|
|
#41
11-01-2005, 01:01 PM
|
|||
|
|||
Re: Challenging the basics behind ICM
[ QUOTE ]
I doubt any intelligent person who's well versed in science and math would prefer a theory that was derived empirically to one derived mathematically. [/ QUOTE ] not to turn this into a pissing contest, but i went to a top school and majored in math and minored in physics. i frankly don't care how a theory was derived, as long as it describes real-world phenomena accurately. my contention is that ICM has not been shown to do that. [ QUOTE ] 2) ICM is backed by ridiculous amounts of empirical data. I actually wrote a program on my computer at home that tests ICM; it's very easy to do. It's very easy to write a program like this one. [/ QUOTE ] does this program use actual tourney data or does it use computer simulations? if it's the former, i'd be very interested to learn more. i'm in the middle of a similar project that has unfortunately stalled. [ QUOTE ] Also, I'm sure data miners can show that results come very close to ICM in actual SnGs (the difference would be because people with larger stacks on the bubble tend to be more skillful, so this would distort things). [/ QUOTE ] i know some people have done this looking at their own statistics, but i'd love to see if anyone has done the analysis looking at observed stt's. i have around 1500 observed stt's and i'd love to try to test icm using them. or if someone has already done this, i would be very interested in learning more. 
|
|
#42
11-01-2005, 01:10 PM
|
|||
|
|||
|
Re: Challenging the basics behind ICM
[ QUOTE ]
I made a computer program in which players start with stacks that I plug in manually and it chooses two players at random to exchange chips as a random percentage of the smaller stack (evenly distributed between 0 and 100%). This is essentially what a skill-less poker tournament is. I admit that this has its flaws, but it's damn good model and it supports ICM. [/ QUOTE ] i didn't know that - that's interesting. have you run it enough to say that it's exactly ICM or just that it's close? if the % of small stack's stack you play for is random between 0 and 100%, doesn't it take forever to bust anyone? i think there's zero debate that ICM underestimates the big stack's equity in this chip distribution: {6999, 1500, 1500, 1} because 6999 can run over the 1500's. (ICM assumes that's a virtually identical equity distribution as 7000, 1500, 1500). that's probably ICM's biggest weakness right now.
|
|
#44
11-01-2005, 01:44 PM
|
|||
|
|||
|
Re: Challenging the basics behind ICM
I'm pretty sure 1500 tourneys would be useless for testing ICM empirically. In order to see what the equities are in a particular configuration of stacks you would have to see something very close to that configuration a very large number of times. I'm just pulling this out from down under, but I'd be surprised if you didn't need at least something on the order of 100,000 tourneys to do this and then you'd better have a lot of computing power and a lot of time.
|
|
#45
11-01-2005, 01:47 PM
|
|||
|
|||
|
Re: Challenging the basics behind ICM
[ QUOTE ]
I'm pretty sure 1500 tourneys would be useless for testing ICM empirically. In order to see what the equities are in a particular configuration of stacks you would have to see something very close to that configuration a very large number of times. I'm just pulling this out from down under, but I'd be surprised if you didn't need at least something on the order of 100,000 tourneys to do this and then you'd better have a lot of computing power and a lot of time. [/ QUOTE ] i don't know much about the stats involved, but the econ professor who was originally going to work with me said we could get some preliminary results at around 500.
|
|
#46
11-01-2005, 01:51 PM
|
|||
|
|||
|
Re: Challenging the basics behind ICM
[ QUOTE ]
[ QUOTE ] I'm pretty sure 1500 tourneys would be useless for testing ICM empirically. In order to see what the equities are in a particular configuration of stacks you would have to see something very close to that configuration a very large number of times. I'm just pulling this out from down under, but I'd be surprised if you didn't need at least something on the order of 100,000 tourneys to do this and then you'd better have a lot of computing power and a lot of time. [/ QUOTE ] i don't know much about the stats involved, but the econ professor who was originally going to work with me said we could get some preliminary results at around 500. [/ QUOTE ] Out of curiosity, how are you going to test it?
