ICM/SNGPT rambling thoughts(long)

[AliasMrJones]

[ QUOTE ]
Raptor is right that playing a certain hand right culminates in getting right the calling ranges of your opponents. But he is only almost there. You have to factor in what your decision will mean to the future calling ranges and how this will affect your EV on your future weighted average decisions. If opponents loosen up, your average EV will very often go down which means that if you stay with the threshhold you will have less situations where you can push for profit.

[/ QUOTE ]

Excellent, excellent summation and something that is very difficult to model, but would be an excellent addition to the current state-of-the-art ICM model (which includes more than just ICM, but all the extra stuff that eastbay has packed into SnGPT) if we could come up with something reasonable.

[Nicholasp27]

it's not just a threshold based on skill level tho...u can be a +5% on a total donk, but u shouldn't wait for a +5% edge, as you will get blinded out before that likely comes along (or it comes along when u have 1bb and so u just go up to 2bb if u win the coinflip)

the +.5% is set to help guard u from pushing too often (loosening up villian's calling ranges too much and too unpredictably) and to compensate for skill edge without causing u to fold yourself out of the tournament

based on play (limited sample size) it seems that +.5% is a good number as long as you exercise caution and don't take it as the gospel due to uncertainty in villian's calling ranges, how your actions affect future actions, etc

if it was higher, then my experience tells me that you'd get blinded out of the tourney too much...or at the very least, your positioning going into itm would lead you to lots of 3rds and not so many 1sts

but if u push every .1/.2 edge u will ruin your FE and their ranges will open to a point where u may not get another +ev hand until you are very low in chips

so .5 is a good marker imo, although in some games you'd want to only push >.9 and some you wanna push >0

[AliasMrJones]

[ QUOTE ]
If you had an +.5%EV push but knew your next hand would be a +1.5%EV opportunity you would obvioulsy not take the first one due to the when you are broke you are done principle.

[/ QUOTE ]

The fallacy here is, "if you knew..." It's kind of like saying if you knew your aces would get cracked this hand, you shouldn't go all-in. The fact is, you don't know. ICM attempts to model this unknown. The fact that the ICM model says it is 1.5% +$EV here means that the model says it is worth more real money to take this opportunity here than to hopefully get a better one later.

[ QUOTE ]
Obviously other factors are important before pushing on the bubble other than the mere existance of a +EV situation.

[/ QUOTE ]

On the bubble? I don't think so. This is the place where SnG prize pool money is won and lost. This is THE time push +$EV. What other factors are important?

[The once and future king]

[ QUOTE ]
That is the entire basis for the ICM model. What is my prize money equity based on the current stack sizes. You might get a more +$EV situation later...you might not...you might get a more +$EV situation later and lose. ICM models these various chances and comes up with prize money equity.


[/ QUOTE ]

This is totaly incorrect. ICM only calculates what your share of the prize pool will be after any specific push, it dosnt predict or factor in, in any way EV situations that may arise later. The Independent Chip Model is just a way to devise $EV from a given stack size.

[Nicholasp27]

we're still attacking this from different angles

imo, you want the most +ev plays possible throughout the bubble...

if u push 6 +.1/.2 ev spot, you may NEVER get a >+.5 ev push the entire bubble, so your average ev would not be as good as pushing the 3 >+.5 hands you got

i realize that icm takes into account you getting called and winning/losing/tying...but it doesn't take into account how what you do affects the next hand...sure u can change the ranges, but that's after the fact...each hand is NOT in a vacuum


take a random hand that sngpt says is +.7 with a tight opp...now change that opp to maniac...depending on the situation, that may change the hand from +.7 to -.3 ...that's a whole $1!...was it worth pushing for 10 cents the previous 2 hands if u have to pass up that $1?

[BadMongo]

Ok, your arguement is all over the place. I'm not really sure exactly what your are trying to say.

[ QUOTE ]
[ QUOTE ]
The more +real-prize-pool-dollar-EV situations you take advantage of, the more real-prize-pool-money you will win in the long term. Period.

[/ QUOTE ]

exactly

if u pass up a +$0.20 push and then get a +$0.50 push and take that, then you followed your last statement and made more money...

[/ QUOTE ]

No, you haven't. You only made +0.50 when you could have made +0.70. You still lose that +0.20 that you passed on.

[ QUOTE ]
u can't just push every +.1% ev situation with impunity unless you can be >99.9% accurate in your assessment of villian's exact calling range based on previous hands

[/ QUOTE ]

First you seem to be saying that you should pass on some +EV situations even if all ICM assumptions have been satisfied. That is clearly false, for reasons that have already been explained. Now you're changing your argument to say that you shouldn't push small edges because you can never be 100% confident of an opponent's calling range. This argument has more merit, but it's not a flaw with ICM, it's a flaw with the input data. If we calculate an EV with ICM based on a hand range that isn't applicable, that's our fault, not the model's.

[AliasMrJones]

[ QUOTE ]
[ QUOTE ]
That is the entire basis for the ICM model. What is my prize money equity based on the current stack sizes. You might get a more +$EV situation later...you might not...you might get a more +$EV situation later and lose. ICM models these various chances and comes up with prize money equity.


[/ QUOTE ]

This is totaly incorrect. ICM only calculates what your share of the prize pool will be after any specific push, it dosnt predict or factor in, in any way EV situations that may arise later. The Independent Chip Model is just a way to devise $EV from a given stack size.

[/ QUOTE ]

EXACTLY! It is a way to estimate $EV from a SET of stack sizes. How can it possibly estimate your share of the prize money at the end of the tournament unless it takes into account future possibilities? What may be confusing is it assumes that everyone will have equal +$EV opportunities in the future. On a long-term (and short-term probability) basis, this is correct. So it does factor in future possibilities by assuming that they will be equally spread across all players. What it doesn't do at present is factor in your opponents' ability to take advantage of their oppotunities vs. your ability to take advantage of your opportunities.

[Nicholasp27]

the existence of a +EV spot is just that...a spot...it's one hand...it's in a Vacuum...therefore if u think a push will loosen your opps' ranges then that is a Factor that should be taken into consideration that ICM does not currently do...therefore u need to do that...


a bot running off ICM in a game where there are 4 players, 800 chips each with blinds starting at 50/100 would NOT be the highest roi player over time if the other 4 used more factors than just +/- ev that ICM does...even if it did accurately change opp's ranges each hand...

why is icm not the only factor?

because it does NOT take into consideration future events...


DEFINITION OF ICM: if all else is equal (skill level, table image, etc) then given a set of chip stacks, it tells you what each stack's expected value is...how much that stack can expect to earn in prize money over the long run

thus it is a GREAT tool...best tool for SNGs, but it's NOT the only factor in the bubble


edit: edited definition of ICM

[Slim Pickens]

I'm fairly sure now that the only reason SNGPT uses threshold values (both positive and negative by the way if you keep track of all the signs and ask it "Should I push?") for decisions is that there is some amount of uncertainty created by your opponents' pushing/calling ranges and the limitation of the ICM wrt blinds/position and such. I was wrong to compare it Sklansky's argument about "taking larger edges later."

SlimP

[Nicholasp27]

no u don't make .70...after you pushed the +.2, the next hand's value is +.1 or even -.4...but if u fold the +.2 then u get +.5

to clarify my argument, even if u input everything perfectly into sngpt, you still should NOT push every +.1% ev hand...as sngpt cannot tell you that taking that +.1% push will change your next hand's ev by .6

i do have a side argument that you also shouldn't use .1% because you are not accurate enough with your inputs...thus that's why i like the .5% threshold as a guidepost instead of .1%...to help make up for your error