Could a room of infinity monkeys beat a GTO bot?
Thought experiment.
Let’s say you had a match between the planet’s current best GTO bot:

And a room full of monkeys smashing buttons for infinity:

*Let’s say the monkeys win if at any point >100k hands their WR surpasses the GTO bot.
As I understand it the infinity monkeys would always win, and easily, for a few reasons:
- 1. eventually infinity would outplay the bot (approximate Nash to a higher %) for longer than it couldn’t. It’s just theoretically inevitable if ‘forever’ is the yardstick.
- 2. there is no such thing as infinite electrical power. Bad news for things that need electrical power to exist and participate eh?
- 3. eventually one of the monkeys would figure out how to yoink the bot’s plug out the socket anyway. GAME OVER mr smarty pants/+1 for opposable thumbs
24 Replies
Yes, obviously.
A more interesting question would be: "How many monkeys would you need to have an x% probability of succeeding?"
The infinite monkeys would effectively try every possible line, and brute force a more accurate GTO solution than a current generation GTO bot.
My understanding is most current solvers use "counterfactual regret minimization" to limit computation time and computing resources to reasonable amounts.
So they kind of test a line, find that it loses EV, and adjust the response until the computer arrives at a reasonable approximation of a Nash equilibrium.
When you plug accuracy parameters into the solver it's programmed to solve for a reasonable solution within those parameters.
The infinite monkeys could find potentially (slightly) higher EV lines. Their solution could also potentially be quite different from the solver, as it's possible to have multiple Nash equilibriums. The monkeys would eventually try every possible bet sizing, so they could find novel bet sizes that outperform in certain spots.
Solvers like PIO operate within limited bet sizing parameters. I know the bet-sizing parameters of more cutting-edge technologies like GTO Wizard are more robust, but I'm assuming they still must limit bet-sizing options that are solved for due to limited computational resources.
More importantly, Infinite Monkeys would make a great screen name.

You lost to Infinite Monkeys.
Lol
Zamadhi I’m going with 50 monkeys for a 50% success rate
But don’t ask me for my calcs. I’m busy. Buying monkeys
Playing a more precise GTO strategy doesn't guarantee an edge.
Having an edge doesn't guarantee a win after 100k hands.
In the grand scheme of infinite button clicking, it is more far likely that the first winning monkey simply out-lucks the bot.
Anyone wanna buy any monkeys?
I think what he means is that there are monkeys constantly jumping in his head and that is for two reasons:
- he feels he invented solution to solve GTO bot (which could be true)
- they are infinite
So if this is like a small discovery, those monkeys will keep jumping forever in his thoughts. If the discovery is confirmed, visiting a psychologist could be a smart option 😉
After running some tests I’m of the opinion ONE cocaine infinity monkey would beat the bot significantly quicker than a whole roomful of non-cocaine infinity monkeys.
The cocaine infinity monkey could maintain absurd frequencies across the tree with a maddening consistency, causing bedlam to all the bot's recalibration algorithms. Timing tells would be an absolute joke. And the whole game would fizzle and flow much quicker. And especially if the infinity monkey starts to associate the buttons with additional cocaine.
Are infinite poker monkeys impossible to defeat ?
I guess it would depend on the type of infinity.
A countable infinity monkey is going to lose big time to an uncountable infinity monkey you'd think. The uncountable infinity monkey able to branch off infinitely into the multiverse for edge.
What could make the whole internet go down and crash because of the data overload ?
Spoiler
Infinite Ceres
GOD HELP US ALL
Infinity hammers. That's how you defeat a bot quicker than an infinity monkey. Nothing can beat an infinity hammer at anything: chess; backgammon; smashing things; curling. They would have Pluribus on soft foods by the year 10^74
The perpetual linearity of time: 2
Stupid dumbass vain 'frequency' bot: 0
Oh ****, I just realised the heat death of the universe invalidates the entire experiment :(
Alarmingly accurate. So why am i not crushing online poker yet with my experimental philosophy horn and miniature veracity cymbal?
a new thread title emerges from the darkness..
Ahhhh... now you tell me. +88.17/100bb today
This is really a language philosophy thought experiment on the miscommunication of imprecise language.
The less precise strategy in this context is obviously not optimal by definition.
I have been out of poker for a long time and it is interesting coming back to poker to see the complete life of its own "GTO" has taken on because of marketing and persuasive language.
Not to mention the bot would probably be doing its own form of infinite monkeys with monte carlo sampling at some point unless we mean infinite monkeys against infinite computing resources .
At that point we may as well just ask if the god of the Christian bible could bluff a "GTO" bot if we aren't going to bother to constrain things to reality.
Can I say what an honour it is to be this good at generating EV whilst also ruling over everybody else and having much better clothes and hair than everybody else?
Good. I needed to get that off my mantle. Now, has anyone seen my trousers anywhere?
First hand history in:
🤖: 100 BB
🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈: 100 BB
🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 posts SB 0.5 BB, 🤖 posts BB 1 BB
Pre Flop: (pot: 1.5 BB) 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 has Q♠ 9♣
🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 raises to 38 BB, fold
🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 🙈 wins 37 BB
