When does ICM become significant in MTTs?

ICM is great at evaluating tournament stacks, but until recently, it was nearly impossible to calculate for large-field MTTs with many players remaining. So instead, everyone has been evaluating tournaments with Chip EV models when many players remain. Recent advances in poker science now allow us to calculate ICM when hundreds or even thousands of players remain.

This begs one important question though – at what point does ICM become significant in tournaments? When is it worth switching to ICM solutions? We ran a series of experiments in partnership with Helmuth, founder of Holdem Resources Calculator, to find out!

• Chip EV = A strategy that maximizes chips
  
• ICM = A strategy that maximizes dollars

To test this, we simulated push/fold tournaments with various competing strategies that would switch from chipEV to ICM evaluations at different points throughout the event. This allows us to test at what point switching to ICM has a significant impact on results.



ICM starts to significantly impact results between 37% and 50% of the field remaining (15% payout structure). I would hazard a guess that ICM becomes critical at about 3x the bubble.

We also ran detailed placement and EV analysis to see where the ICM player's edges came from. This graph shows the cumulative ITM payouts by strategy. The ICM strategies place ITM a lot more often, hanging on to their stack until they can secure a placement.

ChipEV strategies win 1st place more often, at the expense of their ROI%. This is a result of trying to maximize chips rather than maximize money.



Check out the article above for more experiments, details on the simulations, and detailed analysis.

But what about more complex models that allow for postflop play, preflop raises, etc? We compared GTO Wizard solutions at different tournament phases to identify how the strategy changes throughout the MTT. As you can see, the strategy changes significantly even halfway through an event.

BB facing LJ open:




What do you guys think about the significance of ICM in the early to mid stages of tournaments?

somewhere around 1/3 of the field remaining feels like the inflection point but the degree to which you need to care about icm--if at all--goes up and down beyond that point.

with the first set of simulations it would be interesting to isolate how much impact being reckless and stupid around the stone bubble contributes to the overall results. My guess is that alone is a huge driver and together with a lack of icm adjustments at FTs makes up almost all the roi difference

The icm impacts are distibuted nonlinearly throughout the interval where they become relevant. Surely once a stone bubble pots the icm forces are relaxed, right?

Certainly, ICM becomes more impactful near the bubble, then it relaxes a bit before skyrocketing at the final table. A lot of it comes down to the payout structure though.

This graph has been around forever, which demonstrates average bubble factors in a large-field MTT:



The (cEV 75% / ICM 25%) strategy was losing a lot of value being too reckless well before the bubble. This strategy uses chip EV until 75% of players are eliminated, only switching to ICM in the last quarter. But the bubble doesn't occur until 15% of the field remains.

What's the % threshold for a race on 1st hand of WSOP ME?

Assume decent player who doesn't care about "lasting longer" independent of money considerations

I'm thinking 65%... Maybe 62%

Off the top of my head...

ICM always has some impact in a tournament, because chips you win are less valuable than chips you lose, but for most of the tournament it's not so pronounced that you need to go far off chipEV. If it's a spot you think is really close and you'd be calling off a good chunk of your chips, I'd generally fold. When it's really pronounced, like at the bubble and the final table bubble and the early stages of the final table, you should be avoiding situations where you can bust unless you can get it in as a substantial favorite, or you're so short (and this usually means shortest in the tournament, as well) that you just have to make a stand when you find a hand good enough for it.

One reason you see more aggressive lines in tournaments is that there's value in taking pots down with showdown, whereas there might be more chipEV in taking an alternate line that would let a player bluff if they miss. In general, if two lines have the same EV, you'll want to take the one that's more likely to eliminate the risk of you losing the pot, especially the bigger the pot becomes and the later in the tournament you get. (With re-entries pretty common, even if you're deep you can take some chances while you're around starting stack. I don't recommend intentionally gambling it up, and I never rush to double or bust if I lose a lot of chips early and there's still the re-entry option, but it does mean ICM's effects are less pronounced.)

My impression is that small field and big field bubbles are quite different animals.

Case 1 : 200 players - 30 paid - 32 left
Case 2 : 2000 players - 300 paid - 320 left

The bubble will last longer in Case 1, cause the field gets nitty.

So if we are actually closer to the money (and we should make bigger deviations from CEV) in Case 2.

I recognise that graph from Dara O'Kearney's book Endgame Poker Strategy. Not sure if it was first published there?

It's important for players to realise not only what that 1.6 average represents, but also how massively this varies depending on the relative stack sizes. 100K stacks can call off 10K shoves easily. 90K stacks need a monster hand to consider calling an AI shove from the 100K stack.

So, you need to think well beyond the 1.6 average on the FT and note that the actual bubble factors players face can range from 1.06 up to 2.9 or more.

Here's a chart showing the intersecting bubble factors for players on a hypothetical FT with stack sizes 10K 10K 20K 40K 50K 50K 60K 90K100K. There are a few different scenarios presented and this one is for a stone bubble on the final table. Note: I haven't sought permission to reproduce the work here, so I'll just beg forgiveness in advance and encourage everyone to buy Dara's book which is available on your phone via any ebook store

I use the 50% field to start ICM adjustments. The adjustments aren't that drastic at that level, but there is some.