Hacker News Reader

Limit hold'em is popular, and can be found in almost every card room as well as online.

HU limit is probably only online, and theres a fair amount of bots prevalent in the games already today.

> The article also mentions that it's opponent was another bot that was also playing a very strong strategy. That suggests it's not really able to adapt to individual play, which is essential to being at profitable at all but the lowest levels of poker.

The computer bots attempt to play Game Theory Optimal - meaning it plays an (optimal) strategy that does not lose overall, regardless of stake/opponent. Against another GTO player, it will breakeven - otherwise it will be profitable. That is the holy grail of poker (and any game). If you are not playing GTO, there are leaks/mistakes in your game that can be exploited.

Check out https://www.youtube.com/watch?v=VHcrsMPQtgo , a poker theory course that talks about solving a simplified poker limit game

The concept of Nash equilibrium doesn't apply to a poker player facing two or more human opponents at the table, since they are likely not playing optimally or even non-cooperatively, e.g. if there is a communication channel such as facial expressions (however poker-faced) that a robot player is not privy to.

In the case of a poker tournament where there are at least a few non-optimal players, it's advantageous to exploit and take chips from them before other players do. A player who plays as though all of his opponents are playing optimally probably plays too conservatively in that early stage and no doubt faces a significant (non-optimal) disadvantage. Your comments still apply to heads-up (two person) poker, though.

Limit Hold Em pales in comparison to No Limit at every cardroom I've been in the past 10 years. Yes, limit still gets spread, but the limit choices are very poor in most casinos.

I play poker in the card rooms around the Bay Area very regularly, and limit poker is absolutely popular at the card rooms that I go to, as well as in Vegas. To say that it is dead is completely untrue.

It's my understanding that limit is popular at places like the Oaks because No Limit was not allowed for a while, until there were means of getting around it (Oaks has 'cap no limit' for example -- can't bet above the max table buyin; Bay101 has something similar). I see the current popularity as a function more of inertia than of actual interest. A majority of the tournaments and SNGs that the Oaks runs are no limit.

This is also not true. A lot of people enjoy playing limit poker, like 3-6, because they can chase big hands much cheaper. It's a completely different strategy than no-limit, where you could potentially risk your entire stack in any hand you're in.

The other thing is that poker has an unexpected (at least to me) social aspect to it. The people who play against each other are largely friendly to each other, and they play against each other every day so playing 3-6 limit lets you spend more time at the tables than 1-2 or 3-5 No Limit, etc.

California poker isn't typical of most other regional casinos. Look at the Bravo poker app and browse through what games are spread across the country. I have fond memories of limit poker thanks to the old 6max games online, but limit just isn't remotely popular anymore.

>That suggests it's not really able to adapt to individual play, which is essential to being at profitable at all but the lowest levels of poker.

There are two main categories of poker strategy: exploitive and balanced (also known as game theory optimal or GTO). With an exploitive strategy, your goal is to find the mistakes your opponents make (eg their tells) and figure out how to profit from them. With a balanced strategy, your goal is to 'balance' the choices you make (eg don't have any tells) so that it's impossible to your opponent to gain any information about your hand.

A perfectly balanced strategy will never lose money over a sufficient sample size, but it also limits the amount of money you can win. Exploitive strategies will maximize the amount you can win at the risk of being exploited by other players. From my understanding, your best bet against the weakest players is to play a purely exploitive strategy, and against good human players, you'll want to blend exploitive and balanced strategies.

My understanding of poker bots (the Alberta group from the article and PokerSnowie are the best-known) is that their goal is to develop a true balanced strategy. This may not be the way to maximize profits, but as long as humans aren't capable of playing perfectly balanced, the bots will slowly win in the long run, even against top pros.

"making a bot that can beat a full limit table is exponentially harder "

Hard but not impossible. As long as 5-6 years ago there was a number of bots[1] beating mid stake (15/30 & 30/60) 6 max limit holdem on Ultimate Bet (before it imploded). The games were fairly soft but pretty aggressive and those bots were taking a lot of money out of the game. They were also very frustrating to play against.

[1] They all had incredibly similar playing styles so were likely operated by a single person/group.

This is just completely backwards. Full ring limit is much easier to play optimally than heads up. A bot could break even with a good starting hand list and basic actions. Heads up play involves many many situations where the optimal plays are highly non-intuitive and involve both exploiting patterns in your opponents play and "advertising" for the future. This is because both players /usually/ have weak hands. Heads up limit poker requires a very strong understanding of game theory, not just "poker theory". Combine that with the fact that if you have an exploitable leak, it will get exploited over and over again.

Heads up limit poker was /quite/ popular before the US cracked down and online gaming and many players made a ton of money off it. Towards the end, it was clear botters were getting in on the action, and bots were already playing an extremely difficult game.

You are correct that no-limit, especially heads-up no-limit remains a tougher nut to crack from an AI perspective, but this is pretty damn impressive

NB: I played online poker professionally for a few years.

One of the most famous games of all time in the poker world saw swings of $10M+ in each direction and was exclusively heads-up, limit HE: http://www.pokernews.com/news/2006/02/beal-corporation-finis...

> I don't see a bot being close to be able to conquer heads-up no-limit, let alone a full table of no-limit.

PokerSnowie plays NLHE, but only uses 3 bet sizes (half, full, and double pot) to make the number of decisions manageable. I believe it plays heads up and 6max. Of course, I don't know if we have any evidence to suggest it's "conquered" those games, and we don't fully know how much the fixed bet sizes limit its potential.

http://www.pokersnowie.com/faq.html#brain1

Limit is almost a different game than no limit holdem. Simply put: Bluffing is not a problem you need to solve to win in Limit holdem.

Completely wrong on both counts. It is quite literally a different game, and bluffing is arguably more important in limit than in no limit.

Could you elaborate on this? Maybe the group I play with just sets the limit too low, but it's almost to the point that bluffing is pointless, and you only win with good cards. We usually play well into the night, so 5 or 6 hours. On most nights, there are less than 3 hands that don't result in showing cards. What good is bluffing if the threat of losing isn't great enough to make people fold?

It's a nice social outlet, but I would definitely prefer no limit.

Oh. Yes, low limit full ring games are quite different. If every hand comes to showdown you are all playing loose, and the correct strategy is indeed tight aggressive.

The article and my comment are referring to a two player limit holdem game vs an opponent who does not make mistakes. So I presume this is equivalent to a high stakes match with two professionals.

Playing no limit, a bot would have to handle an opponent who makes a big bet in an attempt to induce the bot to fold.

Moreover, bluffing isn't as crucial in Limit b/c you have little leverage to push somebody off their hand. Many people will chase because the pot odds or their own playing style demand it.

Finally, NL or Limit are both poker. In fact, they're both Texas Hold'em. So "quite literally" i respectfully disagree with your assessment.

I'm not a professional poker player, but I did work with one to build a bot about 9 years ago on top of WinHoldem. It was more or less a break even bot on PokerStars if left to play itself, but its real value to that team was as a force multiplier. It made multi-tabling easier, raising interrupts whenever the bot woke up with a hand.

I'm not one for online debates, so feel free to have the last word. To put it another way, I check in the dark.

I guess chalk it up to semantics? I could happily argue that online NLHE and live NLHE are also two different games. Like comparing tennis to squash - similar, but not the same. Of course, this would be better argued on 2+2.