i hear a lot of times that ICM isn't always the best option, but i really dont know in what exactly moment is best ignore ICM suggestions.

if someone can say more about that i'll be grateful.

Mostly it's just something u have to learn with experience. The most common situation however is when hitting the blinds will cripple u to the point that u have no fold equity left, which ICM doesn't account for.

Deviate from ICM when doing so is more +mEV than following it.

some other examples not mentioned above might also include meta-game considerations... Like if I defend my blinds early when the chips are worth less then they won't get stolen as much later when worth more... Not sure if this qualifies or is good, but can be useful to think about versus some villains.

Also, of I push as much as ICM says, then maybe my image will be looser than I think it is.

But I'm just learning too, so what do I know?

ICM is not a strategy. It is an model that equates tournament chips to real dollar equity. By using this model, and by assigning accurate opponent ranges, you can theoretically play late stage push fold poker that is close to optimal.

There are some failings of ICM in assigning true equity to a player. The main one, as mentioned in the first response, is that it usually gives too much equity to short stacks who have lost their fold equity. Where the button is and who will be hit by the blinds first is another one. Another major one is a players skill advantage. ICM suggests that all players have equal equity at the beginning of a tournament, but this of course is not the case.

I know from my own experience of double or nothing tournaments, that ICM fails because of the strategy of that format. ICM works a lot more optimally when all players are trying to accumulate as many chips as possible. Big stacks making calls with the right pot odds is almost always a plus EV (according to ICM) play. But in double or nothings the big stacks don't have any incentive to collect chips. In these tournaments, the chips move from the large stacks to the middle and small stacks on the bubble, which is the exact inverse of how a bubble works on a reg format 9 man. Because players have different incentives with respects to bubble play, there is no way that the same equation for assigning true equity could be equally applied to both formats.

From my experience if ur getting THAT low on blinds <8, SAGE is a better system to use.

I'm pretty sure that SAGE is a strategy, mostly for heads up, that incorporates the equity assigned from ICM and using ranges from Nash.

ICM covers all the math and dollar equity factors for that particular hand. Why to deviate from this is when there are other factors to consider. For example consider this spot:

Blinds: 150/300
CO: 3000chips, fold
D: 3000chips, fold
SB: 6000chips, All-in
BB: 3000chips (hero) ATo

Say you are on this bubble spot and u'd assume the sb's shoving range to be 100%. ICM will tell you that it is slightly correct to fold here, but what factors are there besides ICM. If everyone is ICM knowledgable, it could prove profitable to make the call because of the future equity you'll get in being able to abuse the bubble after your double up. What if there is 1-2 donks in the game who know nothing of ICM with will snap call Ax on the bubble for their entire stack? Then it would be correct to fold a hand that is +EV to call according to ICM (like AKo), because of the future equity you'll get from the likelihood of someone else bubbling themselves.

How to weight non-ICM factors isn't exactly an exact math.

Some things to think about:

Your table image
When are blinds going up
How many opportunities are you getting to be first into the pot
Are your opponents giving you a lot of walks, or putting a lot of pressure on your blinds
Are your opponents tangling a lot with each other
Images other players have of each other
Will you still have FE if you lose / will you have a big enough stack to own bubble if you win / would winning these blinds give you a strategic stack siz advantage

Many of these things will just help you estimate ranges so you can have a better idea of what the "correct" ICM play is, but some considerations extend beyond that

Sometimes, one -EV push will gain you enough chips so that you may make a +EV push the next group of hands. Don't forget the idea of pushing is to keep stronger hands than yours from calling, this works well. When you make these pushes don't use hands with naked aces or kings because those hands will be in the likely calling ranges or players. Instead use 78s, T8o 22 type hands.

I hear this so often (especially from tv "pros"), but there really is no calling range where jamming 78s is better than jamming A2o (they play quite similar in fact)... unless it's some weird calling range, like any pair or ace, but no broadway hands.

Well written

ICM is never accurate as it starts with the assumption that everyone plays equally good which obv. isnt the case.

This is true but only to a very small extent due to card removal effects. Also villains love to call broadway hands and Ax/Kx plays better against these then 78s.

Ilya - v nice post.

op, when you're under 5bbs, don't worry about icm


---
3,636,933,696 games 5.267 secs 690,513,327 games/sec

Board:
Dead:

equity win tie pots won pots tied
Hand 0: 68.128% 63.34% 04.79% 2303715108 174049878.00 { 33+, A3s+, KTs+, A7o+ }
Hand 1: 31.872% 27.09% 04.79% 985118832 174049878.00 { A2o }


---



---
1,356,144,768 games 3.065 secs 442,461,588 games/sec

Board:
Dead:

equity win tie pots won pots tied
Hand 0: 62.866% 62.43% 00.43% 846670104 5881374.00 { 33+, A3s+, KTs+, A7o+ }
Hand 1: 37.134% 36.70% 00.43% 497711916 5881374.00 { 87s }


---

This doesn't quite tell the whole story though as when you shove the A2o hand you'll have more fold equity because you're holding one of the aces and Ax hands make up a very disproportionate section of his range (PokerStove accounts for card removal range weighting, but not for the this - you could work it out exactly if you find how many combinations each hand excludes [sorry, got some odd jobs to do so no time atm to do this]).

