In poker, the classic joke is that the answer to any question you may ask is “it depends”. While it may seem silly at first glance, there is much truth to that statement. Nothing in poker is an absolute, and the ICM is no exception to this, as we've covered in the last article in the series. There are several issues with the Independent Chip Model that should be addressed before anyone delves in the magical world of math and expected value that is the ICM.
The main complaint about the ICM is that it's a complex set of equations, and that no one can solve such a complex series at the table with the amount of time that most poker sites give you. This is a very true statement, for the most part. However, just because it's no use at the table does not mean that it's not useful. Doyle Brunson relied more on feel than on the math of a situation, and it certainly worked for him, as well as many others. But what is typically misunderstood is that players that develop a “feel” for the game have simply internalized the math, as opposed to relying on quantifying variables. Neither way is incorrect, but typically, the math of the situation will explain a play made by a player who “feels” it's the correct play, even though they have no knowledge of the equation.
The fact is that the ICM is extremely difficult to use in real time at the table. What you learn from examining situations using ICM in hindsight, however, has many, many benefits to your game, and the more information we can gather, the better our chances of making the correct decision.
Another situation in which the ICM tends to fail is when the blinds are large in relation to at least one stack on the table. Some people consider a stack of 1-2 big blinds to be where ICM begins to fail because of large blinds. Personally, I feel it's closer to 3 big blinds, and most people feel that “large” is closer to my estimate than 1-2 big blinds. In any situation, this can cause problems with the ICM, especially in bubble situations. In these types of situations, where a folding war defense may be more useful, ICM can be less accurate, although it will often give the correct answer.
Accuracy also comes into play with the number of players remaining in the tournament. ICM is most accurate when you're looking at a range of 3 to 5 players. When there are two players, ICM and actual tournament chips correlate perfectly, so while ICM works just fine with two players, it's quite pointless to calculate, as calculating with chips works just as well in that situation. As the number of players increases, it becomes more difficult to predict the actual outcome, because of the number of variables that come into play in tournaments that ICM doesn't capture, and are very difficult to model in general. While it still works well ten-handed, it works best near the bubble and after the bubble has ended.
The other issue that many have with ICM is the number of potential assumptions you have to make. In a simple ICM calculation, you only need to assume the percentage of time that you'll win the hand, but in more complex ICM situations, you may have to make assumptions about multiple hands, the likelihood that each specific player may call you, and the likelihood someone will call you based on if another player calls you. At this point, critics say that with the number of estimates that you're making, the numbers that ICM will give will not have the precision needed in order to make an accurate decision, so there's no real point in making the calculations to begin with.
ICM is not perfect. In reality, we have no way of actually approximating our average winnings in a tournament precisely and accurately. The number of tournament chips is one way to approximate this payout, but it does a poor job of calculating accurate pot odds. ICM takes into account the actual payouts, and because of this, it tends to be more accurate than estimating using only chips. It's far from perfect, but until something better than ICM comes along, it's a pretty good system, and it does a fairly good job at solving tough decisions.
This is the final article in the series on the Independent Chip Model. Now that you've made it through the third article, my next suggestion would be to start practicing with the ICM. If you get into a difficult decision in a hand, save the hand history, and analyze it using the ICM, and see where the math is coming from. I promise you, the more you practice, the more your tournament game will improve in these situations. If you're ever having problems applying the ICM, or you don't understand why it is telling you to fold when you think it's a pretty clear call, don't hesitate to visit the Sit-N-Go forum. There are many knowledgeable and helpful people there, and will be more than willing to help assist you the best they can.