What Happens When Playing Chip EV and ICM Is Significant

Just how costly is it to play a Chip EV style when ICM is significant? We always knew it was a mistake, but we now have an answer thanks to this recent article from Tombos21. He started his search by mapping out the stages of a tournament. Via the data of single-raised and 3-bet pots, it’s possible to empirically demonstrate the average $EV loss of sticking to a Chip EV strategy.

It turns out that doing this can cost you as much as 30% of your buy-in in EV. It’s likely much more though, as only flop strategy was included in the experiment. Again, I implore you to read the article for more detail.

This was across 225 flops for both single-raised pots and 3-bet pots. I thought it would be fun and instructive to piggyback on the work Tombos21 had already done by going into depth in a couple of specific hands and replicating the experiment. This means nodelocking a player to follow the Chip EV strategy in an ICM hand and then observing how they get exploited by their opponent.

Preflop Ranges: Chip EV vs ICM

The first consideration is preflop. For a given spot, the ICM range will not be the same as the equivalent Chip EV range because ICM aims to maximize money ($EV) while Chip EV maximizes chips (cEV). This leaves us with a quandary regarding our experiment setup.

I say, let’s do both!

Preflop Range Comparisons

First, let’s see what happens when we simply use a Chip EV range when we should use an ICM range, but let the solver play as well as it can postflop afterward.

This is the 30bb Chip EV range for UTG:

And this is the 30bb ICM range for UTG near the bubble:

As you can see, it is slightly tighter (by 1.4%). The shape is also subtly different; we play fewer pocket pairs and medium suited connectors but more Ace-x hands for their blocker value.

In response to the open, this is what the BB does in Chip EV world:

Postflop Strategy: Chip EV vs ICM

Now, let’s compare apples to apples and give both players the ‘correct’ (ICM-adjusted) ranges preflop but not the correct postflop strategy. What we need to do now is give both players the same type of range but run the sim twice, once in the Chip EV model and once in the ICM model. After that, we nodelock the ICM simulation so that one player follows the strategy from the Chip EV sim.

We find ourselves in the following situation: 30bb effective, the CO has opened, the BB has called, and the flop is 932r.

I will study an ICM bubble scenario, so I’ll use the same ranges each time. In this case, 30bb symmetric near the bubble.

1. Chip EV Simulation

To kick things off, let’s look at the Chip EV spot. In all the next examples, the BB checks 100% of the time. This is the CO’s c-bet flop strategy for chips:

A mixed strategy with checking, medium-sized betting, and small betting each ⅓ of the time.

The primary bet size is 33% pot, and this is how the BB responds to it:

2. ICM (Near Bubble) Simulation

Up next is the CO c-bet flop strategy on the bubble of a 200-runner tournament. Both stacks have 30bb, which is also the tournament average on the bubble:

The BB has to play more cautiously on the bubble. They raise around 17% of the time here, compared to 24% of the time in Chip EV world. The CO bets small more often because they are less concerned about being raised. The underlying reason is that hands require more equity to put chips in the pot, resulting in less hands making the cut.

3. Nodelock Chip EV Strategy in ICM Simulation

Let’s now look at what happens when we mimic a Chip EV strategy on the bubble using nodelocking. If you want to do this yourself, I have done it the ‘easy’ way by inputting the Chip EV frequencies for each action rather than adjusting every single hand for the perfect frequencies. As you can see, the solver came very close to mirroring the Chip EV strategy perfectly. The frequencies are probably what matters the most anyway in this experiment, given we almost range-bet in ICM world and we mix everything in Chip EV world.

Now, when we make that 33% pot bet, this is how the BB responds (we have not changed the BB strategy to Chip EV):

The BB folds more often, 40% of the time, compared to 37% of the time when they faced the correct ICM strategy. They do raise more often, however.

This is not where the biggest error is being made. Let’s look at what happens when the CO goes for that 84% pot bet, which is never used in the correct ICM strategy. Let’s first look at the Chip EV response from the BB to that bet:

And this is the response to that same bet but on the bubble:

The BB folds less often, ~45% of the time compared to ~56% in Chip EV world. They also raise less often, more than 13% of the time, compared to ~15% in Chip EV world.

This is the response to the check/raise in Chip EV world:

And this is the response on the bubble:

The CO folds more and raises less. They are forced to play their range more conservatively. Remember, this is the solver adjusting the best way it can after making an initial mistake. A real-life opponent is less likely to see the error of their ways.

For fun, let’s drive it further and see what happens if the CO adopts the same Chip EV strategy when facing the check/raise by nodelocking the same frequencies. Let’s look at how the BB acts on the previous node, facing the 84% pot c-bet:

They fold a little more and raise much less. They raised 14% of the time in the last example but now only 10%. Another big difference is the shape of the range; it is very polarized now. All the best hands raise and a few trashy bluffs (that have some blocker and/or backdoor value). But in the previous example, there were similar value hands that called, and some of the middle-strength hands raised.

The BB has tournament survival to concern themselves with too, so they can no longer raise as wide if their opponent does not respect ICM. However, they do get to play all their best hands precisely how they want without having to look after the middle of their range too much.