Live data from Hacker News

Show HN: Play poker with LLMs, or watch them play against each other

llmholdem.com

81–90 of 101 posts

Re: Show HN: Play poker with LLMs, or watch them play against each other

#81
post #19

I used to play professionally, and I still play in the casinos. These LLMs are playing better than most human players I encounter (low limits). They're kinda bad, but not as criminally bad as the humans.

I just watched for 5 min and no they don't play very well. Deepseek squeezed with K4o against CO open and BTN call with full stacks. Grok 3b AI with 25bb in the button with Q4s. Those are very far from optimal play which is well known since solvers. I wonder how they've been trained.

You and OP are agreeing: "Better than most human players" is quite possibly the lowest skill bar in (at least live) poker.

Re: Show HN: Play poker with LLMs, or watch them play against each other

#82
post #27

Earlier quoted context omitted.

OK so you know how it goes in poker and I should probably read the literature ... How much of a session is based on "reading players" vs "playing the odds"? What I am getting at, is how different is poker than say roulette or blackjack? My initial thoughts are that poker such as TX hold 'em is not a game offered in a casino, so it must be mostly indeterminate. I imagine that the casino versions of poker are not TXHT.…

> My initial thoughts are that poker such as TX hold 'em is not a game offered in a casino Why not? Because you think it's a game where the casino can lose? If so it's not an issue, as casinos that provide poker take "fees" from the stakes. Like how stock exchanges work: there are people making or losing money from stock market, but exchanges are always making profit.

Around where I live, about half the casinos offer poker and about half don't. Poker can be a pretty high cost to a casino. Compared to something like slot machines, it's financially mostly downside: You need a lot more physical space per player. You need more staff. In order for players to come and actually enjoy it, the poker room needs to be located in a relatively quiet corner, ideally enclosed so the buzz of the rest of the casino can't be heard, which is also expensive. And the game is slow, and rakes happen once per hand, so you're making money pretty slowly. And that's just for cash games. Tournaments are worse. If I had to guess, tournaments probably lose money for the casino, and they only exist to get players in so they play at the cash tables. Probably many other things that I'm forgetting because I don't run a casino.

Re: Show HN: Play poker with LLMs, or watch them play against each other

#83
post #19

I used to play professionally, and I still play in the casinos. These LLMs are playing better than most human players I encounter (low limits). They're kinda bad, but not as criminally bad as the humans.

I'm actually surprised at how well they play pre-flop (mostly). Did some initial analysis on VPIP/PFR across positions, and somewhat decent. Post-flop on the other hand is all over the place...

there plenty of published preflop charts and GTO ranges

in fact, a fun project would be take a non-reasoning model, play on a lesser known game format, and see if it learns an "a ha" moment or explicitly simulate moves ahead

Re: Show HN: Play poker with LLMs, or watch them play against each other

#84
post #21
post #10

Earlier quoted context omitted.

I suspect this would only matter much if they also remembered (and cared about) their own prior play.

Not really. Only as far as their table image mattered--in this case, zero. Otherwise, you can and should ignore your own past play. What I'm curious about is if their innate training is enough to give them biases. Like maybe they think Grok is full of shit.

> Not really. Only as far as their table image mattered--in this case, zero.

Right; there's feedback to it. When humans play poker, they do so with common knowledge of the fact that humans have object permanence and can recognize and remember their opponents. The same thing that motivates "profiling" a villain, motivates attempting to project a table image, which in turn motivates being aware of the table image one is projecting.

Re: Show HN: Play poker with LLMs, or watch them play against each other

#85
post #27
post #19

I used to play professionally, and I still play in the casinos. These LLMs are playing better than most human players I encounter (low limits). They're kinda bad, but not as criminally bad as the humans.

