ICM/SNGPT rambling thoughts(long)

[Nicholasp27]

i've seen before using the tools that if the range changes from tight to maniac, the ev can swing >1%...

so if u make a +.1ev move that will result in the opp changing from tight to maniac, then that affected your future ev by 1%, making that +.1ev move not so hot

[BadMongo]

That's fine. You made up an example where the EV of your push depends heavily on folding equity. I'm not saying that push a lot can NEVER decrease your EV on future hands, of course it can. What I'm saying is that it can also INCREASE it.

Consider this example which you seem to have overlooked before:

You have pushed 3 times in a row into the BB of the guy on your left, and you know that his range has loosened substantially. You push a fourth time, holding AA.

Is your EV higher or lower than it would have been had you not pushed those 3 previous hands?

The answer is higher. You want a call when your holding aces, and because of your previous pushes you are now more likely to get it. So passing on those small EV spots not only cost you the EV you would have got directly from those hands, but also the increased EV you now expect to get from this hand.

Therefore, you blanket statement that:

[ QUOTE ]
if your action in hand X causes the opps' ranges to loosen in hand Y, then your ev will be less in hand Y than if you hadn't taken that action in hand X

[/ QUOTE ]

Is clearly false in this case. Since you cannot predict whether pushing now will help or hurt your future EV, I don't agree with your position that those small +EV hands should be folded if the ICM assumptions have all been satisfied.

[Nicholasp27]

your argument is that we can't predict if it will make it higher or lower so just take all of the + and don't worry about it...we may get aces and it helps then...

well we can't predict most things in poker, but we still work to do as best as we can...if i know that pushing now (blinds 75/150) will most likely prevent me from being able to profitably be able to push in 2 hands (blinds 100/200) then i may want to fold this .1ev hand...u have to use judgement

and u are much more likely to get a hand that does worse against the wider range than does better against the wider range...and that's the point of all of this...what's likely to result in more money

again, i don't think that u should be passing up +.8ev situations because u don't wanna look loose...but +.1ev? sure, at times

[AliasMrJones]

[ QUOTE ]
so again, which assumption is incorrect?

[/ QUOTE ]

Well, at this point in your ramblings, mainly that a push now will have a profound affect on your future +$EV opportunities. I think on average the effect will be marginal at best.

The only effect a push now could have is to widen opponents' calling ranges. I think we can agree that tighter opponents = more FE = more +$EV push opportunities. So, looser opponents is bad.

The problem is, a SnG is very, very short. and the push/fold part of a SnG is even shorter. I just don't think there is enough time for there to be much of an effect usually. Also, one push probably has 0 effect. It is combinations of pushes that might have an effect. Again, with the short time factor, the probabilities of having the right combination of pushes to have a significant impact is small.

Based on my experience, I think we're definitely talking about affecting less than one decision per tourney. If I had to guess, I'd probably say we might be talking about one decision per 4 tourneys on average?

(I'm a math major, so I apologize in advance for some of my rude language when proving this; it gets me by when doing assignments.)

1) SnGPT takes into account everything except for skill level and very extreme stack situations. I'll ignore extreme stack sizes because that's relatively rare and playing around it is fairly straightforward (fold A LOT when there's a super-short shorty and you're third stack with little FE).

Remember that what ICM really is is a method to calculate (very accurately) your chances of coming in a given place. Skill level tends to manifest itself in a coefficient before each of these chances. So, if your ICM percent chance for first is A, second is B, and third is C (where 1-(A+B+C)=chances of fourth or worse), your actual chances may be xA + yB + zC. If you're better than your average opponent, then ($first)xA + ($second)yB + ($third)zC > ($first)A + ($second)B + ($third)C. For many of us, this fact is clearly true at the beginning of the tournament because many of us lose less than the rake or even beat the rake and are therefore better than our opposition.

Now comes the important part. Suppose the blinds were raised to 10k/20k at the beginning of the tourny, i.e. every bet for the entire tourny would be all in or fold, and the BB and SB are forced all in no matter what. Do you see how this would seriously lower our x,y, and z? We're now guaranteed to be all in blind in the first 9 hands, and that's if we're lucky (20% of the time we're all in blind on the first hand). Of course, if we properly adapted to this style, we could still have an edge, but clearly not as much as before. This extreme example shows that as the blinds go up x, y, and z approach 1.

Now, imagine that in this extreme example, it's now HU. Obviously it's impossible to gain any edge here, and your EV is exactly what ICM predicts it to be. Thus, as it gets shorter handed, x, y, and z approach 1.

