Decision Making: Unsolvable Decisions in Poker (Crosspost)

[aujoz]

This post brings together a few different thoughts from some recent posts of mine in STTF. However, I think it's more broadly applicable across different forms of poker (and, ironically, probably has the least applicability to the short stack situations of 1 table tournaments)

1) I want to recommend to anyone who has the time (especially over the Christmas/New Year break) to read the book that I believe is probably the best book I've ever read - It's called "The Wisdom of Crowds" by James Surowiecki. The publisher's website is online at http://www.randomhouse.com/features/wisdomofcrowds/

The basic premise of the book is that a large group of independent and diverse people make better decisions than one person, even if that person is an expert.

For me, it has been an excellent unifying theory and explanation of why freedom and choice produces really good decisions - whether it be in selecting what to eat for lunch, what political party should govern, and how to make a decision on other things as well.

(I obviously receive no financial benefit from anyone buying or not buying this book)

2) Secondly, one of the poker players I respect very highly maintains a genuinely excellent blog on poker. At http://zbasic.com/2006/08/friday-aug...mbassador.html he wrote recently on a great idea that recognises that there is rarely one single "correct" and solved answer for poker questions (it is self-evident that some short stacked ICM problems are solved and there is an absolute answer in push/fold situations). This is something that has been discussed by stuff that i've read/listened to by Sklansky and Phil Gordon.

Let me quote at length from him:

[ QUOTE ]

...An outstanding suggestion that was unfortunately never really adopted on a wide enough scale was the proposal that strategy advice should never be doled out in the form of "this is a clear raise, call, fold", but rather should be given in the form of a ratio like x:y:z corresponding to the percentage of times -- in situations like that -- that one should call:raise:fold.

For examnple, imagine a situation where it's folded around to the blinds, the SB completes, the BB raises with AK and the SB calls. Now the flop comes K,7,2 rainbow and the SB leads out with a bet. The BB, sitting there with a very strong holding might make a post on 2+2 wondering how to extract the most from the SB who is unwittingly betting into a very strong hand.

The forums are full of knuckle-draggers arguing about the virtues of raising vs. just calling in order to raise on later streets, as though one were correct, and the other were patently retarded. I think a far more illuminating way of approaching the question is an "action ratio" presented in the form of something like 75:25:0, indicating that in situations like this the BB should raise 75% of the time, just call 25%, and never fold.

Other, more complex scenarios might produce action rations that look something like 60:30:10.? I think this more appropriately reflects the metagame features of a game of incomplete information; there are substantial advantages to changing up one's style of play, and the stark reality of the situation is that without knowing your opponent's hole cards, there is hardly EVER one clearly-correct course of action in a hand (unless, for example, you hold the nuts and have your opponent betting into you on the river.)


[/ QUOTE ]

3) So, combining the two, is it reasonable to tally the responses to a HH post, and then express them as a ratio?

eg: http://forumserver.twoplustwo.com/showfl...e=0#Post8496910

6 people say fold; 4 people say call. Therefore, the correct ratio from #2 above should be 60:40.

Waddaya reckon?

[AaronBrown]

I certainly agree that poker players should (almost) never say never. There are a few exceptions, for example, never fold the nuts. But in most cases, an action that makes so sense, should be done once in a while just because it makes no sense. Otherwise you become predictable.

But I don't believe in the ratio, because that's not the point. It's not that you want to slowplay a top pair 20% of the time, make a moderate raise 20% and make a large raise 60%; it's that there are times for all three tactics. Sometimes I drink water, sometimes I drink wine and sometimes I drink beer; but I don't decide in advance for the next year I'll do 30:50:20, I pick the right drink for the right time.

Even if a ratio is the right answer, a vote is not the way to get it. For one thing, good poker requires surprising people, you can't do that by acting average. A better use of the 60:40 ratio would be to predict what the other player might do, and pick the optimal response. The trouble is you have a lot more information about the other player than random voters on 2+2. Better to use that information.

Finally, aggregation is an important part of the wisdom of crowds. You need a question in which the average is the answer you want, and it's the answer people are trying to give. In this situation, people are giving 100% answers, and you're aggregating them to get a ratio. That's not what crowds are good at.

I think poker is a game that rewards staying a step ahead of the crowd, not being in the middle of it.

[alphatmw]

great response, aaronbrown

[_D&L_]

I'll second the argument that the right way to create a ratio is not combining indvidual wrong answers, such as always raise, always fold. Two wrongs, don't usually make a right. If a ratio was the right answer, why should combining two clearly wrong answers yield the correct ratio?

But i do agree with the quotation in the OP's posts, and that is players need to start thinking and justifying strategies beyond always do x. I really can't force myself to read most the strategy posts, like what would u do in these 5 situations - raise, check, etc. It just doesn't fit into my strategy of poker to have an absolute (or near absolute) answer to those questions, unless the answer is already a no-brainer.

[jetsetboy]

[ QUOTE ]
Sometimes I drink water, sometimes I drink wine and sometimes I drink beer; but I don't decide in advance for the next year I'll do 30:50:20, I pick the right drink for the right time.

[/ QUOTE ]

I think this decision is really read dependant: Which wine? Which beer? Are we still able to read after a month, following this quite aggressive line?

[Alex/Mugaaz]

Wisdom of the crowd is obviously stupid when the crowd is the one who is losing to the minority experts. I'm not trying to be cute or make a one liner comment here, but how much further explanation does this need? This could apply to other fields, but I can't believe the crowd's voice matters on issues that aren't opinion based. I sure as hell don't want the crowd's opinion when I need advice on a difficult medical problem, or when I decide whether I'm a favorite in the game I'm sitting in, or regarding a difficult issue with my taxes. The only opinion I want is the experts. Maybe I want to talk to more than one expert to reduce variance, but I'm not going to to alter a doctor's presciption or regimen because my cousin heard something different on dateline.

[Shandrax]

I have a problem with the basic assumption for similar reasons. Sports Betting would be tough to beat if the crowd would always be right. In fact, some people are making a nice living from betting against public opinion.

[Shandrax]

[ QUOTE ]
Sometimes I drink water, sometimes I drink wine and sometimes I drink beer; but I don't decide in advance for the next year I'll do 30:50:20, I pick the right drink for the right time.

[/ QUOTE ]

If you act upon feelings you will become predictable to someone who knows you well.

In my opinion randomizing decisions is important and if I can, I try to stick to numbers. While I am still searching for an application of this concept for Hold'em, it works well in 7C Stud.

[Jerrod Ankenman]

Instead of mixing the actions, you could just choose the one with higher expectation. That works better usually.

(Maximally) Exploitive play is pure. Any mixing that you do is an attempt to not be exploited, and then you should only mix actions that have equal expectations, or improve your strategy overall so that you don't gain from random mixing.

[_D&L_]

I think poker analysis becomes trivial when it all becomes "read" dependent. Sure, if you know your opponent plays rock too much, you play paper. But what if you don't know? What's your default that prevents you from being exploited, that maximizes return across an array of unknown strategies, or than opens up an opportunity to exploit your opponent?