Can my bot defeat or resist the GTO bot?
This little guy can win with anyone. Something tells me that he can at least "resist" the attack of the perfect GTO bot. Each game would have to be played with antes and only 26 antes per stack, but I think these would still be fair.
I'm currently test-playing the bot with physical cards, but since there's a lot of shuffling and counting going on - it goes kind of slowly and requires some effort. But he plays like a maniac !!! I can win with him, but only if I'm very lucky with the outs or when my crazy actions turn into miracles - at least a couple of times per match. I think I'll focus on the 3 game matches and post some results later.
Please check the graphic, it explains the whole program this AI is based on. Yes, it's a very simple program, but also very tricky and smart in its structure.
Introducing the DPU 5200 heads up NLHE bot:
Sorry for small letters, please click the bar to see full resolution.
Simple GTO bots are built not to be exploitable not built to exploit. A simple one line program that always says all in resists them and defeats them a certain percentage of the time.
I play very well against bots and my bot is really difficult to beat. Considering that the GTO bot wouldn't be aware of my bot's strategy (like I am), I think DPU 5200 would do pretty well and resist in long run.
But if GTO is eternally unbeatable no matter what, then ok, maybe it could also beat my creation.
But what if DPU 5200 can get a better long run results than any human vs perfect GTO ?
I was badly humiliated by DPU 5200 a few minutes ago. It wasn't a long match (who wins 3 games, wins the match), but I just COULDN'T DO ANYTHING !
I lost 1 : 3 so you could say: "Hey, you've won 1 game, so you could also have won the whole match !" Wrong. I've won 1 game because I've been dealt black aces in the 1st hand, so he had to do a mandatory call to my pre-flop all-in. Black aces won, so one point for me. Other than that - misery and pain.
It feels like playing against a wall, a wall that's entirely unpredictable.
From what I noticed yesterday, DPU 5200 can be outplayed even 10 games in a row, but if he wakes up and he gets good cards (or you get bad cards), then you're in trouble. And this is exactly what happened yesterday:
I was really lucky in a couple of key spots and I won three matches in a row (it took about 1,5 hour total to complete them all):
(3 : 0)
(3 : 0)
(3 : 0)
So at this point you can see that I was really doing my best and played well. No mercy for the bot.
Then I said: "One more 3:0 score for me and this will mean that the bot is flawed." I started one more match and this one match took over 2 hours to finish !!! It was some high level intense poker, trust me.
I won that match:
(3:2)
But I think I won only because I flopped quads in the very last hand. And guess what - I was 1 card away from winning it quickly 3 : 0, but he hit a 5-outer.
CONCLUSION: The bot doesn't play at a constant level. He can go from dumb to genius, just like a human (!).
It is alive. It kind of reflects the player's current state of mind - the more you want to win a particular match, the more difficult it will be to defeat him.
In other words, the more afraid you are of losing, the more likely it is that you will lose.
Implying i suppose that the reverse is also true and DPU 5200 can be reverse out-psyched with raw bravery and arrogance?
I am not afraid of losing to the bot. I am not afraid of losing to the bot.
Free coaching there.
If that's all you need to do to defeat this so called 'poker brain', which you already beat quite easily it seems, then one need only stealth in the occasional dud 2♣ with some ludicrously brazen and confident vibes and the dunce deals itself into mathematical oblivion.
Well, yesterday he also beat me easily 3:0, so it may go the other way as well.
It works like this:
You can go all-in in the first hand, pre-flop against his random 2 cards, maybe he will call you and maybe you will win. But you don't want poker turn into roulette, right ? So the more time it takes to end the game - the more you want to win and the more cautious is your game then. You don't want to waste your time. And this is where he starts to play real poker. Today I played 1 game that took 45 minutes. I was happy to win it. And it wasn't a match, it was just 1 game.
I didn't beat him easily last time (penalty Ceres, please read that post again). I was 1 card away from doing it and then he was back from almost dead and showed some serious skills.
The game is No Limit play money, so it all kind of depends on your approach. I play quickly sometimes, but only when I get tired with shuffling and counting.
Bot is gangsta, I have no doubt about it. Know what I'm sayin' ... 😉
That's a solid no-go on the reread soz. As a rule if I'm replying to one of these bot type threads I've required at least three to four deep water bongs to settle my crippling poker bot anxiety.
Truth? If your bot was real 'gangsta ' it would roll me a spliff.
Woaaaah woah woah there officer i was only JOKING haha 
see? just minding my business over here.. oh is that the time? ..must dash!
I've just completed 2 games, I won one and lost the other, each took about 30 minutes, but I want to focus here on something else. While I was playing a question came to my head and my thinking voice asked me: "How much more time are you going to waste pretending that this is just a fun game ? 5 days, 2 weeks or maybe a month ? Can't you see that this is something more ?" And then I jumped into this "zen" type state of mind and focused even more on the game. Then the bot started to play like god. So he kind of responded and reflected my energy back to me. Yes, I'm a good player (occasionally), but so as the bot is. Full spectrum of poker strategy. And real emotions.
Bottom line - the better you play, the better he plays. Simple as that.
I think this phenomenon comes from the fact that:
- if you focus on his card duplicate system, sooner or later you're going to make a big mistake that comes from ignoring basic poker strategy
- if you focus on general poker strategy, sooner or later you're going to make a big mistake that comes from ignoring his system
- and if you focus 50/50 on both things, then if you lose, you're going to complain: "I should have focused on one of the two"
DPU 5200 is really a small revolution. I've never experienced anything like this, I have to get some new plastic decks. What more can I say... he is alive, like Ceres have said.
It's never too soon for a good hint 😀
Introducing the APU 5200 heads up NLHE bot:
Please click the bar if you want to read.
I think the gangster move here would be APU 5200 live-stream crushing highstakes. buying out 2+2.com, and then brute forcing its avatar on randomly selected accounts.
I've noticed that DPU 5200 prefers doing it slowly, while APU 5200 focuses on quick action.
It feels great to be humiliated by both bots, then at the end of the day I can say to myself: "It was a good day !" 😀
I'd love to see Tom Dwan playing against DPU 5200 and Phil Ivey against APU 5200.
Yesterday I lost a game, I don't remember which bot I was playing with, but after I lost I think I made the most stupid face ever in my whole life, hahahah.
I wish I knew whether these two bots make me more a poker player or more an author. (?) Guess I'll have to split in two.
One common idea nowadays, that I first heard from Sauce (
), adapted to a bot vs bot scenario, is having your bot run on gtowizard ai where, as long as the loss in ev is minimal, it will always choose the sizing that any standard, static GTO bot would never see as the optimal one, and solve for it in real time, maybe an in-between sizing etc, that would force the static bot to try to wing it from his libraries. I would bet the static bot would end up in a situation where he is the bad reg as portrayed by tombos in another post, and it is possible to gain an edge over him doing things this way.
Maybe also deliberately leading the game into nodes that are supposed to be the most infrequent ones.
How big would be that edge, if it truly exists? No idea.
Interesting. What do you mean by "wing it from bot's libraries" ? Do you mean that it will be forced to choose from pre-set game tree scenarios ?
Yes. It has in its libraries BTN vs BB SRP, let's say, X, B33, B50, B75 and B125 for the flop cbet, but faces another bot on the BTN who used B102, perfectly balanced, and he has to resort to computational heuristics and approximations to figure out how to respond because his solution does not have that exact sizing.
