No. But you can play HULHE here

Hello Sam ... are you free to answer this question now: is it possible to let Claudico run on standard PC's / laptops ?

Since this hand comes up all the time here: does anyone have the link to it ?

Cepheus is probably known to most of us itt, but interest in LHE is probably several magnitudes behind interest in NLHE for the huge majority here ... given Prof. Sandholms remarks in the Sept. CardPlayer interview, the question whether CMU is going to do something similar with Claudico than UoA did with Cepheus stands. Would be fantastic indeed, and hopefully someone from CMU notices the question and likes to answer ... thanks !!

lol

i guess the results are relatively good news.

they are at least 1 order of magnitude away from competing against top humans.

With any reasonable time limit imposed, that number is much larger.

Based on the fact that Microsoft sponsored the event, I fully expect Claudico's amazingly powerful and perceptive artificial intelligence will be built into the next incarnation of Windows.

No, but a fellow AI is available at Neo Poker Bot. Neo placed third in the Annual Computer Poker Competition in 2012 in the HUNLHE "Bankroll Instant Runoff" section but has not done as well as CMU's Tartanian7 in subsequent competitions.

So rigged. I got it down to $200 in about 10 hands then it sucked out on me three times in a row and I rage-quit.

Doug would have to chime in with the exact link, as I only heard about it when he was speaking to the Professor. I believe that the crux of the hand was that Doug (99) 4-bet shoved after Claudico (A4o) made it 9K to go.

Bot has some glaring leaks. In an unrelated note, after about 75 hands, I stacked him after I 3b w/95bb stacks w/AxJc and c/minr a AKT all club flop and he got ai w/Ax5c.

confirmed bot is a fish

Bot seems to jam all pocket pairs in the few rounds I played. He stacked me when I got it in with QQ vs his 88 and bot rivered an 8.

I'm going to be on Dave Tuchman's "Under The Gun" Twitch podcast on CrushLivePoker tonight at 8pm pst discussing the Brains vs A.I. challenge. If you have any questions, please get in there in chat and I'll try to answer them.

I don't understand this at all. Bot's will shortly thereafter infest online poker at an exponential rate. WR will drop and so will growth of online poker as a business.

From a pure curiosity standpoint I can understand interest in the subject but coming from the Head of Poker Communications at the worlds largest poker site, it's a bit depressing.

meh

You didn't do this for the money and you got more out of this than just 210/h (for yourself and the poker community)

The only thing I wonder is how long poker online will last now that bots are able to beat 99.99% of humans (and some more next year)

Yes, Bot has some glaring leaks. It is very easy to exploit. But it is fun to play!

Ty 'statistical' Stewart

Neobot's interface already seems so much nicer than the one our four humans had to put up with. Microsoft couldn't even copy that ...?!

From UCLA's stats website

Is Neo Poker Bot a good example of what to expect from run-ins with todays bots at HU deep SNG and cash tables? Or are they much better now?

Seems like quite an easily exploitable tight-passive bot, so don't mind playing these all day.

I thnk worrying over a one sided vs two sided t test is neither here nor there. A two sided test tests whether some value is different from zero in either direction and is almost universally used except in rare cases, particularly because it's more conservative. It's fine.

However, we know that if humans lost at 9BB/100 there's no way CMU would be going on about a statistical tie.

Also, it's not clear the assumptions of a t test are even met here. Each hand is not independent and identically distributed, particularly because humans play certainly depended on what came before. See Doug Polk getting tired.

Stats tests are useful but never perfect and not the be all and end all of anything. It's not clear they're appropriate here.

I think a Bayesian approach is better. For now, the evidence tells us humans are better. We can update our hypothesis when another challenge is run.m

Are we sure that the bot improved it's play, or learned? After all, the programmers got to watch the action, then tweak the computer after each session. Just doesn't seem like a "true test" after learning that.

Here's the bot, play the bot, live with the results. Seems like there was some leak fixing as it went on.

I'm going to start using statistical tie all the time now, anytime I lose.

jeskola.net / poker (200bb) is another bot that has only pot size bets but is almost flawless otherwise. It is (or should be) virtually unbeatable. Claudico was probably (educated guess) >1000mb exploitable so there is a big difference.