ICM Discussion - Split thread

[tightpoker]

Darinvg asked my thoughts on ICM in the QQ thread, so I'm starting a new one here:

ICM to me is like that economics game of Free Trade vs Hostile Trade. A
quick sum of this game is like this:

There are 5 countries, each country holds up a Free Trade or Hostile Trade card each turn.

If all countries hold up a Free Trade card, each country gets +100 points.

If 4 countries hold up a Free Trade card, they get +0 points each and Hostile Trade gets +200 points.

If 3 countries hold up a Free Trade card, they get +0 points each and Hostile Trade gets +50 points.

If 1 or 2 countries hold up a Free Trade card, they get -50 points and Hostile countries get +0 points.

If no countries hold a Free Trade card, all countries get -100 points.

You have 10 turns to achieve the most wealth for your country. If your country doesn't receive 800 points, you lose the game - unless you have the most points, then you win the game regardless.

This is a rough recreation in my head of the game, but it's pretty close to prisoner's dillema. If you analze the situation enough, you'll see how the game should be played and how the game is played.

How the game should be played is everyone votes Free Trade and everyone gets 1,000 points. That way everyone wins the game and the people rejoice.

Then there's how the game is actually played. Most are smart enough to figure out that Hostile Trade can screw everyone else over while giving them enough points to win. So the game starts with some people going Free and some going Hostile. The Hostile countries do great until the Free countries get fed up, then drag everyone down. Everyone tries to agree to a truce to save everyone, but there will always be one country that realizes no matter what, they'll get the most points and win if they simply keep vosting Hostile and knife the other countries in the back. Thus, everyone goes down burning in flames except for the one country that stuck to their guns of screwing everyone else.

So, in the end, the dirty fighter wins 100% of the time in this game.

Now, you're probably wondering how this relates to ICM. Well, I see the pro ICM people as people who believe that Free Trade will work in the above model. If everyone plays along nicely, then everyone theoretically gets the most out of the situation. In ICM, it would be retarded not to deal at the final table if you could, because you should get your money out when you can.

In reality, everyone is a bastard. The person that can exploit the others will do better than the ones playing by the rules. That said, if everyone played like the bastard, everyone would suffer as a result.

ICM really expects you to fold in any situation where your chips are in danger, because it's just not worth it to gamble your ability to slide into the money vs your long term possibilities. It doesn't take into account ANY skill, ANY game conditions or ANY regard to bubble theory. Your chips are simply worth your factor of the total chips in play against the prize pool.

I might as well use a club to open my windshield. Or I might as well say that if the Steelers have a 63% win record, then they had a 63% chance of winning against the Dolphins- regardless of any offense/defense matchups, previous record agains the Dolphins, home game/away game advantage, injuries, winning/losing streak and momentum build during the game.

Anyhow, I could go on forever on other stupid theories I've seen on 2p2, but most people believe in something or they don't. My take is that ICM isn't stupid, but it's certainly not applied in the way that it was meant to. I've heard so many proponents say "Yeah, it doesn't take skill into account, but most of the math is right". That's like saying, "Yeah, Seattle just lost Shaun Alexander due to injury, but since they've got solid team stats, it shouldn't affect them as a team."

--> MJ'S HEAD EXPLODES1!!!%@&@% <!--

Just my (humble?) opinion on ICM.

[LadyWrestler]

Previously, I typed in another thread:

"I mostly understand the idea, and believe in ICM, but it would probably serve me well to study it a little more extensively and better apply the principles. Thank you for bringing it up."

---

I said I believed in it because Curtains on 2+2 told me in a pm that I should be studying and applying the principles of ICM. (I respect his poker skills, so I wanted to study ICM more but kept putting it off.) A poster here reminded me of it and I studied it some more.

Sooo...after a more extensive study...I decided I do not like it! It is not a good fit (me and it). I cannot say that applying parts of the theory, at selected times, could not help some players. I am just saying that it cannot help me, right now. Sometimes I think I am already too cautious, if that makes sense as a reason for me not to use ICM.

---

I just read what I typed above and I am not sure it even makes sense to ME! I will try to be clearer:

In the ways that I believe I need to improve my game, ICM would be counter-productive...for me.

---

If none of this post makes any sense to anyone, I am blaming it on too many head traumas, causing neurons to fire when they should be still.

[sfustsh]

ICM is game theory, it's based on infinite samples that ignore specific conditinos (like you said).

Of course, I think that if you played perfect mechanical poker and never built any reads on anyone you could still win. Just not nearly as much and never 'a lot'. So if you applied ICM without taking reads / conditions into account you could win in the long run, but I believe a poker instinct has more to say about that!

Take implied odds for example. There is absolutely no way to know 100% of the time how much your implied odds are. If someone bets into you 100 chips into a 500 chip pot that's 100 chips to call 600 chip pot, 6/1 odds. That's pure mathematical fact. But if calling gives you a chance to bust someone if you hit even if your odds of hitting are less than 1/6 you should call. But how do you know you can bust someone? You can't! If you do hit, there's a chance your opponent will just fold, it's all speculation. It's all 'read dependent'.

