Just had mental images of degens bumhunting the machines waiting for people to leave so they can grim the button then leave.

What are you talking about?!? The dude is obviously an expert at logic gates!

Thats what we love about NVG its just full of experts on subjects like this who share there knowledge opinions with us.

Probably in trial pending approval.

lol, awesome

Sigh! Yet you do not even bother to attempt to correct it.

http://www.pokerlistings.com/strateg...-the-end-um-no

http://www.pokernews.com/news/2008/0...stoxpoker-.htm

2 minutes of browsing got me this info, I'm sure you can find better information if you tried.

I obviosuly no jack **** when it comes to making these machines, but you would be ******ed to think I'm not right on such issues. It boils down to 2 points

1. Given complete information (like chess) it comes down to computational power which computers are better than humans at, LDFO!!. (It should be noted that the best human players could beat the best computers of the past by playing unconventional "wooden" strategies which the computer could not counter effectively. However the latest computer programs can deal with them so now they simply beat all human players)

2. Even in games with incomplete information (like any form of poker) you can still devise strategies which are unexploitable, the more branches (inputs) there are to a game the harder it is to develop such strategies. So bots at NLHE suck arse because it is mostly unsolvable (bar short stacking) and requires game flow and dynamic which computers can't deal with.

Edit: anyone refuting my statement that computer players won't be good at LHE and then talking about the importance of the button in a LHE situation is just LOL!

It can't be bargained with, it can't be reasoned with.
It doesn't feel pity or remorse or fear
and it absolutely will not stop, ever, until you are busto.

fwiw, the post about the button is the most relevant post ITT, and your original post was the least.

HULHE is the form of poker that a bot would most easily be able to play well, but as far as i know, no bot has been built that could beat the best players.

OP is obviously in on the development / profits of these machines and trying to attract suckers.

so f obvious

How long before FTP signs this "robot" up to be a red pro

That probably has to do with the amount of total bull**** being spewed here. Let me be nice and give a serious response.

-LHE is not solved.
-Polaris did beat good human players but did adapt so it is not really relevant here

You said:

This machine... do you know anything about how this machine plays? Do you know anything about the state of the art of non-adapting LHE AI's? I dont but neither do you.

Which is rather lol.

http://en.wikipedia.org/wiki/Computer_Go#Performance

The human mind can currently do a lot of stuff that computers cant yet do.

What things computers can and cant do in theory and in practise is something a lot of very clever people think about. It is nowhere near as simple as "COMPUTORZ ARE FAST! THEY WIN!".

AFAIK that is bull**** and NLHE has a game theory optimal strategy just like LHE had. To find this strategy is what is considered solving the game and the strategy has nothing to do with flow and dynamic. It is just that LHE has much smaller state space so solving it is computationally more feasible (I think I heard polaris uses solutions to sub games of LHE).

Wow, when did Vegas become a clean honest city? lolz

Nobody seems to follow the annual computer poker competition anymore, which I did. The winner this year which is rockhopper is already so much better than Polaris that played Hoss. Actually we can safely assume that the best bot atm CAN beat all human.
For more information, visit: http://www.computerpokercompetition.org/

You actually are in fact completely uninformed



The bot beat some players over 500 hands. That doesn't mean that it played better than those players. I think they used some methods to make variance less of an issue, but it's 500 hands FFS.

No.. actually it's just that you know jack **** about this issue.

In theory a computer could beat a human at any game. In practice, that's no where close to the case.

Your discussion of unconventional wooden strategies just doesn't mean anything, so I'll just ignore it.

Actually, NLHE is completely solvable. It hasn't been solved yet, and it may never be solved, but there is absolutely a solution.



When you publicly call someone a ****** for their opinion, you should know wtf you're talking about.

Actually, HU NLHE is unsolvable if you can bet any sizes, because that makes the game size infiinite.
Game theory only states that a game with finite size, like HULHE, has a MiniMax solution to it.
What ppl acutally do to NLHE, is to bucket the bet sizings, such as you can only bet say, half pot, full pot, 2 times pot, etc.
Read research papers published by UAlberta in case your inrestered.

Any size/infinite? You don't use infinite number of digits in bet sizes, do you...

but theoretically, you can. Just imagine with all those sizings, how big the game tree would look like.

Even if this "robot" can be beat by good players, it may not be long before future versions are able to beat almost all humans.

Ok you could simplify the game to being a maximum stack depth of a billion BBs, and you must bet in increments of 0.0000001 BBs. Happy?

does it go on tilt at all?

Is it legal to bet in currency units that do not exist?

Also, if you are considering the amount of states getting infinite because of betsizes being unbounded... they are bounded by stacksize. We can just solve for any particular effective stacksize. There might be an infinite amount of possible effective stacksizes, any particular stacksize is finite...

(yeah I realize you are not being 100% serious here, just wondering a bit)

Games with infinite game spaces still have solutions. In NLHE the fact that you can sorta maybe theoretically bet any real number isn't a big problem at all because it's quite easy to prove that there are only finitely many optimal bet sizes in any specific spot. If you had a solution for NLHE with 100 BB stacks with bet sizes in BB increments, you'd be one really boring step away from a solution to 100 BB NLHE with real number bet sizes.

Anyway, start another thread if you're interested. This nerdy hijack sucks. I was just insulting the ****** who was calling other people ******s so that he'd shut up.

So actually gobbo ran hot and should quit now that he is up.

Also, I wouldn't be shocked if someone was able to commercialize a bot that can beat virtually every human. What's more, if it loses like 0.5bb/100 vs. the best of the best, breaks even vs. very good players and beats the nl donks who want to give it a shot, I think the machine will have game selected perfectly, because the best players probably won't play it for long anyway.

In the end, the machine would have a lot of hands vs. people it totally owns and fewer hands vs. ppl who either break even or beat it. But if it loses a tiny amount to the best and wtfpwns the others, it will still make a nice profit.

Moreover with a little tweak, when it posts strategy on 2p2, in a HU4ROLLZ situation it will probably bust everyone since Bellagio is overrolled.

Anyone estimate its probable winrate?

They need to bring out a Cribbage version.

I've done the math, and it's theoretical winrate is exactly sqrt(2)/2 bbs/hour.

Our current estimation is that there is at least 33.3 chance, repeating of course, that this is correct. My team is still working on it, tho. I will let you guys know.