ICM Quantified - First Set of Results

[eastbay]

I'm a little confused by this. Where are the other stack sizes, other than your own?

Also, has been discussed several times previously, you can't use all hands from a tournament without introducing a bias into the data. I am sympathetic to your desire to do it anyway, because you have massive sample size problems if you don't. But I think to do this right, you have to remove that bias from the data.

eastbay

[rvg72]

[ QUOTE ]
[ QUOTE ]
The more I think about this, the more I think a skill measure is critical. I could see an argument that the whole difference in the short and big stacks from ICM prediction is attributable to bad players more often having short stacks and good players more often having big stacks. (Also, I would not be surprised to see good players outperform ICM with short stacks and bad players not live up to ICM with big stacks.)

I think ROI is just about the only skill measure you can use. Yes, it will mean maintaining a large database to track ROI, but I think it would be worth it in the long run. (And, might turn out to be useful for other purposes.)

[/ QUOTE ]

I have to agree with everything here. Performance relative to ICM (which assumes no skill advantage) is a strong function of the same thing that allows a player to acquire a big stack: skill. I also agree this is cool as [censored].

[/ QUOTE ]

This could explain some of this or maybe all of this - I'm heavily leaning towards "some" at this point but point well taken.

I'm going to add some basic "skill" data to each data point to give us some way to measure this. I'm going to include Saw Flop %, Pre Flop Raise % and Post Flop Aggressiveness although it will only do it for this tourney. You rarely have good players seeing more than 15-20% of flops so you could use that as a basic way to differentiate or combine different factors to try to draw some conclusions.

You could also use this data to see things like breakdowns of Saw Flop / PFR / Agg for 1st/2nd/3rd and 4th spots in a tourney for example.

rvg

[rvg72]

[ QUOTE ]
I'm a little confused by this. Where are the other stack sizes, other than your own?

Also, has been discussed several times previously, you can't use all hands from a tournament without introducing a bias into the data. I am sympathetic to your desire to do it anyway, because you have massive sample size problems if you don't. But I think to do this right, you have to remove that bias from the data.

eastbay

[/ QUOTE ]

The other stack sizes come from each participant in the tourney.

I understand the bias since the results from each tourney are linked however I think there would be a much greater bias by just using your own data and calculating results based on that vs looking at an SNG and analyzing ICM from each, "random" players perspective.

What do you think? I know it isn't ideal but I don't think it could be done differently. Of course sample size needs to be bigger - hopefully these early findings and the discussions on this thread help fulfill that.

rvg

[mgross]

I think the best way to get unbiased results is to create a simulation where players start at a set chip stack at a push/fold level of blinds. They play according to ICM and then take those simulated results and we will see if the actual fits to the ICM model.

[rvg72]

[ QUOTE ]
I think the best way to get unbiased results is to create a simulation where players start at a set chip stack at a push/fold level of blinds. They play according to ICM and then take those simulated results and we will see if the actual fits to the ICM model.

[/ QUOTE ]

That is a very interesting idea...

rvg

[mgross]

The question is, do we include blinds going up? How does that affect everything? Position relative to blinds when they go up, and how to adjust right before they go up. Can we fit that into the model?

[zabt]

[ QUOTE ]
I'm going to add some basic "skill" data to each data point to give us some way to measure this. I'm going to include Saw Flop %, Pre Flop Raise % and Post Flop Aggressiveness although it will only do it for this tourney. You rarely have good players seeing more than 15-20% of flops so you could use that as a basic way to differentiate or combine different factors to try to draw some conclusions.

[/ QUOTE ]
I don't want to sound overly negative -- I think this is outstanding work that could turn out to be very valuable. But, I don't think these "skill" measures will be useful. You rarely have good players seeing more than 15%-20% of flops, but that doesn't mean every good player falls outside that range, especially not during a single SNG. A good (or bad) player may have a run of good cards in a single SNG which causes them to see more flops. Any skill measure which is limited to the individual SNG in question is going to be of questionable value.

[zabt]

[ QUOTE ]
I think the best way to get unbiased results is to create a simulation where players start at a set chip stack at a push/fold level of blinds. They play according to ICM and then take those simulated results and we will see if the actual fits to the ICM model.

[/ QUOTE ]
Use ICM to see if ICM measures equity correctly? I think there may be some problems with this.

[mgross]

Use ICM-style play to judge if ICM models correctly is not bad. Not everyone in the results from this thread play perfect or anything. We are just making play consistent by using a style. We will then see how the results play out.

[pr0crast]

[ QUOTE ]
I think the best way to get unbiased results is to create a simulation where players start at a set chip stack at a push/fold level of blinds. They play according to ICM and then take those simulated results and we will see if the actual fits to the ICM model.

[/ QUOTE ]
I completely agree with this. A simulation would provide purer data than simply amassing huge samples.

Great work rvg, I look forward to seeing where this goes.