Pokernomics

[Rudbaeck]

[ QUOTE ]
[ QUOTE ]
Levitt's paper on the NFL gambling markets was not good. He used poor data to come to incorrect conclusions. So I am not convinced he can do a good job analyzing poker data.

[/ QUOTE ]

I'm even less than "not convinced." Part of my evidence for this conclusion is that he asks for only 10k or more hands. Frankly, I don't think he has a clue about what he's getting into.

[/ QUOTE ]

Which means that someone with a clue can still write the real version. 2+2 and PokerRoom could probably accomplish some fairly interesting datamining... *nudge nudge wink wink*

[Ed Miller]

[ QUOTE ]
Which means that someone with a clue can still write the real version. 2+2 and PokerRoom could probably accomplish some fairly interesting datamining... *nudge nudge wink wink*

[/ QUOTE ]

I'd love to see something like this done right. And perhaps Mr. Pokernomics will do a fine job, and I'll eat my words.

To be honest, though, I think a "real" implementation of this sort of analysis requires a melding of both "human" understanding and purely statistical. That is, the statistical stuff is most useful to determine the right play in situations that people can't easily derive theoretically. For instance, what to do with A9s UTG in a typical low-limit online game. [img]/images/graemlins/smile.gif[/img]

But the tone of Mr. Pokernomics's little spiel portends another flaw I see with his approach. He seems to poopoo the work that those quaint little theoreticians have done. It's almost as if he's saying, "For years we've had to make stuff up about poker theory because we didn't have real data. But now that the data is available, let's let the big boys step in and do the job right."

Unless he hires some sharp theoretical minds to tell him where his models are going haywire, I don't think he's got a chance. He's gonna come up with some very silly results in many places. Of course his book will probably sell well regardless, so all is not lost. [img]/images/graemlins/tongue.gif[/img]

So yes, I think many people at 2+2 could probably do a better job of this sort of thing than this guy. Anyone want to step up to the plate? [img]/images/graemlins/smile.gif[/img]

[Ed Miller]

[ QUOTE ]
[ QUOTE ]
Have you tried it? I'm wondering if they give quality, accurate and useful feedback.

[/ QUOTE ]

Sent in a database of some 35k hands today, so haven't had a chance to get the analysis yet. I'll get back on it.

[/ QUOTE ]

Maybe I'll play a few hands at $1-$2 and intentionally make some modest, systematic errors. Then I'll submit them and see what they have to say.

[Rudbaeck]

[ QUOTE ]
For instance, what to do with A9s UTG in a typical low-limit online game. [img]/images/graemlins/smile.gif[/img]

[/ QUOTE ]

Going off topic here, but I think it honestly doesn't matter if you raise or call. The difference is going to be so fine that it will be darn near impossible to see a difference even in a humongous sample.

[Jordan Olsommer]

[ QUOTE ]

Maybe I'll play a few hands at $1-$2 and intentionally make some modest, systematic errors. Then I'll submit them and see what they have to say.

[/ QUOTE ]

Just out of curiosity, but how would you go about challenging any claims this guy will make at the end of his project which you deem incorrect? I don't mean to offend; I'm genuinely curious, and can only see this project as a Good Thing overall - if he actually comes up with something useful and it helps people to understand the game better, great! If he just comes up with a bunch of absurd conclusions that help people to understand the game better by understanding why they are absurd, that's great too. But it just seems to me that there would be an awfully great temptation for a theoretician to hear Levitt say something like "based on this and that data, I conclude that you should come in for a raise with A9o UTG" and simply reply "hogwash!" without ever properly addressing the argument, you know? (in fact, upon consulting gocee, that is a conceivable argument that a solely statistics-minded person might make, since A9o will win 1.9% more than its fair share of pots against 9 random hands)

[Kama45]

Judging from his last book, this work will probably try to appeal to the masses including/especially non-poker players. Even if all his analysis and data-collecting methods are 100% sound, there is unlikely going to be anything useful for a typical 2+2 poster. Hopefully, I am wrong.

[djack]

[ QUOTE ]
Judging from his last book, this work will probably try to appeal to the masses including/especially non-poker players. Even if all his analysis and data-collecting methods are 100% sound, there is unlikely going to be anything useful for a typical 2+2 poster. Hopefully, I am wrong.

[/ QUOTE ]

Definitely agreed. I figure it's a way for me to pick up a free book and tshirt.

[Rudbaeck]

[ QUOTE ]
Judging from his last book, this work will probably try to appeal to the masses including/especially non-poker players.

[/ QUOTE ]

If you read the post about in in Poker Theory it seems clear it will first be published as a research article. Hence pretty unpenetrable to laymen. It might develop into a book though, but that doesn't seem to be the primary goal.

[Ed Miller]

My thoughts have crystalized somewhat:

1. I'd love to see someone take a shot at this.

2. I think drawing reliable conclusions about the complex stuff that we'd like to see addressed will be really hard, much harder than it sounds. I liken the complexity of the problem to that of speech recognition or natural language. When I worked at Microsoft, I initially worked on the search engine, MSN Search. About six months in, I was moved to a new project that was supposed to develop a natural language search solution for the next version of Windows. Ultimately I left to play poker, and the project failed. (In that order, of course) The crux of the problem is that natural language is a lot tougher problem that it sounds like it should be. It seems like if we just gather enough training data from enough people about the way they use language and what they mean by it, it just becomes a classification problem to make sense of everything. But in practice it doesn't quite work like that. And people have been trying for the better part of 30 years to do decent natural language and speech recognition with still limited successes.

So don't have blind faith in large datasets. There's a lot more to solving these problems than merely gathering lots of data. Brilliant minds have been pouring over natural language data for decades and still haven't really come close to "solving" the problem.

3. I'm really uncomfortable with the promises Levitt is making regarding giving hand history donors insights into their play. Frankly, and perhaps I'm jumping to conclusions here, but I feel I'm probably not, he's got no idea at this point whether he's going to draw meaningful conclusions or mere garbage.

People play poker for money, and bad advice is expensive. It behooves anyone who gives people advice on their games generally to know what they are talking about. Now that's obviously not to say that many advice-peddlers presently active aren't failing in exactly that way. But it is to say that Dr. Levitt, unlike some other poker types, clearly aspires to be an upstanding citizen, as it were. His interests are probably academic in nature. So he should seek to "do no harm."

I think he should dump the "I'll give you advice on your game" promise and just go with the T-shirts. If he wants to keep that aspect of the incentives, I think he should hire someone who really knows how to play poker to review the recommendations for reasonability before they go out.

Sure, the whole point of the project is to uncover things that people like me have missed in our books. But when you cross over from mere observer to advice-giver, you take on a LOT more responsibility.

4. My initial reactions sound disturbingly (to me) like how the baseball old guard reacted to Billy Beane in Moneyball. (Though I think the analogy is very incomplete.)

5. I think this is a cool project, and I hope it's successful.

[steamboatin]

Maybe it will be like just after Phil Helmuth's "Play Poker like the Pros". A bunch of fools faithfully following terrible advice and Steamboatin gets to run good until they run out of money.

Immediately after this book came out the play at Caesar's IN changed dramactically for the worse. Much more aggression with terrible hands and capping preflop with any pocket pair, etc.

I never read the book and most likely never will but for a few months after it's debut, low limit poker was a whole lot of fun so I drew the conclusion from listening to other player's discussions and posts from 2+2 that it was most likely to blame for the sudden change in Low Limit poker.

That is why, Phil Helmuth is my favorite author for other people to read and I anxiously await his next masterpiece.