Friday, July 27, 2007

Polaris

I've been reading a good bit about this poker bot that two pros, including Phil Laak, managed to beat. Here's an article from The New York Times and a thread on 2+2. Also check out the Slashdot link, which is hilarious because of some of the ignorant comments.

If anything, this bot makes me worry less about computers taking over poker. Why? Because this bot didn't win, even though it was playing a form of poker that should be easiest for programmers to master, heads-up limit hold 'em.

In heads-up limit games, the decision tree seems to be limited, and the absence of multiway pots removes some of the complexity brought on by several new, unknown variables. I feel better about the chances that I'm up against bots online if this one, supposedly one of the best in the world, couldn't even win in a relatively simple heads-up match.

Now, I'm not saying this bot sucks at poker. This one went against solid players who were able to achieve a decent winrate over a small sample size. I'm guessing this bot could probably do well in lower-stakes heads-up limit games.

But that's a long way from the dire imagined future of an online poker world filled with tables populated by machines rather than humans.

Look, chess and checkers have pretty much been mastered by computers, in large part because those are games of absolute information, where there are no unknowns. All the pieces are on the table, and the computer can figure out with a high degree of accuracy the best move based on many future outcomes. Unlike poker, there are no hidden cards, nor is there such a thing as bluffing. I can understand how these concepts might be more difficult for computers to understand.

Bots are a problem, and they'll only get better as time goes on.

But we don't seem close to a point in time where bots can outsmart thinking players in 6-max no limit games. I just don't see it right now.

5 comments:

Pseudo_Doctor said...

seriously bots are beatable because you know why? they are programed by people and most people are donkeys...I'll take my odds that the programmer's suck

Wes said...

Obviously this is not the best bot out right now. There is far more reasons why the best bot would to be made to just have a HU bot grind 30/60 LHE all day and never be outed than there is for this college to make a bot to beat a some known pro in an exhibition.

And pseudo_doctor, you are dumb.

SubZero said...

I agree Wes. The best designs go to the biggest buyer, look at the stuff designed for the US military.
I think in future the best bots will incorporate a neural net that will 'learn' in real-time how a player operates, even a player who can change their style mid-game.

syferium said...

I saw your website and all the blgs are very good, so you have the opportunity to sell space ads on you blogs... And i can't find another way to contact you,sorry...


i receive 72$ every month doing nothing, there is just 4 links at my website)))And you're website is much more better then mine.
Payments via PAYPAL and CHECK!!

http://www.text-link-ads.com/?ref=29747

No matter what language your website or blog has. No matter how many visitors a day visit your website, it's only depends on the PR!
PR MUST BE AT LEAST 2!!!!

IF SO YOU CAN EARN MONEY!!! LOT OF MONEY!!

try them...


After you'll sign up here write to me... and i'll tell you how to optimize it for well...

You can find several good tips here http://googlegeotargeting.blogspot.com

Hammer Player a.k.a Hoyazo said...

Gnome I agree 100%. So far I have neither seen nor heard of a bot that can consistently beat good players in nlh. LHE is only for donks anyways so I don't care if a bot can be made that beats those games. In general though when I fire up the 6max nlh tables, I hope there's a bot at my table. Eeeeasy money.