Barry Carter may have just written a book about ICM but he also wanted to make sure people know when they should not trust it 100%.
In case you have missed it, I am going to be talking about ICM a lot in the next few months. ICM is the most important strategic consideration for a tournament player and understanding it makes the difference when you are playing for big money.
With that said, there are limitations to ICM and I would not be doing my due diligence if I didn’t acknowledge them. Here are three example for how ICM does not always have the answers in tournaments and why you should have a more robust approach to decision making.
ICM doesn’t account for skill
An ICM calculation gives you an aggregate valuation of your chip stack based on each remaining players likelihood of finishing in each position. It doesn’t factor on which players are better. So if Michael Addamo and myself both had the same stack, they would be worth the same in ICM terms.
For the most part this doesn’t matter too much because the later you are in a tournament the smaller the skill edges. A top pro might have a 60% edge with 100BB stacks but only a 10% edge when the average stack is 25BBs, for example.
Those edges still do exist, however. Doing a straight ICM based deal at a final table might not be a good idea for a top pro who should expect to win more on average than their equity suggests.
Edge also should impact your ranges. If you have a big edge, a comfortable stack and a slow structure, you probably should not risk your stack with the bottom of your range, even if it is the ICM approved range. You can wait for better spots. Likewise, if you are the underdog skillwise at the table and perhaps bottom in chips (so you can’t easily ladder) you might be better off calling wider than ICM dictates if it gives you a chance to get back in the game at a table of crushers.
ICM does not consider future game
The calculations we provide in my book Endgame Poker Strategy and most of the ranges you will see in ICM solvers give you the best decision you can make right now. They might, for example, say that you can profitably shove 18% of your hands under the gun based on your stack size, your opponents’ stack sizes and the payouts.
What these calculations do not account for is what happens after the hand in question. They don’t account for the blinds going up and/or you posting the big blind next hand.
Let’s say you have 3,000 chips and the big blind is 1,000. Your shoving range here might be something in the region of 35% of hands. However, if the big blind is going to be 2,000 next hand and you are the one who is going to be posting it, that should impact your range now, which most ICM calculations don’t factor in. You should probably be shoving much wider than 35% of hands here because the prospect of posting 66% of your stack blind the next hand is terrible. It’s much better to take the risk now.
ICM breaks down when stacks are incredibly shallow
There is a fundamental principle of ICM which is the fewer chips you have the more each is worth, and that you should play tighter, not looser, than you might think as the short stack.
There are inflection points, however, where this is not the case. We have already seen this with the future game example but you will find when you look at solvers a lot that there becomes a stack depth where it’s more profitable to play closer to a ChipEV style. It’s relative, however, if you have 5 big blinds when everyone else has 30 big blinds this might be the case, if you have 5 big blinds when everyone else has 7 big blinds then ICM principles will hold up.
To give an extreme example why ICM breaks down in these spots, imagine a spot three handed where the chip leader had 30,000, the 2nd stack had 20,000 and the short stack had 10,000. A quick ICM distribution would suggest the chip leader wins 50% of the time and comes third 15% of the time.
|Player 1 (30k)||50%||35%||15%|
|Player 2 (20k)||33.3%||40%||26.7%|
|Player 3 (10k)||16.7%||25%||58.3%|
But what if the big blind was 30,000, forcing everyone all-in? In that example in order for the chip leader to come last Player 3 needs the best hand, Player 2 needs the second best hand, then after that the chip leader would have to lose the following hand to be the first player to bust. That would actually happen roughly 6.2% of the time, not 15% of the time. The chip leader wins the tournament way more often than ICM suggests in this example.
This is a very extreme example and skill is not involved. However, it’s not as ridiculous as it sounds, I recently found myself at the final table of a super fast freeroll and the chip leader had four big blinds with the remaining four players all had under two big blinds. Everyone was basically forced all-in within an orbit and the dynamic wasn’t unlike the one I just suggested.
I think I have picked the three most common examples of the limitations of ICM but broadly speaking I still believe it is the most important strategic consideration in poker tournaments. However, it is important to know these limitations so you can think reflexively at the tables and while making final table deals.
What do you want to know about ICM? Let us know in the comments: