Yes, but it's worse than you'd think, from what I understand. The bots will try to get more than one seat at a table and share information, so that it's even MORE unfair.
Cash games in person are pretty bad too. Casinos need to make money and they only get it off rake, so they have lately turned a blind eye to collusion. The only safe poker (except among friends) is tournaments, in person, so they randomize the tables properly.
Collusion in live poker games in casinos is not a widespread problem. There is a problem with poker where people always think they are being cheated every time they lose. If you are playing in a casino in person it is very unlikely you are being cheated. If you are playing in a regulated website online that verifies the identities of the customers it is also unlikely you are being cheated.
The vast majority of people that play poker absolutely suck and think they are being cheated because they lose money very quickly. Most bad poker players would literally be better off playing blackjack.
I used to play poker back in the 2000s. The online game was getting harder then and I can only imagine it's gotten worse? Also GTO solvers are a thing now? I don't know what stakes you are referring to but I feel like the overall quality of poker play has never been higher.
Played thousands of hours in casinos. Saw some asshole show down cards to someone still in the hand, stuff like that, but never anything I thought was collusion.
I've played for years, in locations all across the US, and know many other people who've played much more. This is basically not a thing. I encountered it once, decades ago, at Hawaiian Gardens in SoCal. If it's happening regularly it's very obvious, and the floor will quickly become aware of it. As far as making a living, it's easier to learn how to beat live low stakes than to successfully collude, and the latter would be unlikely to help much if you weren't already beating the games.
Surprisingly, no. Most sites do a good job of finding and banning bots.
It's also fairly easy to spot a bot. They will make odd sized bets at times. You check to see if that betting line is taken in a solver.
I'm a software engineer with 10+ years of experience. I'm also a poker player that has a very deep understanding of the game. Writing a poker bot that can beat the game is absolutely not trivial. There are "solvers" that use counterfactual regret minimization to solve a constrained version of the game for specific scenarios. These are useful for understanding the principles of the game but they are not the cheat sheet people think they are.
I think people fundamentally don't get that poker is not like chess. The vast majority of money I win is from identifying when players are too attached to their hand and never folding or when they just give up on their hand and fold to any bet.
I'm an ex online poker pro. You probably don't have the deep understanding of the game you think you have. Bots were already destroying the field up to mid-stakes 10 years ago.
I'm a current online poker pro but probably not for much longer. Bots are a serious and real problem and they do beat the games for a good winrate. But it's still possible to make money even in environments with some bots as long as you can find games with fish. And some games on geofenced sites (the OP said they play in Michigan) or other small pools don't appear to have bot problems.
I'm sure there is some, but it's standard practice to keep at database of all your hands.
I've sanity checked the winning regs (at my stakes) and they all make mistakes.
I think it helps that in the US all the sites are geography based. It makes it harder and less financially viable to run a bot ring.
I don't know what a reg is but I would assume the first thing you do after you get your bot working is to add some imperfect play so that you don't get banned by the most trivial heuristics.
Reg means regular, ie someone who plays a lot. Usually but not always it implies someone who is a winning or at least attempting to be a winning player
Yeah. Poker isn’t actually hard to play perfectly if you’re writing a bot. The hard part is making it look authentic and not emit any patterns or other signals that could be detected in the data.
It's exceedingly hard to play perfectly. Nobody knows how except in limited toy games, like heads up at 7BBs or less. And perfect play varies drastically from opponent to opponent, this isn't blackjack.
So it's a cat and mouse game like most things, except the spread between the cat and the mouse is quite high but is still (somewhat) in favour of the mouse as the online poker world currently stands?
Hmm… there are multiple variants of poker, at least one was weakly solved in 2015. I guess one could implement their algorithm. But I don’t know if the weakly solved variant of poker is popular?
Online poker is very much beatable. Poker isn't solved in the same way chess is. It also depends on the site and the rake. Some unregulated sites don't do KYC so collusion is possible.
What I meant by "poker isn't solved the same way chess is" is that if you take a solver and follow all the actions it tells you are "best" you will not make the most money. It isn't like stockfish where by using solver outputs you will automatically make more money than the best pros in the world. 99% of poker is understanding the unique ways your opponents are bad and adjusting your strategy to profit the most from them. Even the best pros in the world still make mistakes.
I used to work for an online poker outfit. The boss wanted weak bots populating the tables so that we looked popular. Of course, he had a “crack team” of bot writers for playing on “other sites” to make money, too.