Juk

Yup, SAGE is a strategy for headsup, and as such there is no ICM needed for it, simply because in hu your chipEV = $EV. It's not using perfact nash, but something close to it, with an easy way of calculating/remembering. You should almost always use Nash anyway hu, no need for an easy to remember solution if you can just play with a Nash table open somewhere in the back.

I have heared people saying that you coul just shove Nash/SAGE from the sb and call from the bb in a stt, but thats not really 100% true, exactly for the reason that you have to factor in ICM for a situation with more than 2 players left.

You mean card removal right? Since we hold an ace, our opponent is much less likely to hold an ace.

Also, the following stoves show A/2 and 87s vs 10%, 20% and 30%. It's pretty obvious to see that as our opponents calling range tightens A/2 and 7/8s converge (both do equally as bad against top hands) but as opponents calling range loosens A/2 does much better:

10%

Hand 0: 33.645% { 87s }
Hand 1: 66.355% { 77+, A9s+, KTs+, QTs+, AJo+, KQo }

---

Hand 0: 34.853% { A2o }
Hand 1: 65.147% { 77+, A9s+, KTs+, QTs+, AJo+, KQo }

20%

Hand 0: 36.738% { 87s }
Hand 1: 63.262% { 66+, A4s+, K8s+, Q9s+, J9s+, T9s, A9o+, KTo+, QTo+, JTo }


---

Hand 0: 41.409% { A2o }
Hand 1: 58.591% { 66+, A4s+, K8s+, Q9s+, J9s+, T9s, A9o+, KTo+, QTo+, JTo }

30%

Hand 0: 37.641% { 87s }
Hand 1: 62.359% { 55+, A2s+, K5s+, Q7s+, J8s+, T8s+, 98s, A7o+, A5o, K9o+, Q9o+, J9o+, T9o }

---

Hand 0: 44.366% { A2o }
Hand 1: 55.634% { 55+, A2s+, K5s+, Q7s+, J8s+, T8s+, 98s, A7o+, A5o, K9o+, Q9o+, J9o+, T9o }

Michty, are you certain you used the correct ranges? I'm for example surprised QTs is supposed to be better than ATo. You may want to use heads-up ranges, those should be more applicable for SnGs.

I just typed in 10%, 20% and 30% into stove and let it do the rest of the work. But you're right they seem a little off...

Yep.

Juk

Using the PS Equilator in the "vs. one random hand" setting I get the following results:

10%

Hand 0: 34.257% { 87s }
Hand 1: 65.743% { 66+, A8s+, KTs+, ATo+ }

---

Hand 0: 29.495% { A2o }
Hand 1: 70.505% { 66+, A8s+, KTs+, ATo+ }

20%

Hand 0: 36.871% { 87s }
Hand 1: 63.129% { 55+, A3s+, K8s+, QTs+, A7o+, KTo+, QTo+ }

---

Hand 0: 38.781% { A2o }
Hand 1: 61.219% { 55+, A3s+, K8s+, QTs+, A7o+, KTo+, QTo+ }

30%

Hand 0: 38.851% { 87s }
Hand 1: 61.149% { 44+, A2s+, K5s+, Q8s+, J9s+, A3o+, A5o, K8o+, Q9o+ }

---

Hand 0: 43.179% { A2o }
Hand 1: 56.821% { 44+, A2s+, K5s+, Q8s+, J9s+, A3o+, A5o, K8o+, Q9o+ }


As we see, if we expect to be called by just one player (as will usually be the case in SnGs) and put the players to our left on tight ranges then 78s is indeed significantly better than A2o.

Cool thanks - this is what I was expecting - 78s to do better against a tight range but worse against a looser range. Have no idea how stove got the ranges I put in my hand when I put in 10% etc...

I'm guessing at ~15% they play about the same in terms of equity, which is good to know. Below 15% 76s should do better and above 15% A2 should...

I think Poker Stove ranks the hands in the order of value for cash games where you have to expect to play against multiple hands. The Equilator I use has several options to choose from with "against one random hand" beiong the most useful for SnGs. I don't know if Stove has those options, too (I don't use Stove myself).

From http://www.pokerstove.com/pokerstove/faq.php

Juk

pokerstove junkie +1