So I don't believe in that sort of thing too much, since it's not "cold hard math" like poker odds are. It's game theory.

As for game theory itself, I've read that when computer algorithms were pushed heads up against each other in game theory "games" the ones that were most successful were the ones that always choose the "greater good" options and kept a log of which opponents cheated against them. That's probably a winning poker strategy, I suppose.

[Linuxrocks]

Just my (humble?) opinion on ICM.

Wow, MJ ! That's a long-winded explanation of why ICM sucks.

For the first time on this forum, I have to disagree with you. You have some valid points, but seems like you misunderstood what it is. It doesn't take skill, game conditions, maniac meter etc. into consideration. That's why there's a person clicking the button instead of a computer.

BUT, ICM applies perfectly when these things don't matter. What skill do you have to consider, when you have 4xBB to decide whether to push or fold in SB ? NOTHING. How did we get to fold or push ? ICM.

Yes, it doesn't help you if someone limps, you are itching to push your AT in SB (everyone in that thread concluded that it is a push, but I wanted to call) and it doesn't help in myriad of other situations.

As explained in NLTAP, there is always a mathematically correct strategy, but no one can get there, because no one knows the exact percentages. ICM gives you a tool in a very specific situations to determine what you should do if you know the percentages perfectly.

In short, I like ICM, but I don't blindly follow it.

[flash]

ICM simply tells us what virtually all poker players recognize intuitively: the more chips you have relative to your opponents, the better your chances of winning. Like pot odds it is simply another use of probability to aid the poker player in making a decision.

But if ICM was 100% mathematically correct, winning poker would be impossible (don't believe me? just try applying ICM to the initial state of a given tournament). If you are reading this forum, I will assume that you believe winning poker is possible, thus we have to assume that there are limits to ICM.

As a poker player our task is to determine the right tool for a given situation.

[flash]

BUT, ICM applies perfectly when these things don't matter. What skill do you have to consider, when you have 4xBB to decide whether to push or fold in SB ? NOTHING. How did we get to fold or push ? ICM.

Why do we need ICM for fold or push? Isn't that straight chip EV?

[darinvg]

Interesting.
Not sure where to go from here. It seems to me that ICM works pretty well in most situations. However, it HATES bubble calls example.
Therefore, does it just come down to putting the villian on a range and determining if we are ahead? I guess some more examples might be the best to figure this out.

[Linuxrocks]

ICM simply tells us what virtually all poker players recognize intuitively: the more chips you have relative to your opponents, the better your chances of winning.

No you don't. For humans it's easy to think linearly, but the equation is not linear. For ex. you gain 50% more chips, does your chance to win increase by 50% ? Does your chance to place second increase by 50% ?

[Linuxrocks]

BUT, ICM applies perfectly when these things don't matter. What skill do you have to consider, when you have 4xBB to decide whether to push or fold in SB ? NOTHING. How did we get to fold or push ? ICM.

Why do we need ICM for fold or push? Isn't that straight chip EV?

Nope, if there's a smaller stack ready to bust in the next hand on the bubble, it's a fold, unless you are like MJ or me who want to take that kind of -EV risk.

[Linuxrocks]

Interesting.
Not sure where to go from here. It seems to me that ICM works pretty well in most situations. However, it HATES bubble calls example.
Therefore, does it just come down to putting the villian on a range and determining if we are ahead? I guess some more examples might be the best to figure this out.

Like I said before, adjust the skill factor in SNGPT and try this again. I call in an instant in that hand. Yes, it's -EV. So, what ? If I get the chips, I am going for first. ICM by default doesn't take that skill into account.

You might have already seen this that is similar but I have AK

To open up some of your game, try calling with what looks like -EV calls in SNGPT but slightly +EV with adjusted skill level.

[tightpoker]

For the first time on this forum, I have to disagree with you. You have some valid points, but seems like you misunderstood what it is. It doesn't take skill, game conditions, maniac meter etc. into consideration. That's why there's a person clicking the button instead of a computer.

We might not disagree that much. In my rant, it probably appears that I hate all applications of ICM, but the main ship I was trying to torpedo is the blind following of ICM ala SNGPT. Situations where people plug in the necessary variables in tricky situations and simply say 'Ok, the math says fold/call/fold' just drive me bonkers.

I think that a skilled person, understanding ICM, could apply ICM is a weighted variable in intuitively figuring out if a call is good or bad. For an unskilled person to use ICM not as a varaible, but as a means, that's where it because a blunt tool and not one of refinement.

BUT, ICM applies perfectly when these things don't matter. What skill do you have to consider, when you have 4xBB to decide whether to push or fold in SB ? NOTHING. How did we get to fold or push ? ICM.

