I'm just thinking of limitations of the icm model and in 5 minutes have thought of quite a few, surprising as we all seam to use it as the complete truth.

- proximity to the bb (often of more important when)-
- immanent blind increases
- model assumes equal abilities
- spots where we can obtain the chip lead and the icm bubble pawnage spot (more valuable vs people who understand icm)
- the preservation of fe (not getting blinded out )
- reg on reg dynamics
- psychology ( for example at 50/100 blinds a 1019 seams to have a disproportionate fe vs a 989 bb open shove its over 97 percent of the other chips stack but I bet it doesn't have 97% of the fe
-assumes opponent has a decent internet connection and isn't just about to be sent to bed by there wife/mother lol

any you can thing of please add to post.

Not sure what you are looking for. Yes it has limitations. No, you shouldn't blindly follow icm calculators. Some people know this, some don't. The ones who don't can be taken advantage of. And the ones who knows a little bit more that everyone else here probably wont post it for everyone to see.

"any you can think of please add to post".I think this topic is too general and just seems like free massinformation about something that people would rather keep to themselves. Perhaps if you posted specific hand or more specific setting someone would go into detail on this. But this topic is to broad.

peace

ICM just models the value of your stack under certain assumptions. It's a starting point not a destination.

How are you supposed to take advantage of someone who is following a ICM calculator strategy? I would like to know this so I can crush Russian bots.

Several of these are captured by an icm ev calculation some are present and although a limitation are simply due to viewing one hand in isolation, chipev has the same issue.
Immanent Blind increase/proximity to the bb would be the same for any calculation. Even if in the early stages and roughly chip ev you still have to take this into account it is not really just an icm model issue.

reg on reg dynamics/psychology are captured by the ranges you use when you perform the icm ev calculation. eg if you have 12bb in SB and it's folded to you you have to decide if the reg BB is going to call you light, this is not an icm issue and an icm calculator with the correct reg ranges will give you a decent result.

ICM is kinda the nuts. If I could i'd change my name to IM_ICM.

ppl always look for reasons to ignore ICM.

Yeah that was badly formulated on my part. I didn't mean you can abuse someone playing perfect . Think I might have missunderstood OPs question.

Err. Not sure I get what your saying? If we keep all the good information to ourselves surely this forum stops being usefull to anybody, as we all try to be information parasites, digesting any useful information posted by others but never trying to contribute anything.

If you feel this topic may touch on something so valuable to the individual they would hate to divulge that information. then it seams like a good topic to me.

Its not like by contributing you would allow yourself to be exploited. Its a general topic about the limitations of ICM. What about posts where a betting line is revealed that is massively exploitable.

What percent of the player pool do you think Is reading 2 plus 2, imo we are not trying to profit from each other apart from by sharing knowledge. We are just trying to better exploit the tendencies and errors of the more recreational poker players.

peace

exactly what i was thinking...

You know, F.E. doesn't stand for fold equity.

It obviously stands for fast expand...

FGS though

I live breath eat and sleep ICM if all else fails I have ICM

LOL. Good one

ok this thread didn't work out as intended if a moderator could please pull thread. I was expecting responses like not taking into account when we have a tiny stack and our opponents implicitly collude against us or when we do this to someone else. nvm. Guess everyone wants to analyze particular hands and abandon general reasoning and the limitation of a mathematical model we all use every day. Sod trying to learn anything lets get back to point scoring lol.
just reread this and i sound like a gripey child. Hope i can find an emoticon of tongue out

found it yay

ICM overvalues very short stacks.

ICM does not account for card removal/bunching.

im sorry please elaborate. i agree ICM does overvalue very short stacks but my thinking is regards to implicit collusion of the two bigger stacks and factors like proximity to bb which factor more heavily when very shallow.
What do you mean by card removal /bunching?

Check out the value given to a 1-chip stack by ICM. Yes, the overvaluing has to do with future considerations as you illustrate.

Thanks for the followup because card removal/bunching is really an issue for some ICM calculators and not ICM itself. It has to do with information you can gain by knowing the action prior to your decision. For example, if a lot of people fold to you in late position, the likelihood of a strong hand behind you is somewhat greater. So while a given villain's calling range will be the same, the better hands in his range will be slightly more common. The hands are not weighted in many calculators.

ICM really only models the value of a stack at a table and personally I don't really think that icm overvalues small stacks that much at all.

Here is an example of the type you suggest, for a standard STT type with 4 players left.
Player: 0, 4500, Equity: 33.3303%
Player: 1, 4500, Equity: 33.3303%
Player: 2, 4499, Equity: 33.3279%
Player: 3, 1, Equity: 0.0115%

So player 3 has roughly 1 hundreth of 1% of the prize pool compared to 33.3% for the others, this is a lot more than the chip value of 1/13500 suggests but here there is always the chance of a crazy AA v KK style clash to give a decent payout to this 1 chip, seems reasonable.