|
|
#47
11-01-2005, 02:23 PM
|
|||
|
|||
|
Re: Challenging the basics behind ICM
[ QUOTE ]
[ QUOTE ] [ QUOTE ] I'm pretty sure 1500 tourneys would be useless for testing ICM empirically. In order to see what the equities are in a particular configuration of stacks you would have to see something very close to that configuration a very large number of times. I'm just pulling this out from down under, but I'd be surprised if you didn't need at least something on the order of 100,000 tourneys to do this and then you'd better have a lot of computing power and a lot of time. [/ QUOTE ] i don't know much about the stats involved, but the econ professor who was originally going to work with me said we could get some preliminary results at around 500. [/ QUOTE ] Out of curiosity, how are you going to test it? [/ QUOTE ] this was going to be the prof's contribution, until he became too busy [img]/images/graemlins/frown.gif[/img] i, unfortunately, don't have the graduate level econometrics background to do it myself.
|
|
#48
11-01-2005, 03:07 PM
|
|||
|
|||
|
Re: Challenging the basics behind ICM
wahooriver,
I'm not trying to be an ass. I'm definitely not a fan of discouraging people trying to post questions and become better players. However, you really do not understand ICM or how to apply it to decision making in SNGs. It is painfully obvious from your posts in this thread. Look in the FAQ for the ICM Calculator. There are links about its history and how to do calculations based on it. That could help provide some idea. Run a search on 2+2 (I know it's not the greatest search function in the world, but if you work at it you'll usually find something) for some old threads explaining the underlying idea of ICM and why ICM itself doesn't tell us how to make decisions, but rather we look at how each decision changes our ICM expectation and find the way to maximize that. This often involves putting our opponents on hand ranges, which accounts for [ QUOTE ] Has the villian been stealing the blinds at every opportunity? How loose are the other players? Who just had a bad beat? How do the other players assess my play (am I considered tight and solid or loosey goosey)? [/ QUOTE ] The actual Independent Chip Model does not account for this, but making decisions based on it does. I hope you take a longer look to understand what all of these posters are saying. Best of luck. -SonnyJay
|
|
#49
11-01-2005, 03:43 PM
|
|||
|
|||
|
Re: Challenging the basics behind ICM
SonnyJay,
I have ignored all the insults and immature remarks - this is an open forum and I assume that we each have a soapbox (like in Hyde Park). I do believe that I understand how ICM works. I respectfully disagree that it represents a major breakthrough in understanding for me! I suspect that some users respond best to this type of simulation process (for in fact a model is no more than a mathematical simulation). Many pros would run over most of us in SNG play, even if we are consistent winners. They are probably not using a simulation model, but rather understanding pot odds, stack sizes and situations. I mentioned the book Blink (no I am not a bookseller) because Gladwell presents situations where mathematical modeling fails and situations in which it excels. My premise (which I probably presented poorly) remains that relying solely on a mathematical model <font color="red"> without </font> understanding your opponents will lead to suboptimal results. Certainly many supporters of ICM do incorporate reads in their analyses. The other potential problem of a mathematical model is the everchanging conditions. Players change their tendencies based on stack sizes, position, recent hands. I suspect that these factors are indeed too complex to model. ICM probably helps some devotees understand stack sizes and dynamics. Certainly others learn these factors more intuitively. It probably depends on how one tries to analyze poker. Despite my love of mathematics, I enjoy understanding situational plays. I will throw in the occasional bluff, because the situation makes it likely to work. I understand that the bluff need only work a reasonable percentage to be worthwhile (that is a math proof). So I have no problem with ICM as a learning tool, but still submit that one can become an excellent player using other learning strategies. As I read these posts, the most successful and ardent supporters of ICM do use it as a tool but not the only tool. The most successful work to understand the other players and the current situation. I worry that we must make that point more clearly. Playing SNG poker requires a broad knowledge of styles, situations and probabilities. We cannot overemphasize one factor over the others.
|
|
#50
11-01-2005, 03:54 PM
|
|||
|
|||
|
Re: Challenging the basics behind ICM
u don't understand ICM
ICM doesn't have anything to do with opponents or reads or cards or anything All ICM does is tell u the equity that a chipstack has in the tournament prize pool based on the stack sizes of everyone at the table that's it we then use calling ranges, reads, probability, etc to determine the possible chip stack outcomes if we push/call vs if we fold...we then use ICM to see what our equity would be in those various outcomes...if the equity average (Weighted by probability of each outcome occuring) is higher when we push/call than the equity of us folding, then we say it's +$ev to push/call nobody relies solely on ICM because all ICM does is tell u the equity of a chip stack the better u read the players and determine what they will do, the better you give accurate probabilities of the outcome, so the more often u will make the correct move nobody emphasizes ICM over anything else, as it is merely an equity calculator...
|
 |
|
|
Linear Mode