Well, in these situations, the math is so concrete that it can hold back the Colorado river. Usually there's very little disagreement at the critical conditions between ICM theory and opinionated 'optimal' theory. It's at the borderline junctions like bubble, short-handed, 10BB, high variable chaos situations where ICM becomes a lot more wobbly. These are also the situations that are the most + or - $EV in a game in my opinion, because they have so many potential paths that there is a greater possibility for optimal play. With a short stack, the number of nodes is pretty much 1.

As explained in NLTAP, there is always a mathematically correct strategy, but no one can get there, because no one knows the exact percentages. ICM gives you a tool in a very specific situations to determine what you should do if you know the percentages perfectly.

I agree here. I think most good players naturally are able to figure out the 'rough' estimates of the outcomes of each potential path they take. A lot of this has to due with experience, but knowing how to look ahead and feel out the $EV of a situation is what it takes to harness ICM correctly, but the ability to feel out $EV pretty much means that there is already a prerequiste skill of thinking beyond it already.

[LadyWrestler]

For the first time on this forum, I have to disagree with you. You have some valid points, but seems like you misunderstood what it is. It doesn't take skill, game conditions, maniac meter etc. into consideration. That's why there's a person clicking the button instead of a computer.

We might not disagree that much. In my rant, it probably appears that I hate all applications of ICM, but the main ship I was trying to torpedo is the blind following of ICM ala SNGPT. Situations where people plug in the necessary variables in tricky situations and simply say 'Ok, the math says fold/call/fold' just drive me bonkers.

I think that a skilled person, understanding ICM, could apply ICM is a weighted variable in intuitively figuring out if a call is good or bad. For an unskilled person to use ICM not as a varaible, but as a means, that's where it because a blunt tool and not one of refinement.

BUT, ICM applies perfectly when these things don't matter. What skill do you have to consider, when you have 4xBB to decide whether to push or fold in SB ? NOTHING. How did we get to fold or push ? ICM.

Well, in these situations, the math is so concrete that it can hold back the Colorado river. Usually there's very little disagreement at the critical conditions between ICM theory and opinionated 'optimal' theory. It's at the borderline junctions like bubble, short-handed, 10BB, high variable chaos situations where ICM becomes a lot more wobbly. These are also the situations that are the most + or - $EV in a game in my opinion, because they have so many potential paths that there is a greater possibility for optimal play. With a short stack, the number of nodes is pretty much 1.

As explained in NLTAP, there is always a mathematically correct strategy, but no one can get there, because no one knows the exact percentages. ICM gives you a tool in a very specific situations to determine what you should do if you know the percentages perfectly.

I agree here. I think most good players naturally are able to figure out the 'rough' estimates of the outcomes of each potential path they take. A lot of this has to due with experience, but knowing how to look ahead and feel out the $EV of a situation is what it takes to harness ICM correctly, but the ability to feel out $EV pretty much means that there is already a prerequiste skill of thinking beyond it already.

LadyWrestler's head just...
EXPLODED!

[Linuxrocks]

We might not disagree that much. In my rant, it probably appears that I hate all applications of ICM, but the main ship I was trying to torpedo is the blind following of ICM ala SNGPT. Situations where people plug in the necessary variables in tricky situations and simply say 'Ok, the math says fold/call/fold' just drive me bonkers.

......

I guess you are right. Seems like I was actually agreeing with you. Darn, my one chance to look clever disappeared j/k

I go back to my old idiom I said once in the old forum MJ is always right !

[flash]

BUT, ICM applies perfectly when these things don't matter. What skill do you have to consider, when you have 4xBB to decide whether to push or fold in SB ? NOTHING. How did we get to fold or push ? ICM.

Why do we need ICM for fold or push? Isn't that straight chip EV?

Nope, if there's a smaller stack ready to bust in the next hand on the bubble, it's a fold, unless you are like MJ or me who want to take that kind of -EV risk.

Ok, I agree but you didn't mention the small stack ready to bust in your original post. I was referring to fold or push as opposed to fold, call, raise or push which is straight chip EV, correct?

[darinvg]

Interesting.
Not sure where to go from here. It seems to me that ICM works pretty well in most situations. However, it HATES bubble calls example.
Therefore, does it just come down to putting the villian on a range and determining if we are ahead? I guess some more examples might be the best to figure this out.

Like I said before, adjust the skill factor in SNGPT and try this again. I call in an instant in that hand. Yes, it's -EV. So, what ? If I get the chips, I am going for first. ICM by default doesn't take that skill into account.

You might have already seen this that is similar but I have AK

To open up some of your game, try calling with what looks like -EV calls in SNGPT but slightly +EV with adjusted skill level.

Overall, I like the idea of finding ways to get away from ICM.
However, I've seen a couple of references to skill. It doesn't take any skill to call an allin; therefore, I don't understand how that changes things.

Also, are all of the non ICM people saying that your AK hand is an auto call?