things by sitting in an armchair and thinking about them. Theory is fundamentally primary. Data is secondary. No data speaks for itself, and in some circumstances theories can be known to be true despite empirical data to the contrary. The empiricists, roughly speaking, say the opposite. They say knowledge is ultimately based on sensory information we gather about the world. We have to experience the world before we can understand it. All hypotheses have to be tested to see whether they're true. We can't simply think in an armchair and learn something about the world. Theory must conform to data and not the other way around. Now I am a rationalist, and poker is a great example why. Let's start with the basics. There are objectively correct and incorrect ways to play poker. If your goal is winning hands and money, then certain techniques are superior to others. You can loosely categorize the techniques into two fields, mathematics and psychology. Mathematics is the straightforward one. Card games are fundamentally probabilistic. There are only 52 cards in a deck. If you're looking at the Ace of Spades, then you have a 1 in 52 chance of guessing it right on your first draw. The more cards you draw, the higher your chances. From a mathematical perspective, poker is simply a complex analysis of probabilities. What are the chances that your opponent has a better hand than you given the cards that you can see? The difficult part of poker for me is the psychological part. Master poker players aren't just calculating probabilities. They are aware of your previous betting patterns, any tells that you have your general disposition, your likelihood of bluffing, the way that you put your chips into the pot, etc. Plus, they're aware of their own psychological projections, how they think they're being perceived by others. The master poker player might establish a reputation of being loose at the table, meaning he plays weaker hands than he probably should. Then just when you think you've got a beat, he'll lay down a killer hand. Or vice versa, he'll establish a reputation of only playing strong hands, but then he'll get away with bluffing it just the right time. The goal is to outplay your opponent both mathematically and psychologically. You want your opponents to confidently bet into your strong hands, and you want to know when you should fold because you're beaten. Ultimately, what determines a strong hand from a weak one is mathematics. A strong hand versus a weak hand means that the vast majority of the time the strong hand will win. For example, say the cards were just dealt, you've got pocket aces, and your opponent has a 2-7 offsuit. If it's just the two of you playing, the pocket aces will win around 85% of the time. Now it's possible that your opponent could get lucky and make two pair, but it's not likely. Therefore, if your opponent bluffs and bets all his chips, the objectively correct thing to do is call him. But here's where things get tricky. Imagine that your opponent gets lucky. Two 7s come down, and he beats your aces with three of a kind. Ah, bummer, bad luck. Maybe you'll get him next time. But imagine it happens again. He keeps getting lucky over and over. Instead of 85% odds, you get him to bet everything with 98% odds in your favor. Total domination. But he gets lucky again and takes all of your money. This isn't unheard of. Poker players have some pretty incredible stories of so-called bad beats, where they get strings of terrible luck. Sometimes it costs them a gigantic sum of money. This happens because of variance. The inherent randomness in card games. It's like flipping a coin. If you flip long enough, eventually you're going to have an unusual streak of 20 heads in a row. Let's examine this a little deeper. After taking all your money, imagine your opponent is interested in understanding the game of poker, and he's a strict empiricist. You explain, my friend, the reason you won isn't because you were playing better poker. In fact, I was out playing you every single hand. You simply got lucky. I can fully explain why your moves were incorrect. And he responds, ah, but the data isn't on your side. Clearly, I kept winning, which means my moves were more correct than yours. If your theory were right, then you would have beaten me. And you respond, no, no, my theory fully explains why you beat me, and it isn't because you were playing better moves. I understand this despite losing all of my money to you. He responds again, well, why do you believe your theory is true? I am treating your claims as empirical hypotheses, and as such, all of the data contradicts you. Your theory should adjust accordingly. You respond, but I am appealing to the logic of the game. Just by analyzing the concepts involved in poker and understanding the related mathematics, I clearly understand why I played a better game than you. It was just variance. Eventually, my theory will be demonstrated as superior. And he responds, OK, prove it. Let's test your theory some more. So you play more games, and the exact same story unfolds. He keeps getting lucky and taking your money. You understand it's just variance, but he challenges you and says, at what point will you abandon your theory? If I keep beating you, won't that prove that your theory is wrong? Are you going to dogmatically cling to your