Pokernomics

[sfer]

It will almost certainly not be a book but rather a paper with very generalized findings.

[Robk]

[ QUOTE ]
his "theory" that he's unqualified is based entirely on speculation. The best anyone has managed to show as evidence is King's points about the NFL study

[/ QUOTE ]

people who spend hundreds of hours on 2+2 reading things that strangers have written about poker develop an innate sense for a writers' poker knowledge that is strikingly accurate. its based on the phrases a writer uses, the questions he asks, and the manner in which he asks them. i think that much of the skepticism you are receiving on here is due to the fact that (to me at least) the pokernomics page reads like it was written by someone who knows little to nothing about poker. (i can point out the specific passages that make me think this if you want.)

i dont think this means the study will be bad. for all i know levitt didnt even write the page and is a poker expert. or will acquire more knowledge before/during the study, or will do a great job in spite of a lack of poker expertise. im just trying to explain why some may not have total confidence. since the page was meant to appeal (or adverise, even) to hardcore poker players (those who play many hands and have saved datatbases of them) i do feel it was a mistake to put up such underdeveloped thoughts without consulting one of those members of the u of c family who know more about the game.

also you state that we have scant evidence, yet have not responded to the most damaging comment imo, namely Ed Millers point that advice generated from 10,000 hands will be misleading and perhaps dangerously so. also king yaos point about the flawed nfl study is much more damaging than i think you are giving it credit for, although perhaps thats only because you havent read the study.

in any event im optimistic about the study and will be sending in 50k hands in a few days. good luck to you.

[Ed Miller]

[ QUOTE ]
also you state that we have scant evidence, yet have not responded to the most damaging comment imo, namely Ed Millers point that advice generated from 10,000 hands will be misleading and perhaps dangerously so. also king yaos point about the flawed nfl study is much more damaging than i think you are giving it credit for, although perhaps thats only because you havent read the study.

[/ QUOTE ]

I went back and forth once with Dr. Levitt, and I'm still not convinced the feedback he will give will be at all valuable for the large majority of his donors. He said that the feedback will be limited in scope and specificity. But it still could easily be "bad" feedback.. and even if the feedback is technically correct, it could send players to the wrong parts of their game looking for things to change.

[djack]

First off, Scott, my first reaction is this: "methinks the lady doth protest too much." I realize that this is in part because you're an econometrician and you're defending your field. That's fine.

After all, we'll see who is right. I have an open mind about whether Levitt can do something useful. I won't judge until I get my results. But that doesn't mean I can objectively decide that I think the odds are against Levitt offering us much truly worthwhile (did you like the ambiguity in that phrasing? [img]/images/graemlins/wink.gif[/img] )


One more thing. Do you have a source for this paragraph?

[ QUOTE ]
Also, one good thing to come of this to keep in mind is that even if you do think Levitt is unqualified, a database on millions of hands and (hopefully) a large sample of individual players (thousands maybe? We can only hope) will come into existence.

[/ QUOTE ]

I missed this on the website (and don't have time now for a perusal). If Levitt is planning to make my database public and accessible, then I am definitely not submitting.

[sfer]

[ QUOTE ]
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.

[/ QUOTE ]

I've already suggested as much to the Grad Student likely doing most of the work here.

[disjunction]

[ QUOTE ]

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.

[/ QUOTE ]

Exactly. My first reaction is that he will produce a nice playbook for what to do against a certain VPIP/PFR player after a certain flop, but the result will be kind of a "static" strategy which misses the rock/paper/scissors aspect of poker.

[Edit: Not to imply that he won't say anything interesting, or do bad science. More that he'll only be able to tackle a small piece of a difficult problem. I'd compare it to some of the old Peter_Rus math posts.]

[MuckerFish]

I've been seeing this all over the forum. I'm a little skeptical. Seems to me this collection method is flawed, wouldn't this be a victim of collection bias? People who send in histories looking for advice may be better than the general population, or who knows, maybe a little worse? What if you sent in 10K of your best hands, or 10K of your worst? What if you intentionally misplay 10K hands and then send them in. Also what does your results over 10K or even 50K hands really mean? The web site said it will look for what a winning player does, but that sample of hands doesnt say anything about whether or not you are a winning player. Winners can go on 10K losing streaks and losers can get hot. Seems like you would need to collect info about the player also. What about variables like short handed vs full ring? What if you lose consistently in low stakes cash ring games for 9K hands or so, but cash in a MTT that played for the last 1000 hands of the sample - is this person a 'winning' player that my hands would be compared against?
I don't know, seems like too many variables to control for, and too much selection bias in the collection method. But, I'm not much of a stats guy, so maybe these things aren't a big deal. But before I'd put much stock in his conclusions about my play, or play of 'winners' in general, I'd need to see more info on the validity of the statistical methods and sampling methods.

[uuDevil]

From http://www.pokernomics.com/moreontheproject.htm :

[ QUOTE ]
We also understand that poker players have reason to guard their hand histories. Therefore, we guarantee that only people who are working directly on the project will have access to people's hand histories. NO ONE'S HAND HISTORY WILL EVER BE MADE AVAILABLE TO ANYONE NOT ON THE PROJECT.

[/ QUOTE ]

I'm taking them at their word, but even if somehow my hand histories became public despite this promise, I'm not too worried that anyone would bother to analyze my play and show up at my table to take (added) advantage of me for what would probably amount to some small fraction of a BB/100 at $1/2.

Edit: From Levitt's blog:

[ QUOTE ]
Rest assured, we will handle the data with great care. We will not give or sell your data to anyone else. Nearly forty players have already contributed more than two million hands to me.

[/ QUOTE ]

That last part made me smile, for some reason....

[vulturesrow]

Ed,

I will submit the same hand history to you that I did to the pokernomics site and then compare the analysis. I wont even charge you. [img]/images/graemlins/smile.gif[/img] Seriously though, I wasnt expecting much out of the analysis, I really just wanted the free stuff and I was pretty unconcerned about someone looking me up at the Party Poker low limit tables to exploit my hand histories. But it will be interesting to see what the results are and how it compares with what I see as my weak areas.

[djack]

What if all those that might submit our databases to Levitt gave them to someone like Ed?

I don't know if mr. majorkong would be interested, of course. It'd have to be someone that the majority of posters here trust.

Ed just seems to be like the perfect guy because he doesn't play much online (at least I don't think so), he's respected by people here, and he's a better theorist than Levitt. heh.

Would anyone actually be interested in doing this?