ROI and ICM

[microbet]

Has anyone tried any ICM/chip valuation calcs that take either ROI or finishing distributions into account?

Also, there is very little analysis here that takes $/hr into account. Do any of you make decisions in SNG's giving any consideration to $/hr? Or is it all about pride (or possibly fun)?

[quinn]

The chances of an average player winning a tournament at any given point seem to be roughly their number of chips divided by the total number of chips in the tournament..
plus a somewhat linear factor of their percent deviation from the median chip stack, where average is zero..

So, if your stack is 4000 and there are 8000 chips in the tournament left, with the median stack at 1500, your stack is 2.6 times the median stack.
So your chances are .5 + 2.6*x, with x = approx .04 (x is constant)
(that would put your chances at about 60%)



I value $/hr very highly.

That's why I play up to 8 tables at a time. My ROI suffers, but ROI doesn't buy stuff.

[spentrent]

[ QUOTE ]
That's why I play up to 8 tables at a time. My ROI suffers, but ROI doesn't buy stuff.

[/ QUOTE ]

Can I get a tattoo of that?

[microbet]

[ QUOTE ]
The chances of an average player winning a tournament at any given point seem to be roughly their number of chips divided by the total number of chips in the tournament..
plus a somewhat linear factor of their percent deviation from the median chip stack, where average is zero..

[/ QUOTE ]

Sure, roughly, and I'm partly responding as a bump to the thread, but I'm not just talking about chances of winning. I mean total equity in the tournament for any given chip distribution and whether anyone has tried to alter the ICM calculations with the subject's ROI or finishing distribution. Particularly I'm interested if the people who are programming ICM tools (eastbay, ChrisV, dethgrind ...) have thought about doing this. I have been playing around with it a little, but haven't really come up with anything that good yet.

[ QUOTE ]
I value $/hr very highly.

That's why I play up to 8 tables at a time. My ROI suffers, but ROI doesn't buy stuff.

[/ QUOTE ]

Do you ever make decisions differently in a particular tournament because of this? Like take a little bit extra of a chance early?

[eastbay]

[ QUOTE ]
[ QUOTE ]
The chances of an average player winning a tournament at any given point seem to be roughly their number of chips divided by the total number of chips in the tournament..
plus a somewhat linear factor of their percent deviation from the median chip stack, where average is zero..

[/ QUOTE ]

Sure, roughly, and I'm partly responding as a bump to the thread, but I'm not just talking about chances of winning. I mean total equity in the tournament for any given chip distribution and whether anyone has tried to alter the ICM calculations with the subject's ROI or finishing distribution. Particularly I'm interested if the people who are programming ICM tools (eastbay, ChrisV, dethgrind ...) have thought about doing this. I have been playing around with it a little, but haven't really come up with anything that good yet.


[/ QUOTE ]

I've thought about it quite a bit. I don't have a real compelling solution yet. You may remember that I started a thread about looking at the issue empirically. I haven't returned to that since I've been hard at work trying to get my program ready for release.

As soon as things settle down a bit, I intend to go back and complete that study, although I worry that even with many thousands of hands, it isn't enough to draw any real hard conclusions. We shall see.

eastbay

[stupidsucker]

Blinds are what change the outcome by a large error though.

If you take 4 stacks
3500
2700
1300
2000

Now take out skill and take out blinds and you have a rough idea already of ITM(you can break it all the way down to 1,2,3 as well).

The simplest way to simulate no blinds and no skill is simple a 4 way all in fest every time.

the three highest stacks all have an ITM win% above 75%, and the lowest has an ITM win% of less then 25%...You can figure it out from there..

The biggest problem is position and blinds. What good is a 300 chip lead over the small stack if the blinds are 200/400 and you are next? Throw in the complication of not knowing what someones push/call/fold standards are and you have a big equation.

In the end, The equation would have a lot of variables to be perfect. I am not sure what things you can take out to slim the equation, but maintain its integrity.

