Thoughts on ICM

[trojanrabbit]

In thinking about ICM, there is something that never occurred to me before. Don’t know if this has been discussed before, but I couldn’t find any reference on it.

Say you’re on the bubble with 150/300 blinds and these stacks:

SB 1800
BB 1600
CO 2600
Button 2000

Let’s say you’re in the CO and you have a borderline hand that you think might be slightly +EV to push. When you calculate what is your baseline prize equity after you fold, you use the current chip stacks. However, there’s a pretty good chance that some action is going to happen after you fold. Maybe the button will steal the blinds shrinking both of the short stacks or maybe there will be an all-in confrontation. Any sort of action improves your EV regardless of who wins. In fact, every time the other stacks are relatively equal you gain anytime there is action, regardless of the outcome. That makes the EV of folding actually better than the model predicts. Thoughts?

Tysen

[tigerite]

Yeah, it's something ICM doesn't consider; I try to allow for it myself before making a push, too.

However I think it's generally more of a factor when the button has more chips than you, rather than less.

[kevkev60614]

This is something I don't think about often enough. Most often I compare my $EV of pushing to the $EV of folding it to the BB.

DOes SNGPT just calculate your $EV as if nothing happened? So no action at all behind you, just the situation like it was before the blinds were posted?

[trojanrabbit]

[ QUOTE ]
DOes SNGPT just calculate your $EV as if nothing happened? So no action at all behind you, just the situation like it was before the blinds were posted?

[/ QUOTE ]
I believe that's the case.

Tysen

[eastbay]

True.

Now as an exercise it's kind of interesting to posit strategies for those behind you, and work out the EV based on the possible ways the hand can play out, and compare that to the fold around assumption. It's generally a smallish delta from the fold around assumption, smaller than the uncertainties associated with the range assumptions.

This is one of the reasons there's a "min edge" setting in SnGPT, to account for such effects on the baseline model. If the players behind you were wild, pushing and spite calling a lot, for example, this kind of effect might be significant. The alternative of course is to actually work out the consequences of assumed strategies behind you, but that's a hell of a lot of complexity for a nominal return.

eastbay

[eastbay]

[ QUOTE ]
DOes SNGPT just calculate your $EV as if nothing happened? So no action at all behind you, just the situation like it was before the blinds were posted?

[/ QUOTE ]

Not quite. "Fold $EV" assumes the hand is folded around, which isn't the same as the using pre-post chipstacks. Those numbers are also listed.

eastbay

[trojanrabbit]

Just trying to get an order of magnitude view on this.

If someone (button or SB) steals the blinds you gain +0.2% or +0.3%
If there's an all-in confrontation, you gain +2.3% to +4.4% depending on who it is.

What would you estimate is the chance of an all-in confrontation (either pre or post flop)? If it's only 8% of the time, then you are already at the 0.5% "min edge." I would suspect the chance would be much closer to 30% which would give you a +1.2% bonus. That's quite significant!

Don't want to slam ICM, I'm just brainstorming here... /images/graemlins/laugh.gif

Tysen

[eastbay]

[ QUOTE ]
Just trying to get an order of magnitude view on this.

If someone (button or SB) steals the blinds you gain +0.2% or +0.3%
If there's an all-in confrontation, you gain +2.3% to +4.4% depending on who it is.

What would you estimate is the chance of an all-in confrontation (either pre or post flop)? If it's only 8% of the time, then you are already at the 0.5% "min edge." I would suspect the chance would be much closer to 30%

[/ QUOTE ]

For SB and BB behind you, I would say something nominal would be pushing top 30%, calling top 15%, which is less than 5% showdown. 30% seems like a crazy number to me. That would mean nearly even-stacked bubbles would last about 3 hands. Is that what you see? My even-stacked bubbles usually last a lot more than 3 hands.

BTW, there is no "the" min edge. 0.5% is just a default.

For UTG, it's higher, but then you also have to consider the fact that you're taking the blinds next hand, which further erodes your equity if you fold, which tends to offset any action effect.

eastbay