Time to kill

Waiting for the airport shuttle ... so time for a little venting.

I ran across a post by Bill Rini, which referenced a New Scientist article, allegedly concerning a breakthrough in bot research. It seemed like exciting stuff, more so when you read the first paragraph of the article:

IMAGINE being up against a poker player who can calculate the exact odds of a hand being a winner, play it with a straight face, and if necessary bluff with the best of them. Such a player exists, but you won't find him wearing a Stetson or hiding behind a pair of dark glasses. This player lurks within a computer, created by a pair of academics who have succeeded in making a software agent that can bluff just like a human player can.

The body of the article contains the usual unsubstantiated claims from "experts" such as: "Given the current state of poker bots, if you are losing to them you should be ashamed". Aside from the fact that there seems to be lots of evidence that this is not true -- should one really expect that those people who might be in possession of winning bots would advertise the fact? That is, aside from academic groups like those at the University of Alberta. These groups are handicapped by not being able to test and develop their bots in real play, as it would violate the terms and conditions at most sites, which I have to add is fair enough (though I think they should be allowed to play freerolls, with some provision for not actually collecting prize money).

Anyhow, my olfactory sense was further offended by the smell of bullshit when I read a quote from one of the authors of the original academic paper on which the New Scientist article is based:

That's because the bots now playing poker can't bluff convincingly. "Computers are programmed to perform the best strategy, but bluffing is based on unexpected, illogical actions," says Evan Hurwitz, a computer scientist at the University of the Witwatersrand in South Africa.

Now that's just so wrong. For most poker style games, bluffing is provably part of a "best strategy" (however you wish to interpret that phrase). The quote implies that incorporating "unexpected, illogical actions" is impossible in a program, when really it's simply a matter of making a few judiciously placed random number choices.

I was pleased that the original article was also referenced, so I went to have a look at it. Seriously underwhelming. The first point is that it has nothing to do with poker, contrary to the claims in the first paragraph of the New Scientist article. Then, the paper itself is just a train wreck. Here is a selection of a few of the things I found to disagree with or be annoyed by. Some are mere annoyances, but several are substantial:

• The abstract and introduction make out that the bluffing result is the main part of the paper, when in fact its mentioned in a short section at the end.


• The authors couldn't be bothered to learn how to incorporate graphics without nasty gray backgrounds.


• Even if they had, the only graph that can reasonably be interpreted contains no information. The ones which might contain information have differently coloured or shaded lines that can't be told apart (ye old paste from Excel bug), and don't support the claims made about them in the text. They have no legends, axis labels, nor explanatory captions. In a high school science project they'd be given a 0 for that part of the assessment.


• The game considered has absurd extra rules (it may well be a commonly played game in South Africa, but for an academic paper it could really be stripped to its essential form).


• The modeling ignores what would be a key feature in actual play (basically, in poker terms, that there would occasionally be "kill" pots).


• Finally, the example of bluffing behaviour would come down in poker terms to folding KK in the BB because of a late position raise. That is, it's an indication of the stupidity of the folding agent, not the cleverness of the bluffing one.
  

And one final point -- the description of the actual bluffing scenario in the New Scientist article is substantially different (and a better story) than that in the actual paper. Perhaps that's the way it actually happened, but an unpleasant odour lingers.