If you understand ICM really well, perhaps you could produce a document which will enlighten many of us. I have been through the ICM descriptions that I could find and think about it in specific situations, but the core concept still eludes me. Ramifications of an aging mind no doubt, but it isn't yet a dead mind.
Caveats:
ICM is only a model, not reality. The assumption that the relative size of your stack is a good predictor of your chances to finish first is often a dubious one. IMO, the two main flaws are
1. At the start of a tournament, ICM says that everyone involved has equal equity. Obviously this does not take into account the relative skills of the players. if you have a proven track record of having 20% ROI at the stake you are playing, then you should adjust your model to take that into account. This means calling even tighter than ICM suggests. This is exactly what you see happen on the first days of the
wsop main event. Most of the pros are playing excessively prudently as they value themselves much higher than their amateur opponents and that affects their $EV guesstimations
Incidentally, this correction may also justify that a fish takes ridiculous chances early on. If you adjust ICM estimates to a very negative ROI, the resulting $EV should actually push you to decrease your calling requirements. Actually, if I had to play a table full of pros such as you do in the Poker after dark show, for instance, I'd be delighted to take 50-50 coin flips against them. That's a much better chance then playing skill vs. skill.
2. ICM does not take into account the dynamics of the table but only the current hand. For instance, if you are the smaller stack and you are being bullied by your opponents, it may be strategically better to take a chance earlier with improper ICM odds rather than wait until your stack becomes so small ICM
tells you to call. If you are a big stack and the table lets you bully it on the bubble, then it may be better to not break that bubble too early.