OK so you know how it goes in poker and I should probably read the literature ... How much of a session is based on "reading players" vs "playing the odds"? What I am getting at, is how different is poker than say roulette or blackjack? My initial thoughts are that poker such as TX hold 'em is not a game offered in a casino, so it must be mostly indeterminate. I imagine that the casino versions of poker are not TXHT.…

> How much of a session is based on "reading players" vs "playing the odds"?

Several good answers here, but I will add my own take.

“Playing the odds” is basically playing good, fundamental poker. This is a baseline that most players will use when sitting down at a table of unknown players. This is often called a “balanced” strategy (note some people erroneously call this “GTO strategy, assuming the balance, which is not actually what GTO is).

“Reading players” is a thing, but it can be broken up into (at least) two categories: 1) physical tells, and 2) player habits.

Physical tells is not a big thing. Some people give off a lot of tells, but some folks are also decent (not good) at giving reverse tells. Honestly, you can be a wildly successful poker player while knowing nothing about physical tells.

(Side note: One of the most reliable tells is bet timing tells. This can sometimes even be a tell online, especially when people are taking shots at higher stakes or are deep in a tournament. It can also be faked, but some folks are super reliable with timing tells, and they don’t realize it.)

The second kind of tell is player tendencies — things like when players play too many hands or when they play too few (e.g., fold too often).

One very reliable way for good players to smooth out their earnings curve is figure out which players fold too often and in which spots. Once they’ve figured that out, they try to set up that spot and basically print low-risk (sometimes even no-risk) chips.

Taking advantage of these tendencies is called an “exploitative” strategy (as opposed to the “balanced” strategy mentioned above).

Really good players can take rec players on a journey through a series of emotions (and accompanying predictable gameplay) such that the good player can read the rec player like a book. The odds tip heavily in favor of the good player at this point.

Pro player and strong amateur players are so far ahead of recs in ways that the recs don’t even realize.

Re: Show HN: Play poker with LLMs, or watch them play against each other

#86
post #27
post #19

I used to play professionally, and I still play in the casinos. These LLMs are playing better than most human players I encounter (low limits). They're kinda bad, but not as criminally bad as the humans.

OK so you know how it goes in poker and I should probably read the literature ... How much of a session is based on "reading players" vs "playing the odds"? What I am getting at, is how different is poker than say roulette or blackjack? My initial thoughts are that poker such as TX hold 'em is not a game offered in a casino, so it must be mostly indeterminate. I imagine that the casino versions of poker are not TXHT.…

Just an fyi they make apps like https://apps.apple.com/app/id1530767783 to train on betting based on expected value.Not the whole picture but trains the math side.

Re: Show HN: Play poker with LLMs, or watch them play against each other

#87
post #72

I just saw GPT 5.2 do something absurd. It has a crazy amount of money ($26k) but folded with a 4-pair before the flop. That's insanely conservative, when it would have cost just $20 to see the flop. But even worse, on the very next hand it decided to place $20 down with a 5 and 4 of different suits. In fact, all of them love folding before the flop. Most of the hands I'm seeing go like - $10 (small blind), $20 (big…

Modern poker (which tbf not sure if these LLMs are acting according to modern GTO or not) is highly dependent on position. Things change a lot too when/if you are in SB/BB.

Yes, the prompt tries to get them to play GTO. I do think their preflop play is the closest to mirroring this compared to postflop behavior.

Re: Show HN: Play poker with LLMs, or watch them play against each other

#89

Earlier quoted context omitted.

Modern poker (which tbf not sure if these LLMs are acting according to modern GTO or not) is highly dependent on position. Things change a lot too when/if you are in SB/BB.

Yes, the prompt tries to get them to play GTO. I do think their preflop play is the closest to mirroring this compared to postflop behavior.

Is this tuned to tournament or cash GTO? To the OP's shock about pocket 4's (I think this is what they meant by 4-pair(?)), folding 4's pre flop in early position to no raise would be fairly standard in tournament GTO (although the stage of the tournament and # BBs can change things up significantly), but less standard for sure in cash (almost never probably).
Post reply on HN