The point of all of this is to show that when it's push/fold time on the bubble, our skill level advantage is much smaller than it is earlier on. Therefore, we don't have nearly as much of an argument for passing up marginal EV edges (Again, as Alias keeps trying to say, EV and cEV are different. This is incredibly important to understand, and is the entire point of ICM.) because our chips are no longer worth significantly more than our opponents' chips. If I had a large enough sample size, I'd demonstrate this empirically to anyone who's skeptical.

2)

The argument that our opponents' calling ranges will loosen after we push is sound. However, its effect is pretty insignificant.

Consider all the possible outcomes of a push (I'm gonna ignore three-way pots and things):

a) You get called by a bigger stack and lose. Clearly, widened opponents' calling ranges don't mean anything to you after this.
b) You get called by a bigger stack and win. Here, widened calling ranges do matter. (The argument that they couldn't matter nearly as much as doubling up holds no weight because doubling up was factored into our EV calculated by SnGPT the first time, and therefore should not be factored in again.)
c) We get called by a shorty and win. They do affect us here.
d) We get called by a shorty and lose. Here, it affects us.
e) We win the blinds. They affect us here.

Now, further consider the situations in which widened calling ranges actually affect us.

Case b is a rare circumstance because of the gap concept. Big stack's calling calling range should be significantly tighter than our pushing range, so he shouldn't be calling often, and when we push with marginal hands (we're not debating non-marginal pushes, of course), we don't figure to win more than 30% of the few times we get called. Similarly, if we win this, we obviously have a bigger stack than before, and widened calling ranges hurt larger stacks less than smaller stacks because larger stacks have less chance of being busted out and need to push less frequently.

Case c involves the same two factors as case b (it's rare, and we have a larger stack) with the very important additional factor that you're now ITM. ITM, it's much less dangerous to get called, so the widened calling ranges have a much smaller effect.

Case d is essentially the worst case scenario in terms of getting hurt by widened calling ranges. However, it's rather rare because, again, shorty should be very tight calling us (more so than a big stack). Also, very few pushes into a short stack are marginal, so this should only affect you when considering to make a very marginal push into a short stack. Most of these marginal pushes are marginal because shorty's stack size is pretty big compared to our own (>50% or so). In these cases, the FE lost from a shrunken stack size is much more significant than that do to widened calling ranges.

Case e is insignificant because our lost EV from widened calling ranges can never be larger than the amount we gained from stealing the blinds. This is true because if the push did lose more than that amount in EV, it would become a fold (which obviously can't cost us more than the blinds).

I apologize for the very long reply, and feel free to disagree, but please take this into account before furthering making your arguments. Also, please please please refrain from the idea that EV isn't always a good thing; it always is.

[AliasMrJones]

[ QUOTE ]
if i know that pushing now (blinds 75/150) will most likely prevent me from being able to profitably be able to push in 2 hands (blinds 100/200) then i may want to fold this .1ev hand...u have to use judgement

again, i don't think that u should be passing up +.8ev situations because u don't wanna look loose...but +.1ev? sure, at times

[/ QUOTE ]

Well, eastbay's program comes preset with a threshold of .5, so you wouldn't be pushing a +.1ev if you use the defaults anyway. You say you shouldn't pass up +.8 because you don't want to look loose. What about +.5? If you shouldn't pass up +.5 because you don't want to look loose it looks like eastbay already put in place enough fudge factor to overcome the table image factor.

[Nicholasp27]

ok, so now we agree that it does matter...so we agree that ICM neglects change in table image in it's calculations...

i'm not saying it has a 'profound effect'...i'm simply stating that it can widen your opp's ranges, which can have an effect ranging from .1ev to 1.1ev+...so i'm saying that if u know this push will have an effect on their range, it may be best to fold this hand

especially if blinds change next hand, u'd rather push the next one with tighter opp ranges than push it with a slightly wider range that comes from pushing 2+ times in a row

and i think the bubble is plenty long enough...many times it can last for 4+ levels!...all it takes tho is 3 hands...if sngpt tells u to push hand 1 and hand 2, then hand 3 will have a diff range...so if hand 2 is +.1ev, then maybe u should fold...

[Nicholasp27]

i've said i think .5 is a good guidepost

but there are times when .5 should be folded...remember, when it calculates u being called, it just calculates your static icm after the results based on probabilities...it doesnt take into consideration that being called with 27o is much worse than being called by kk in terms of future ev and table image

[ZeroPointMachine]

[ QUOTE ]


The argument that our opponents' calling ranges will loosen after we push is sound. However, its effect is pretty insignificant.