Admittedy, when you are far from a bubble icm does overvalue a single chip
eg, same STT
Player: 0, 2250, Equity: 16.6662%
Player: 1, 2250, Equity: 16.6662%
Player: 2, 2250, Equity: 16.6662%
Player: 3, 2250, Equity: 16.6662%
Player: 4, 2250, Equity: 16.6662%
Player: 5, 2249, Equity: 16.6602%
Player: 6, 1, Equity: 0.0086%

but in this case I think chip ev is also overvaluing a single chip, it's then 0.0074%. The only real chance of getting some money is a couple of really flukey hands with multiple big clashes before we hit the bb or while somehow our 32o hitts hard and wins a few times

Imo the algorithm for icm does drop the $ value in a reasonable way as this small stack size shrinks toward 1 chip.

There will be plenty of old posts about the icm algorithm, I haven't looked but I suspect the search feature will find plenty of good info.

ICM assumes "perfect" play, if you think your opponents won't play "perfect" you should exploit him.
However it's only "perfect" in it's narrow limitations. The bigest limitation by far is the ignorance of position. Let's say you end up with 2bb on a 9handed table. ICM assumes that you have a certain percantage EV of the price pool, that doesn't depend on position. Obv. thats not correct. To be on the button in the next round is way better than being hit by the blinds. Future Game Simulation (FGS) adresses this issue. It's supposed to be more accurate but not perfect. Good calculators (at least HRC) come also with unrestricted ranges, which makes calculations more accurate.
I'm more than happy with my Holdem Ressources Calculator, as it let me do everything, is fast and accurate but there are other products on the market as well.
For other exotic influences like bad internet connections, people have to leave: If they will sometimes call and sometimes fold when they shouldn't their calling range in percent stays the same but it gets weaker which lets you push wider. I've experimented with it a bit. E.g. if villain is supposed to call with KTo, you can let him call with some K9o combos while he is folding some KJo combos. A pretty cool function of my new toy but it really doesn't change that much...
You may wanna download the trial. The topic was always fascinating for me and I'm always happy to discuss it.

My HRC takes card removal into account, even in ICM mode, I don't know if others calculators do the same though.
What is bunching?

I don't actually know the technical definitions and whether they are true synonyms. I always think of card removal as adjusting ranges based on known cards that cannot be in the range (e.g., hero's hole cards) and bunching as adjusting ranges based on known actions (e.g., folded to the SB, where we gain probabilistic information on BB's holdings). I probably used the terms synonymously where I shouldn't have. (I'm really paying for posting ITT without thinking carefully ).

Since I have us off the topic of ICM limitations already, how big is the weakness of popular calculator's FGS mode where they assume Nash ranges for future rounds?

As a serious response while evaluating probable player tournament scenarios, the ICM does not take into consideration the player’s current position at the table. Player skills, as in the ability to read hands, call bluffs and intimidate your opponents cannot be factored into the program.

The ICM does not consider status of blinds either, thus the approximate amount of your chip stack might probably not be as close to home as you like. Without the addition of these assumptions and limitations to the ICM and its computational process, the model is far too complex to create finish scenarios as well as add up probable stack value.

In my mind the biggest limitation doesn't lie in the program itself but in the limited system ressources we have on our computers. In order to get accurate the program has to simulate many future hands and thus needs a lot of calculation power.
To demonstrate that I simulated this hands in various modi:
4 players, everyone has 500c, blinds are 50/100. Payout 70/30 For simplicity I will report only the results for when everyone folds to the button and he has to decide wether to push or to fold.

ICM:
But pushes 36,4%
Sb defends: 14,5%
BB defends 26,9%

FGS1:
But: 33,3%
SB 11,5%
BB 27%

FGS2
But: 40,9%
SB 11,3%
BB 31,5%

FGS 3
But: 39,7%
SB 17,6%
BB 26,4%

FGS 4
But 34,4%
SB 14,9%
BB 28,3%

Note that the "unexploitable" defending range of the SB is between 11.3% and 17.6%. While this seems pretty random it actually makes perfect sense once you understand how these programs work. Let's look at it from the perspective of the SB. In Depth1 he will be on the button, so he should defend less than ICM suggests, in order to get in this favourable position. In Depth2 he will be in the CO, so he should call even tighter, because the next 2 rounds are very +EV. IN Depth3 he will be hit by the Blinds which makes him want to call more in the actual situation, cause if he calls and looses he won't have to pay the BB. In Depth4 he will be in the very same position again, so he should call with a range, similar to ICM. In fact he calls somewhat wider even if the But is pushing less. This is mainly due to increased FE when he doubles up.
Having that said, I still have no idea what the best response would actually be. What would be everyones range for that scenario? For a definite answer we would need super computers that can solve the entire game.