Here are some

1)Stack sizes
In relation to the other stacks
and
In relation to the blinds
2)# of opponents
3)Position(and where the action is at the time of decision)
4)average idea of range of hands others would push with
5)average idea of range of hands others would call with
6)Blind sizes
7)How many hands left before blind increase
8)What is your position in relation to the blind increase


I wish I knew how to put my simpleton explaination into an equation, but I dont. I dont even know if I spelled simpleton correctly. But I do know the idea is sound when only using stack sizes/#of opponents the variables. If I am wrong let me know cause it fascinates me.

[microbet]

[ QUOTE ]
The equation would have a lot of variables to be perfect.

[/ QUOTE ]

When doing an ICM calc, you are usually doing some estimation like, 'what range of hands does he have?', or 'how likely is he to fold?' and these will never be perfect. And, especially late in the game, you will always have complex considerations based on player reads and what will happen in future hands.

So, there will always be innaccuracy, and I think it is possibly fair to say ICM is not helpful and one is better using an approach based on personal and shared experience.

But, if you are going to use ICM as a tool, I think it would be more accurate than it is if you just make a very simple adjustment where you base your equity at buyin on your expected return and have a uniform adjustment until the equity of all the chips equals the first place payout.

Does that make sense? I just mean keep lowering your skill-added equity as you go on so that you never have more than the maximum possible equity. Whether 'as you go on' is determined by number of players left, chip stacks, blinds, or some combo, I don't know.

[eastbay]

[ QUOTE ]
[ QUOTE ]
The equation would have a lot of variables to be perfect.

[/ QUOTE ]

When doing an ICM calc, you are usually doing some estimation like, 'what range of hands does he have?', or 'how likely is he to fold?' and these will never be perfect. And, especially late in the game, you will always have complex considerations based on player reads and what will happen in future hands.

So, there will always be innaccuracy, and I think it is possibly fair to say ICM is not helpful and one is better using an approach based on personal and shared experience.


[/ QUOTE ]

I think you might be very surprised how often stock ICM leads you very close to what the most experienced and successful players will say they do by feel.

I know I was.

Let's make that more concrete: Try to construct a push/fold situation where everybody has at least a few blinds in their stack that you think ICM would mislead you about.

eastbay

[microbet]

[ QUOTE ]
Let's make that more concrete: Try to construct a push/fold situation where everybody has at least a few blinds in their stack that you think ICM would mislead you about.

eastbay


[/ QUOTE ]

Well, this is call/fold situation, and is artificial, but it is what I was thinking about.

First hand of a $11 SNG, you are in BB, everyone folds to SB who pushes and shows you AKo. You have 88.

You will win 55.5% of the time.

You call and win - $18.44 in equity
You fold - $9.83 in equity

These positions are equivalent if you win 53.5% of the time.

So, I'm not talking about anything like folding a 70% edge, but does a winning player pass up this 55.5% edge, eventhough it is +$EV. If so, then perhaps the ICM calcs can be modified. This does not take $/hr into account. If you take that into account I think it is a call, no matter how good the player is.

[Apathy]

[ QUOTE ]
[ QUOTE ]
Let's make that more concrete: Try to construct a push/fold situation where everybody has at least a few blinds in their stack that you think ICM would mislead you about.

eastbay


[/ QUOTE ]

Well, this is call/fold situation, and is artificial, but it is what I was thinking about.

First hand of a $11 SNG, you are in BB, everyone folds to SB who pushes and shows you AKo. You have 88.

You will win 55.5% of the time.

You call and win - $18.44 in equity
You fold - $9.83 in equity

These positions are equivalent if you win 53.5% of the time.

So, I'm not talking about anything like folding a 70% edge, but does a winning player pass up this 55.5% edge, eventhough it is +$EV. If so, then perhaps the ICM calcs can be modified. This does not take $/hr into account. If you take that into account I think it is a call, no matter how good the player is.

[/ QUOTE ]

Well sure it is, since you can open up a new table right away, but even then (when only considering hourly rate) its close in mt mind.

Also the swings you will experince by playing this way will be a lot greater.