I'm curious if it's possible to make a profitable GPT Poker bot, I have seen a few GitHub repos but not experimented with it. Obviously legal/ethical concerns there aside. In my experience you see a lot of the same names when playing and they could be bots. But you can interact with people in the game by chat or "throwing" objects at them in the game like a horseshoe or cards. And when they react back maybe that's a sign they're not bots. Regardless you want to avoid playing with the good people and seek out the bad players aka fish so you can play against them instead. Table selection is key.
Given how awful LLM are at chess, I'd say GPT sucks at poker. Making a profitable bot using state of the art poker software, like stockfish for chess? That's already done.
Sure you could use an AI agent to write that code, but it wouldn't actually being that AI agent in the hot loop constantly coordinating your UI interactions with the poker engine...
Comments like these make me feel a bit safer from AI in my engineering job. People think it's a perfect no brainer fit for so many inane situations.
Bots mostly don't play fair. One strategy I've seen is having multiple bots play passively to minimize losses, unless at least 2 get placed in a single game. In that case the bots can share information bully the rest of the table by playing aggressively.
The problem is specifically with unregulated sites that don't verify the identities of players. The bots aren't so much the problem as the fact that they can collude and share hole cards. But fyi, bots aren't actually good at playing poker outside of specific scenarios vs bad players or in scenarios where the decision tree is not large (ie short stack tournaments where the decisions are pre computed, you can imagine how massive your edge can be when you have a pair of 9s and you know there are already 3 dead aces and your decision is only all in or fold)
Not really. Maybe in very specific applications of limit (fixed bet size) hold'em, but no limit texas hold'em, the most popular variant online, is very much unsolved, especially in multi-way pots. There are simply too many variables and strategies involved to calculate quickly enough on the fly. For games like omaha, which uses 4 hole cards, this is even harder.
Due to advancement of theory and study and popularity over the last ~20 years though, it's definitely much harder to be successful than it used to be.
Heads up no limit holdem bots crush the best players in the world even 200bbs deep. So kind of like the same situation as chess. Not solved but not beatable.
There haven’t been any real online NLHE heads up games online in at least a decade. so it’s kind of a moot point. Not sure which bots or players you are referring to - these games don’t exist online in any meaningful volume.
It's not really my thing to give too much thought about macro-trends that are out of my control or worry about what negative consequences they might have on my life.
The short answer, I really don't know what this means for the future of the career of programming, the business of software, or anything else.
Instead of worrying about that I'm going to try to focus on the here and now, the upside potential, and the unique set of advantages that I have available to me to build something valuable, have fun, and maybe profit.
I'm going to do what I enjoy doing, try to learn some new skills and create things.”
I read doyle's Super System back in the day and used that as the basis for my poker strategy from high school to mid-twenties. In talking to some friends who play competitively, they say SS is just super out dated and you would get eaten alive at any cash game. I'm curious what, in your opinion, is the "standard" playing strategy that is most effective in today's poker rooms? I'm curious if that answer is different online vs in person.
The "standard" strategy is to play GTO (game theory optimal). There are solvers out there (like GTO Wizard) that show you the "optimal" play for every situation, which is used as a baseline, and then players deviate to exploit specific player tendencies.
GTO trees are far too complex to fully memorize, so nobody can play perfect GTO. But you can do a lot of solver work to get reasonably close.
I won a tournament on a cruise against a guy who I think was doing this. He was one of the people at the table who had a consistent tell and that helped me beat him heads up.
Super System is old, but I wouldn't call it outdated. Definitely still worth a read. The more books you've read, the bigger your tool set. The key here is that it works in both directions because each style you learn, is a style you may recognize other players using. The pitfall of Super System, is by now, everyone's read it, and it's quite easily recognized:
- limp-shoving under the gun
- always trying to go on runs
- over playing suited connectors (JTs specifically)
But, you still get the advantage of being able to recognize it. There's lots of good wisdom in there that isn't as prescriptive either. Read as many books as you can. Poker is information warfare.
No idea, obviously you can see I'm a n00b and not a profitable poker player but on the strategy side, and this is very oversimplified, but you should be folding a lot of the time other than when you get AA-22, AK, AQ, AJ, AT, KQ, KJ, QJ, JTs, T9s, 87s you call or 3 bet pre-flop because you have good odds. When you're up against tight players you can make a small bluff on the flop and scare them away most of the time, if they raise you fold though. Position is very important in the game, when you're on the button you have odds in your favor because everyone else has to check or bet before you so you play more loose and aggressive in that position and more tight and passive in early position. There is no one single strategy to memorize and apply, that's why it's great. 5 minutes to learn the game, a lifetime to master.