theory, regardless of the data? And you again respond, no, my friend, you're just getting lucky. If you understood the game of poker, then you would understand why I don't have to appeal to simple empirical tests. I am appealing to the logic of the game. Ultimately, it doesn't matter how much I lose. I know that I understand what's happening. And so on. This is precisely the argument between rationalists and empiricists. The rationalists say you can know some things just by logical analysis, even if the data suggests otherwise. And the empiricists say, at some point, if your theory does not accord with the data, then you must abandon it. Notice how the rationalist and empiricist cannot make headway. The rationalist keeps appealing to his theory. The empiricist keeps appealing to his data. The rationalist won't be persuaded by more data. And the empiricist won't be persuaded by more theory. Neither side can prove his claims to the satisfaction of the other. The rationalist claims to fully understand what's going on, while the empiricist does not. The empiricist keeps his options open, just in case the data changes in the future. A rationalist might say, you can understand correct principles of poker, simply by sitting in an armchair and thinking about them. While the empiricist says, no, you have to actually play poker to get the real world feedback, thinking that you can deduce the principles of poker in an armchair is dogmatic and naive. Now, as a rationalist, I can't help but conclude that the empiricist is dead wrong. And he's sticking his head into sand. You can indeed understand certain truths in poker, regardless of the data. And no data will convince me otherwise. That isn't because I'm being dogmatic. It's because I see the logical framework behind the game of poker. And logical truths, as I keep writing about, are not open to revision. They are necessarily true. You can rationally deduce certain principles of poker for the exact same reason that you could know square circles don't exist. Logic. It's a matter of pure logic. You don't need to go out and test whether square circles exist. They certainly don't, nor will they ever. However, this is where rationalists can end up being dogmatic. There is one respect in which any careful rationalist needs to revise his theory. What's a more plausible explanation that your friend who keeps beating you even with inferior hands is simply defying the laws of probability over and over and over or that he's cheating? Perhaps it's the case that one of your fundamental rationalist assumptions, the randomness of the 52 cards in the deck, is mistaken. Maybe your friend is playing with a couple aces up his sleeve. If the rationalist isn't careful, he will be sticking his head in the sand and deny what is a much more likely theory that the poker he's playing in his head is not the poker he's playing in the world. The poker in his head is dealing with an untampered with deck of 52 cards and the poker in the world is dealing with a tampered deck. So if you have a theory that is not matching up to the data that you're gathering, it doesn't matter how logically airtight and pristine your theory is, you've got one of the axioms wrong. One of your basic assumptions is simply incorrect. Now this phenomenon of understanding the logical framework of a topic is not unique to poker. Every area of thought comes with an inescapable framework that you can analyze. You can call it conceptual analysis. It's the realm of pure logical reasoning, not empirical hypothesizing. Economics is my favorite example. A limited set of truths can be known with certainty. They are logically necessary given the concepts involved. There are plenty of empirical hypotheses in economics, but none of them contradict economic theory. You might even say that the theories are the presupposed conceptual framework into which empirical data gets added. Specific economic data doesn't necessarily contradict sound economic theory, just like the specific results of poker hands will never contradict the theory of poker. The data only makes sense in the context of the theory. It doesn't matter how many times in a row pocket aces loses to two seven offsuit, the truth is a simple matter of logic. But the same principle also applies to theories of physics, which is why I wrote my article on quantum physics. Simply by analyzing the concepts involved, you can know that the abused interpretation of the Copenhagen interpretation of quantum physics is wrong insofar as people claim that reality is logically contradictory because two particles are in a superposition. That is nonsense and it's a matter of logic. Now, note in this circumstance, pure logic doesn't tell you what the correct theory of physics is, it just tells you what the correct theory isn't. There was certainly a role for testing theories and coming up with concepts that are empirically derived rather than logically derived, but this is a topic for another article. The case for rationalism is most clearly made with strictly logical examples. In certain circumstances, you can indeed understand something regardless of the data and that doesn't mean that somehow you're an enemy of the scientific method. The resolution is simple. Not every proposition is an empirical hypothesis and some theories can be accurately deduced from an armchair.