“I always say this — I think we’ll see better AI models in the coming years that outperform current ICM models. ICM is just a bit too one-dimensional. It doesn’t fully capture the depth of actual poker play — especially when it comes to stack dynamics.
Certain stack sizes clearly have advantages or disadvantages, but ICM just chops up the prize pool and says, ‘You get this much, you get that much.’ It doesn’t account for the pressure players are under or the edge that bigger stacks can apply.
So yeah, big stacks are often undervalued, and short stacks are usually a bit overvalued — especially in high-pressure spots. ICM is still a valuable tool for understanding direction, but you always have to take it with a grain of salt and apply real poker logic on top.”