Consider all the possible outcomes of a push (I'm gonna ignore three-way pots and things):

a) You get called by a bigger stack and lose. Clearly, widened opponents' calling ranges don't mean anything to you after this.
b) You get called by a bigger stack and win. Here, widened calling ranges do matter. (The argument that they couldn't matter nearly as much as doubling up holds no weight because doubling up was factored into our EV calculated by SnGPT the first time, and therefore should not be factored in again.)
c) We get called by a shorty and win. They do affect us here.
d) We get called by a shorty and lose. Here, it affects us.
e) We win the blinds. They affect us here.

Now, further consider the situations in which widened calling ranges actually affect us.

Case b is a rare circumstance because of the gap concept. Big stack's calling calling range should be significantly tighter than our pushing range, so he shouldn't be calling often, and when we push with marginal hands (we're not debating non-marginal pushes, of course), we don't figure to win more than 30% of the few times we get called. Similarly, if we win this, we obviously have a bigger stack than before, and widened calling ranges hurt larger stacks less than smaller stacks because larger stacks have less chance of being busted out and need to push less frequently.

Case c involves the same two factors as case b (it's rare, and we have a larger stack) with the very important additional factor that you're now ITM. ITM, it's much less dangerous to get called, so the widened calling ranges have a much smaller effect.

Case d is essentially the worst case scenario in terms of getting hurt by widened calling ranges. However, it's rather rare because, again, shorty should be very tight calling us (more so than a big stack). Also, very few pushes into a short stack are marginal, so this should only affect you when considering to make a very marginal push into a short stack. Most of these marginal pushes are marginal because shorty's stack size is pretty big compared to our own (>50% or so). In these cases, the FE lost from a shrunken stack size is much more significant than that do to widened calling ranges.

Case e is insignificant because our lost EV from widened calling ranges can never be larger than the amount we gained from stealing the blinds. This is true because if the push did lose more than that amount in EV, it would become a fold (which obviously can't cost us more than the blinds).

I apologize for the very long reply, and feel free to disagree, but please take this into account before furthering making your arguments. Also, please please please refrain from the idea that EV isn't always a good thing; it always is.

[/ QUOTE ]

Good analysis. But I think it focuses on bubble hands only. The assumption in c. that you are ITM if you win in particular. In my experience the days of cruising to four handed with 50/100 blinds are pretty much over. ICM plays consistently begins with 5/6/7 players at the table and quite a few hands between you and the money. Can controlling your opponents calling ranges in the early part of ICM play significantly alter your expectation for the tourney? I don't know. But, I think it is worth examining.

I appreciate your input on this discussion. I think I mentioned earlier that my OP was way off track. This current discussion is where I was headed. I have been making some adjustments in my game along these lines and it seemed to work. Obviously, "seemed to work" is not a terribly compelling argument. But it started me thinking and led me down this crooked road. My initial theories on why this might work were more or less garbage. But the resulting discussion was of benefit to me and hopefully others. This current discussion may be equally as valid. Still not sure. When my thinking is fuzzy on something it haunts me until I can resolve it. (Wish I'd never read the words quantum mechanics)

[AliasMrJones]

[ QUOTE ]
ok, so now we agree that it does matter...so we agree that ICM neglects change in table image in it's calculations...

i'm not saying it has a 'profound effect'...i'm simply stating that it can widen your opp's ranges, which can have an effect ranging from .1ev to 1.1ev+...so i'm saying that if u know this push will have an effect on their range, it may be best to fold this hand

especially if blinds change next hand, u'd rather push the next one with tighter opp ranges than push it with a slightly wider range that comes from pushing 2+ times in a row

and i think the bubble is plenty long enough...many times it can last for 4+ levels!...all it takes tho is 3 hands...if sngpt tells u to push hand 1 and hand 2, then hand 3 will have a diff range...so if hand 2 is +.1ev, then maybe u should fold...

[/ QUOTE ]

A typical SnG is about 60 hands. You're saying the bubble can last 40 of those? I guess it's possible. It's also possible to get dealt AA 4 times in a row. But, how important it is is dependent on how often it will happen.

I don't think the bubble will be long enough often enough and I don't think if you're using correct ICM push/fold play you'll be pushing enough often enough for it to have much of a real effect on your longterm ROI. Some effect? Probably. Enough that it is worth investigating? Sure. Enough that most players should be frequently deviating from ICM? I really don't think so. Like I said before, I think for most people, it is perhaps one decision per 4 tournaments of difference and I think there is a pretty good chance of actually hurting your ROI rather than helping by deviating from ICM suggested push/fold.