The effectiveness of SNG-W "future games" simulator

[darom03]

I am writing this in the hopes of a small discussion about SNG-W new "future games" simulator.

The simulator supposedly takes into account future actions in the coming rounds. That way it should be a counter for the shortcomings of ICM.

However, my experiments with SNG-W shows otherwise.

Here is an example of what I think is bad advise from SNG-W with the simulator turned on.


If I were to turn off the simulator it would be a clear 0.42 push.

Which of the two advices would you rely on in this situation?

Another example, where the opposite is the case:


If we were to turn the simulator off it would be a -0.43 fold.

Which of the two would you trust?

EDIT: Understand me correctly: I think the simulator is a good step in a right direction, so I would hate to discourage any development of the feature.

[RexWoo]

I think the simulator is in early version.

I think you'd better post directly to the developper to be helpfull.

[darom03]

[ QUOTE ]
I think the simulator is in early version.

I think you'd better post directly to the developper to be helpfull.

[/ QUOTE ]

Yes, but I would rather have a discussion among you guys. Maybe we could all learn something from it.

[BradleyT]

Let me see if I have this right -

In the first example you have 4042 chips, blinds are 100/200 and you have Q5o and pushing into the BB (370 chips) who is calling 100%
Simulator off = +.42%
Simulator on = -.2%

Correct?

[IFoldPktOnes]

The opponent models have a huge effect on the simulations (even for the opponents who have folded).

It seems you can change the folding players opponent models by setting them to call, changing it, then setting them back to fold. I think the default opponent model setting in options might even set your own player model for the simulations.

For example, I changed every player model to very loose for the 1st hand and it made it a +0.16 push.

[darom03]

[ QUOTE ]
Let me see if I have this right -

In the first example you have 4042 chips, blinds are 100/200 and you have Q5o and pushing into the BB (370 chips) who is calling 100%
Simulator off = +.42%
Simulator on = -.2%

Correct?

[/ QUOTE ]

Yes, that is correct.

[Ace_Ren]

Disclaimer 1: I'm not really familiar with that piece of software and I'm not 100% sure of what's going on, but I think I get it. In the first example before posting blinds, your stack is 4042, not 4142, right? Metro is at 570, not 370, right?
Disclaimer 2: ICM calcuations are not my strong point, so take my advice lightly.

In the first example, note that even if you do push, your equity goes up from 42.5% to 42.86%. Q5o is a 50.12% winner against a random hand (does Pokerstove include hand probability distribution in it's random setting?). The gain seems obvious. However, if you fold, your immediate equity will go down, but you will be able to get into a more profitable situation later.

As the dominant big stack, you can wait this hand out, then push on the next hand with any two. Nemo probably doesn't want to call with 2 shorter stacks around. So, you pick up 300 chips. Next is Blue's turn, you push again. Blue probably doesn't want to call with Metro being so low, he is looking to fold too. As a big stack, you can increase your own stack by attacking med-stacks with almost impunity as long as there is a short stack in play. If you bust the short stack now, you give away that opportunity.

In the second example, the simulator must be expecting correct play from Russ and Mark. Let's say you fold to Sheep. He picks up 200 more chips. Sheep probably folds in the SB (200 to Russ). It quickly gets around to your BB and someone is pushing you all-in (crushing your equity). If you fold in the BB, you're down to 500 chips(?). The SB would take you down to 300 chips if you don't push again (maybe someone else pushed first). Since you didn't push on this hand in question, Sheep may be able to last through another couple blinds. In the end, you get into a situation where you are the one who is all-in in the BB and the big stack only needs to foot part of the bill for all your chips.

I've seen that scenario play out hundreds of times where one small stack tries to wait out the other and it ends up being the non-aggressive one on the chopping block (actually, both are... the big stacks gain all the equity). The calculator seems to think that you should take your shot now (46.53% of winning) and hope. Even if you lose... you still have a little left.

In the second one, I personally agree with the simulator turned on. If it assumes correct play from Mark and Russ, you will be bleeding away equity with every hand. Imagine the ICM calculations in the next couple hands from their perspectives, who do you think that equity is coming from? The calculator says to take your shot now and catch up in stack size.

In the first one, interestingly, it seems like your equity goes up no matter what you do. It seems to conclude something that I've concluded for quite a long time: The big stack has an even greater equity than people realize because of his ability to push people around. You can afford to wait the first hand out because you'll be able to push later at an even higher gain.

(edit)However, in the first one, considering how close the numbers are. I would take the sure thing now and push rather than make a whole bunch of assumptions and gain only .02% equity(/edit)

[darom03]

[ QUOTE ]
However, if you fold, your immediate equity will go down, but you will be able to get into a more profitable situation later.

[/ QUOTE ]

Yes, but even if I loose the coinflip against the random hand in the BB, I have a dominant stack and will be able to push them around next hand and the next, just as if I had folded.

[ QUOTE ]
In the second one, I personally agree with the simulator turned on.

[/ QUOTE ]

I do too. I actually have trouble understanding the advice to fold with the simulator turned off, as it would leave us as the short stack.

[darom03]

[ QUOTE ]
The opponent models have a huge effect on the simulations (even for the opponents who have folded).

It seems you can change the folding players opponent models by setting them to call, changing it, then setting them back to fold.

[/ QUOTE ]

If correct, this is a big issue.

[SitNGo Wizard]

A couple of quick comments.

Make sure you are using release 1.0.1.65 (or later). I fixed a significant bug in FGS and the results of previous releases were wrong. That said, I still consider this a beta release.

If you have a medium to big stack, FGS will tend to increase your equity both when you push and when you fold. In the first hand your equity increases more when you fold than when you push. In the second hand your equity increases more when you push than when you fold. I have not verified that the results are correct, but I am not surprised by them.

As mentioned, FGS uses the current opponent models. So a “Loose” opponent will play looser in FGS. I plan to add the ability to override opponent models for FGS.

I do not claim that FGS solves all of ICM’s limitations, but I do think it is a step in the right direction. I welcome all comments and hope this discussion will inspire future improvements.