ICM

[rurollin]

Can anyone please explain ICM, what it is, and how it works in the latter stages of a sng..?

I just read about it in a recent Bluff, but they never went explain it, nor did I pull up anything with a search..

[Mingdu]

I could try ... but my attempt couldn't come close to the info and links in the FAQ

[Beerocrat]

read FAQ
IBL

[wiggs73]

Here's the link that's in the FAQ.
http://archiveserver.twoplustwo.com/...te_id/1#import

[eurythmech]

In after lock

[DevinLake]

Here's something I wrote for the new FAQ. Please anyone feel free to critque it (Slim?).


ICM

What is ICM?

ICM models your equity in the tournament based a stack sizes, where your equity is a representation of your share of the prize pool. So, if you stack relates to 50% equity. It means you should win, on average (assuming equal skill), 50% of the prize pool.

ICM uses the various stacks sizes at the table to determine the probability of you having a given finish distribution. Based on that distribution, your equity in the tournament can be calculated. So, for a single table tournament with a 50/30/20 payout structure, ICM is used to determine the probability of you coming 1/2/3 or out of the money. With these probabilities, a dollar value can be associated with your stack size.

If you are really inclined, here’s the math:

P^m_i = prob. of mth place for ith player
P^m_i = sum_{k!=i} P^1_k*P^m-1_i(S_~k)
where S_~k is the stacks after removing the k'th stack.

The recursion is closed with:
P^1_i = S_i/sum_k S_k

So, if you know how much equity your stack gives you, how does that help you make decisions at the table?

Well, most people use software like SNGPT and other programs, which uses ICM as part of the analysis. After a session, you can review all your hands, or hands that gave you trouble, to determine if you were making +$EV decisions.

So, when deciding if a 'move' at the table is +$EV you have to determine the equity you have based on the different scenarios that occur from that 'move'.

Looking at the simplest case where you open raise all-in, the following scenarios can occur:
-Everyone folds;
-You get called and win; or
-You get called and lose.

Programs like SNGPT assume only one person will call. Each of these scenarios results in a distinct chip distribution. So, ICM must be used for each scenario to determine your equity for each.

You then use calling ranges to determine the percentage of the time everyone folds or someone calls. Those hand ranges are then used to determine how often hero wins or loses the hand.

So, hand ranges are used in conjunction with ICM to determine the $EV of a given 'move'. Dethgrind gives a more detailed example here .

[eurythmech]

[ QUOTE ]

So, if you stack relates to 50% equity. It means you should win, on average (assuming equal skill and ignoring relative stack positions), 50% of the prize pool.


[/ QUOTE ]

I would propose this change.

[DevinLake]

[ QUOTE ]
[ QUOTE ]

So, if you stack relates to 50% equity. It means you should win, on average (assuming equal skill and ignoring relative stack positions), 50% of the prize pool.


[/ QUOTE ]

I would propose this change.

[/ QUOTE ]

Ughhh....will anyone reading the FAQ actually understand what is meant by that? It needs to be pretty simple IMO. Also, understanding the importance of position is a skill!?

[eurythmech]

It might be true that this would be too complicated.
But ignoring skill, relative stack positions have effect on your EV, but not your ICM value.

[gulon]

In layman's terms, ICM compares your chances at getting ITM by shoving and the chances of getting ITM if you fold. If your chances of ITM by shoving > chances of getting ITM by folding then you shove.