if i thought id get a nobel prize for it i wouldn't be as lazy in trying to prove it. either nash is wrong, or (much more likely) ur interpretation of what he said/proved is wrong.

ya... they have action (when i get back outside the us)

i just keep it simple and play my instincts, this whole thread is an eye sore.


but i hate math in general so to each his own

To me it seems like the only way to play GTO HUNL is if one opponent disclosed his full strategy to the other player and agreed to never adjust...what am I missing about GTO?

Not if I decide I am going to PAT every single hand, then I am exploiting a supposed GTO strategy, because it cannot adjust to that. Basically I dont think there is a calling frequency for this spot that is both going to have positive expectation vs
1.someone who is patting every hand,
2. someone who is patting some made hands and some bluffs

3.someone who is only patting made hands

I don't think a calling frequency exists that will have positive expectation against all 3 of those patting strategies simultaneously.

It's just funny you posted the same exact thing last year. People explained why you were wrong very clearly, and you still maintain the same exact position. Makes me think you might be trolling.

Any University of Alberta guys around to get this HU challenge happening?

Yeah, we can definitely do this but we'd need to work out some details.

First of all, I need to get permission from someone to use their bot. This shouldn't be a problem but could take a little while. The bot would definitely not change from the start. The code would be fixed in advance and it wouldn't be learning your tendencies. It would literally be playing according to a (very big) chart.

I don't think Stars or FTP allow bots, but maybe they'd make an exception.

More likely it would have to be on some platform other than a normal, licensed poker site. I'd obviously be willing to take every precaution to make sure the RNG and everything else about the setup is on the up and up. We could have whatever the setup is audited by some third party.

Playing 100k hands might not be practical. It should be either a smaller sample and/or there should be a buyout clause that is very likely to get used.

Any of the U of A guys reading this thread? PM me if so!

lol seriously can people stop posting about gto when they quite obviously have no knowledge of what gto even means? Seriously it takes like 5-10 minutes max to just google it and at least understand the basics. A game theory optimal strategy is optimal whether 1 hand is being played or a billion hands are being played.

You are not willing to book him on the IGT machine?

You'd be wrong. Can we simplify the game a bit more?

1. both players ante 1 unit and are dealt a random number 1-3
2. Player 1 gets a chance to either stand pat or re-draw, then gets the chance to bet 2 units (pot-sized bet) or check
3. If faced with a bet, player 2 may fold or call, if player 1 checks hand goes to showdown
4. the player whose hand is the bigger number wins the pot, they split the pot if they have the same number

I currently have no idea what the gto solution would be but I'd be willing to bet (or if you don't want to bet might just do it for fun anyway) that I can come up with a strategy that is gto, that is no matter what you do assuming we each get to be player 1 and player 2 an equal number of times you will not come out ahead. Interested in a bet?

I don't know enough about the underlying AI but there are also a bunch of practical problems. It's not a format that lends itself to getting max hands/hr. It requires him to hang out in a casino all day with another independent observer to verify results. It's not very spectator friendly. Tom has expressed interest in using a HUD.

All that being said, if we can't find a better solution, I'd probably do that.

Nah, because I think the simplification fundamentally changes the game. I think your simple game can be solved, and I still don't think 5 card draw high can.
In your simple game, the decision is much easier, because if you have the nuts(3) you obviously call, and when you have the 1 you obviously fold. Your only decision would be whether or not to call with the 2 or not and how often.
It isn't even the same concept because the game tree is far too small. In 5 Card Draw, the bot isn't going to have the stone cold nuts 1/3 of the time. The point is he is not going to have a good enough hand to call a PAT very often, unless he is adjusting to how often the opponent who is doing it, which the bot cannot do. There is not going to be a calling frequency with an edge over the three separate pat strategies i listed.

jusgivithere,

You have no idea what you are talking about and couldn't be more wrong!

If you were right, you could win a nobel prize!

For sure Jungleman falls into this camp.

Durrr rocking the 'I'm savant who could prove all you NE math fools wrong if I wasn't busy making skrillah' angle again...

With only two cards in hand and 100bb in stack, he has seen things we can only dream of.

No, chess is not totally soved.

However, top computers play far better than top humans. Even so, not ALL games between top computers or computers end in a draw.

"1.someone who is patting every hand,
2. someone who is patting some made hands and some bluffs
3.someone who is only patting made hands

I don't think a calling frequency exists that will have positive expectation against all 3 of those patting strategies simultaneously. "

Ok then come up with a calling range/frequency that has an edge over those 3 strategies at the same time,that cannot adjust. You simply can't. Because any frequency that has an edge against 2 and 3 can be exploited by 1, and any frequency that has an edge against 1 can be exploited by 3.

If there was a game where the numbers were from 1-100 and there was 1 draw, and I knew what the bots calling range was against a stand pat, i would have an edge.

So the only reason one is solvable and the other isn't is because it's simpler? What if we use the numbers 1-5 instead? I'm trying to come up with something that still has the element of standing pat you seem to think defies having a NE but being simple enough for me to actually come up with what that NE (or if there is none prove there is none, but that won't happen because you're wrong) is. It would probably take a computer to come up with NE in that other example whereas the 1-3 game I could solve pretty easily by hand.

Good mention.

One simple example of Nash's idea would be the game of Rock-Paper-Scissors. I could tell you my strategy and..

The problem is that gto isn't profitable against every single range, it's profitable against the overall sum of actions. So if you are patting every hand you are giving up the value that is drawing. So it might not call enough in an exploitative environment and you're gaining a small amount of ev in that they are folding too much in this particular instance, but then when the gto bot is in that situation it gets to draw and make a better hand with more bad starting hands than you do, and that's where it will make its money back.

The bots GTo strategy would be more complex than just [x and above] and involve randomization

If you are right, then prove it, and collect your Nobel Prize!

Edit: or what was already posted

How can anyone with a shred of intelligence read the wiki on the Nash Equilibrium and think that Nl hu/this number draw game don't have a nash solution? I mean the one argument is that 'poker isn't a game as defined in the proof' which I'm not expert enough to answer, but people smarter than me say it is, so ya!

Obv. more GTO river calls are made when the pot lays 10 to 1 than when the pot lays 2 to 1.

Zadeh dissects this idea for the games of Draw Poker and Low Ball in his book Winning Poker Systems, 1974.

Also, he discusses deviating when you have information.

The term Game Theory never appears in the book. Zadeh speaks in terms of best Uninformed Strategies and best Informed Strategies- however, he attempts to solve via GT.

Dont understand why people keep trying to argue with ppl who obviously have no clue (except durrrr I guess since he wants to bet). Ike I would pm egj (author of slumbot)