> No idea, obviously you can see I'm a n00b and not a profitable poker player but on the strategy side
Sorry, why are you answering a question if your first response is “no idea”? Am I missing something? If you have little information, my feeling is that your response is at best just BS? I know that sounds very rude, I’m sorry for that.
You’re probably right, I am just sharing what I have been studying and from my experience playing but I’m a losing player so it should be taken with that context.
I don't think so. I don't play nearly as often as I used to, but I still do alright. Though, I usually play tournaments where "game theory optimal" can only get you so far. There is a lot more nuance in a tournament where your style should change as it progresses.
Probably a dumb question but when I watch poker on TV I see that the aggressive players tend to win, so why do the losers let themselves get intimidated?
Aggression generally wins the day but pay attention to the size of the bets and their position. Generally, you want to make a lot of small bets to show action (they can fold for a minimal investment) and leverage table position to make other players make hard decisions.
Oversimplifying for sure but if you're loose and aggressive against a tight aggressive player, you're going to make them fold most of the time and win a small amount by applying pressure on bluffs but every once in a while if you get too aggressive and they call you because they have a monster hand then you get wrecked
what do you think about OP's 40% VPIP? It seems to me that in low-stakes online play you'd want to play tighter than that, but I know very little about poker strategy beyond what I've absorbed from seeing people talk about it.
It's a fairly meaningless stat without knowing the number of the players at the table. At a quick glance he seems to be playing 6-max, but sometimes 3-handed. In any case 40% is within the reasonable range for 6-max.
Assuming they are playing 6 max with full tables 40% vpip is egregious and I do not see how they could have a winning strategy playing like that. (Looking at their results they are not winning).
To be fair, 1k hands is a pretty meaningless sample - I think most pros would say you need at least 50k if not 100k hands for the results to be any reliable signal as to whether or not a player is actually winning or losing in the long run.
Have you cross-referenced with the other hand trackers whether the numbers add up? Alternately, could someone explain why wouldn't a LLM hallucinate with numbers in an application like this?
Yeah I used PokerTracker 4 to cross reference and kept working with Cursor until it got very close like within 1% of accuracy but there are still some edge cases I might not have found yet. In the beginning it was hallucinating a bit by “estimating” what the percentages “should be” etc but I kept working it until it was doing things right.
> Then I started building my own Python script automations to export my hand history from PokerStars, import it into PokerTracker 4, check my balance, stuff like that.
If it works like it did with ASR (Advanced Speech Recognition) back in the day, then doesn't the app now have all of your decision bias? Restated, isn't the app a reflection of how you play poker, not how an AI would play if it were truly artificially intelligent?
If you're looking for a tool that may be a bit better than Cursor for UX, you could potentially look into Lovable. If you know what you want and the proper design terminology, you can potentially make some slick looking UIs.
Potentially. Like the other poster said, you can go through Github, but if worst comes to worst you could take screenshots of what Lovable makes and have Cursor modify the UI based on the screenshots. If you're using Sonnet 4.5 as a model it should be able to handle it pretty easily.
You can have Lovable do the design and then sync it to a GitHub repo. Then you can pull it down locally and further refine with Cursor.
Just be aware that under the hood Lovable is strictly react (or at least it was the last time I checked it) so that might be a important variable to consider since I saw that you were using Laravel.
Yup, sw engineering is a slow march to being commoditized. Some things will remain hard (only because it's cutting edge and pushing the limits of something) but known patterns and services will be just-yell-at-ai to stand up. A lot of businesses can run on the latter, i guess - but at that point the challenge is having a viable business, not the software development of X.
I still think that might be oversimplifying what software creation is which is being able to explain to a computer what it is you want. I think of Cursor as Python was to C. It's a higher level language but you still have to be able to think like a hacker, which will always be a rare skill.
