NLHE Tournament Hand Ranking for All-In

[PairTheBoard]

Yes of course. What was I thinking. We expect Aisthesis to calculate a larger stack which makes the K-S/A ratio slightly smaller. So that's great. That's just what we have and I think it shows Continuity between Aisthesis results for several players and the K-S results for the SB. I believe that's just what we need to have confidence in Aisthesis's results.

It looks like both you and Aisthesis treat the Small Blind Bet the same so there should be no descripancy between the two results in the SB case due to that.

I also think you explain the larger descrepancy for the 22 case.

I'm afraid I don't have the patience to deal with computer code these days.

Did David ever writeup your results in the CardPlayer Karlson?

The correlation between the K-S and Aisthesis results taking into account the non-fundamentally different methods gives me confidence in the accuracy of both.

PairTheBoard

[PairTheBoard]

From Karlson's post below it looks to me like you both treat the Small Blind bet the same. So that's great. I think that's the best way for continuity going from the AV2 case to the full table cases.

And Karlson is of course right about the DF being smaller than the expected 3 because the AV2 stack tends higher due to the restricted calling requirments. I wasn't thinking clearly on that. So that's great. Your's and his results correlate as we would expect them to which I think confirms the validity of your results. I think you're in business Aisthesis.

I intend to study your actual results more closely now as I believe we can have pretty good confidence in them after their favorable comparison to K-S in the SB case.

PairTheBoard

[Aisthesis]

I agree fully. I will say that Karlson's results are definitely "cleaner" (actually completely clean, as far as I would be able to tell) from a game-theoretical standpoint.

However, some of the things that make it a completely accurate analysis of the given scenario may in some cases tend to distort the actual practice of poker.

A good example is again the extreme case of 22. Against that hand, if it is shown, 73o will want to call against most stack-sizes from BB and might also have odds to call from the button (at least if 22 were the only hand to worry about) if 22 were to move in from the CO. But in practice, that hand is surely going to fold every time because it is so horrible against the RANGE of hands justifying the all-in. So, some of my assumptions which tend to detract from a completely precise game-theoretical approach to the problem may tend to give a more accurate reflection of expected practice.

[Aisthesis]

You are in a NLHE tournament in the middle phase. Blinds are at 200/400 with 25 in antes. There are 9 players at your table, and you are the shortest stack with 2,400 in tournament chips. Somewhat more than half of the players have been eliminated, but you still have a long way to go before the money starts. I'll further assume that you've just been moved to this table just to avoid the additional complications of reads, but since everyone has successfully survived this far, you can at least assume that the other players at the table are pretty good.

You are dealt 77 in early MP (UTG+3), and it is folded to you. What is your best play here?

[Jerrod Ankenman]

Jam, and it isn't close.

Jerrod

[Jerrod Ankenman]

I usually use the rule of thumb of about five to six times the pot. So if the antes are 1 BB, then as early as 15 times the BB. If the antes are (as is more usual), half a big blind, then near 10-12 blinds.

[Aisthesis]

Ok, you actually start a little earlier than I do (I've backed down to about 10 BB, sometimes even a little less until antes start). I posted the same question on UPF and had several people who started WAY later (4 BB) or not at all. I don't exactly see how they survive that way, but if they can do it, more power to 'em!

I'm trying here to get a range of stack-sizes that applies to most players for figuring out which hands to analyse on all this--like JJ or AKo are obviously great from any position short-stacked. But AJo at a 9- or 10-player table doesn't appear to be quite enough UTG, if your stack is as much as 5 times the total pot. And truly weak aces seem to become only viable in LP if it's folded to you.

Anyhow, thanks for the very helpful response!

[karlson]

"Basically, what would it take to run the truly precise answer for various numbers of players?"

Once you figure out the exact question you're trying to answer, it's not all that difficult. For the KS rankings, it was very well defined -- at what point are you better off flipping your hand face up and going all-in than folding, given that your opponents play perfectly. While the question is basically the same here, defining perfect paly for your opponents is a tad trickier. If you give up on this, I might pursue myself at this point, but if you come up with the exact conditions, I could definitely help you do the simulations.

Hey Jerrod, great to have you join the forums. Can I hereafter refer to you as Noted Game Theory Authority JA?

[Aisthesis]

I think I could give an exact definition of perfect play--but it's also not clear to me whether it provides a better or worse approximation of practical situations, and it also clearly makes the simulation completely impossible by hand.

But why not use essentially the same definition the you did on the SB case?

The definition would then be this (?): You turn your cards face up, and any opponent calls if he has +EV given the size of the pot coming to him. I would further assume that all opponents have you covered, but no further betting is allowed (a lot like Sklanksy's recent "unusual hypothetical situation"). Further, you are always UTG, but we simply run the question for tables with 10 all the way down to just 3 players.

But in addition to the unusual situation resulting on the 22 hand, there are also going to be complications from the simulation standpoint: If you move in UTG and flip over your cards, say UTG+1 has a good hand giving pot odds for a call. The calling criteria are now going to go way down for all subsequent players.

As I say, I really have no clue how one puts together the kind of code you run to get results on something like that, but it would definitely have to get data off of a lot of unusual situations, many of which have quite low probabilities.

That problem looks to me like it is well-defined, anyway. And if you are willing to write code to get clean results on that problem, the results would be EXTREMELY interesting imo. I'd be particularly interested in how those results match up with the ones I've gotten by hand.

I also have some somewhat vague ideas as to how one might define a game-theoretical situation without the "open cards" assumption. That approach would imo be better in principle, as it also would incorporate the "gap concept," but I still can't quite figure out how to make the calling criteria truly well-defined given the variables involved--specifically, stack-size, since the smaller your stack, the larger your range of hands is going to be for the initial all-in. That type of solution would imo be the "perfect" game-theoretical model for practice, but, even if one could define a precise range of hands for "hero" given his stack-size (hence giving the opponents at least a well-defined decision), the solution to that problem would create a lot of complications of its own, I would think.

[Aisthesis]

Well, blast! Was this one too easy?? [img]/images/graemlins/smile.gif[/img]

Anyhow, I agree, although I was hoping someone might want to limp here or raise less than all-in...

By the way, while I definitely prefer the all-in in this situation, I can't provide any objective reason why limping or a raise less than all-in is a bad play (hence, the "survey" aspect in the subject line).

Many of the players on UPF, anyway, seem a lot more conservative on "all-in or fold" mode, so I was also trying to get a feel for where I stand on this issue relative to the rest of this board.

Does anyone here DISLIKE the all-in? Would everyone here still move in with a stack as large as 4,000?