And the best hackers at any level abstraction will always be the ones that actually understand what's going on in every lower layer in order to diagnose when the abstraction is failing them. Anyone that thinks you can be up at the level of vibes without understanding how an LLM thinks, without knowing how to review and factor your vibed Python or whatever high-ish level language, can make it performant without knowing if or when to write something in lower levels like C or need to be using a library where all the hard work is in something like C, make it secure without understanding how that gets turned into instructions for an incredibly obedient but ignorant machine (like the LLM is but in the exact opposite ways, buffer overruns and free before use and stuff)... It's a holistic practice. The guys that produce code and don't know which parts are happening in the browser or in the client, think they can trust the values of cookies not to be tampered with and junk are able to be productive, probably more so with LLMs these days but they simply can't make quality software and never will be able to. Corporations love them because nobody's accountable to resiliency (securty, quality, reliability) until something actually breaks and those guys can get thrown under busses easily when and if that happens because they're cheap cogs. Hackers love them because we'll always have work to do to improve (or compete with, or exploit) what folks like that make.
our industry has existed on the cutting edge doing what's hard since its inception. it's just that there was a time when sending a piece of text across a wire was hard. Now that's easy, so we do more with the tools that make that easy. When what's hard today becomes easy we'll do that quickly with the tools that make it easy and then do more hard stuff. We can say we've achieved AGI when the tools are doing better on their own than a tool plus an engineer would do, and I think that's a long way off.
Exactly. This is how it's always been. LLMs make it easy to spit out boilerplate code, which drives the price of boilerplate down to free. But good engineers will add a lot more value to that which raises the bar for everyone. The things you can create with an LLM become boring and worthless (honestly they mostly already were before coding agents came out) and the hirable skills become everything else that engineers need to do.
The vast majority of people that play poker absolutely suck and think they are being cheated because they lose money very quickly. Most bad poker players would literally be better off playing blackjack.
Played thousands of hours in casinos. Saw some asshole show down cards to someone still in the hand, stuff like that, but never anything I thought was collusion.
Plenty of other angles, though.
I've rarely met players whom I think could even act properly on the knowledge.
I've also never been in a room that wouldn't take it very, very seriously.
Even some people who are victims of scams admit that at the time they sent some/all of money they knew it was a scam but did it anyway.
I think people fundamentally don't get that poker is not like chess. The vast majority of money I win is from identifying when players are too attached to their hand and never folding or when they just give up on their hand and fold to any bet.
How so? It's not like the entire game state is visible to all players, and a big part of poker is playing the other players.
Writing a winning poker bot is not trivial, you are unknowingly spreading false information.
https://en.wikipedia.org/wiki/Cepheus_(poker_bot)
The major poker sites claim that they have really good (and very top secret) bot detection. I'm skeptical.
Sure you could use an AI agent to write that code, but it wouldn't actually being that AI agent in the hot loop constantly coordinating your UI interactions with the poker engine...
Comments like these make me feel a bit safer from AI in my engineering job. People think it's a perfect no brainer fit for so many inane situations.
This decline was underway a full decade before bots really came on the scene.
Due to advancement of theory and study and popularity over the last ~20 years though, it's definitely much harder to be successful than it used to be.
“ What it all means for the future
It's not really my thing to give too much thought about macro-trends that are out of my control or worry about what negative consequences they might have on my life.
The short answer, I really don't know what this means for the future of the career of programming, the business of software, or anything else. Instead of worrying about that I'm going to try to focus on the here and now, the upside potential, and the unique set of advantages that I have available to me to build something valuable, have fun, and maybe profit.
I'm going to do what I enjoy doing, try to learn some new skills and create things.”
GTO trees are far too complex to fully memorize, so nobody can play perfect GTO. But you can do a lot of solver work to get reasonably close.
- limp-shoving under the gun
- always trying to go on runs
- over playing suited connectors (JTs specifically)
But, you still get the advantage of being able to recognize it. There's lots of good wisdom in there that isn't as prescriptive either. Read as many books as you can. Poker is information warfare.
Sorry, why are you answering a question if your first response is “no idea”? Am I missing something? If you have little information, my feeling is that your response is at best just BS? I know that sounds very rude, I’m sorry for that.
Are today's online tables simply impossible to win? (bots, collusion)
Or are players simply too evenly matched and the house rake/fees kills you anyway?
If it works like it did with ASR (Advanced Speech Recognition) back in the day, then doesn't the app now have all of your decision bias? Restated, isn't the app a reflection of how you play poker, not how an AI would play if it were truly artificially intelligent?
Just be aware that under the hood Lovable is strictly react (or at least it was the last time I checked it) so that might be a important variable to consider since I saw that you were using Laravel.
https://andreasthinks.me/posts/ai-at-play/
On top of that, LLM output is so mediocre that even marketing firms are doing most “copy(s)” by hand.