How much $EV will you give up?

[Moonsugar]

An argument with adanthar in another thread got me thinking. Instead of hijacking that thread even more than I already have I will start this new one.

Here is the situation:

Typical Party SnG payout, 6 people left, you are UTG with 1000 other stack sizes are: 1400 , 600, 1000, 1800, 2200 (bb). Blinds are 100/200. Blinds will increase to 150/300 in 9 hands. Opponents are about as good as you.

If pushing is -$EV (according to ICM) it could still be a worthwhile move so that you preserve your FE.

Questions: How much -$EV will you accept in a push here to preserve your FE? Can you quantify how much future, uncertain FE is worth? How?

Follow up: If the blinds were going to 150/300 next hand how much more $EV would you be willing to give up in a push here?

[golfcchs]

Whats FE?

[Moonsugar]

Fold equity. That is your equity in a pot by having enough chips to make your opponents fold their hands when you go all-in.

[Moonsugar]

Where are all you "push any 2 UTG" people telling me how to calculate how much my FE is worth?

[raptor517]

i dont like doing that. people realize that you can be pushing a LOT of hands there, and will loosen up their calling reqs. if yer gonna push any 2 in a spot like that, do it utg +1 at least, though i sometimes will just eat a blind. 700 DOES still have fe. holla

[adanthar]

First off, blinds increasing in 9 hands are different than blinds increasing in any other number. But no matter.

Second, if something is -$EV, by definition, you shouldn't do it. The point is that the ICM/eastbay's tool do not work properly in close situations and almost everything that they say is very close or even plain close is actually a push or a call.

This does not mean you should be pushing 32o, because that's not close. QJ is kinda close *according to the ICM* in the other direction and is actually totally automatic.

I suspect what's actually 'close' when you do take all of this into account is a hand like JT.

edit: or, you could put it that way in far less sentences vvv

[Bigwig]

Your analysis is flawed.

If preserving your fold equity is valuable enough to push in an ICM -$EV situation, then it is not a -$EV decision.

I'll expand if you need me too.

[Moonsugar]

Can you answer the questions I asked?

I don't think YOU can logically make a case why QJ push should be 'totally automatic'. I believe it is 'totally automatic' because you have been doing it and it is habit. But that doesn't make it correct. I want to know how you know 32 is so bad, QJ is automatic and JT is close. What factors go into your decision?

If the bb was t30 you would not think JT was a close all-in decision here. It is obvious that the reduction in future FE caused by the oncoming blinds is what is driving your decision. How does one know that a certain reduction in stack size is too much to accept? When you offer your opp. less than 4:1? Less than 2:1? For 10 steal opportunities? For 5?

A discussion on these factors would be very helpful.

[Moonsugar]

No, it isn't. I understand that. That is why I posted this whole thread.

Edit: Oh, I see the problem. Whenever I say -$EV, I guess I should have added (according to ICM) instead of just after the first use of the term. Sorry.

Can you answer my questions?

[Bigwig]

[ QUOTE ]
No, it isn't. I understand that. That is why I posted this whole thread.

Edit: Oh, I see the problem. Whenever I say -$EV, I guess I should have added (according to ICM) instead of just after the first use of the term. Sorry.

Can you answer my questions?

[/ QUOTE ]

But you're asking me about how much $EV I'd 'give up.' The answer is none. At least, not intentionally.

[adanthar]

You're missing the forest for the trees. Loss of FE is one of several factors but by no means the only one, and they are not even all quantifiable.

A weird example that should still intuitively make sense: It's an hour into an SNG with 300/600 blinds and, through an amazing coincidence involving eight way all ins preflop, the chip stacks look like this:

SB: 7000
You (BB): 2992
8 other people: 1 chip apiece

It's folded to the SB, who goes all in with two cards or possibly one card. You have aces. What do you do?

In this situation, I don't even have to look at the ICM to know it will say to fold, as will eastbay's tool. And yet, if you fold, three things will happen:

1)You will lose this 600 chips and the next 300 on the SB
2)The short stacks will sometimes bust and sometimes double up to 2, 4, or 8 chips while the blinds circle the table faster and faster and you keep folding every hand
3)You will eventually wind up all in with the luckiest short stack in a race for third while the big stack laughs at you and wins

So, of course you have to call. FE here is a non-factor except as to the big stack, and since he'll still have 4K chips left when he loses you won't be able to bust him before you hit HU. But guess what? In the meantime, you'll have done the same to him and taken away a lot more chips than just the ones you won with the aces.

Going back to the QJ hand, in other words, the reason QJ is not close is because of the winning percentage it has against a likely calling range. Against a certain opponent who folds Q9o but calls K8o, it's not so good because it will never be over 40-45% to win when you are called. Against almost everyone else, you either have way too much FE *or* are closer to even or favored in a flip that will give you even more FE when you do win it.

That's about as much as I'm ever going to write on this subject until I publish a book about it. If you don't get it from here, start by figuring out exactly why it is a bad idea to open push AK in level 1.

[FatalError]

i frequently decide to push with any 2 in this spot, or any 2 without a duece or trey

[Moonsugar]

No, I am asking how much HYPOTHETICAL $EV ACCORDING TO ICM you would give up. Are you saying that whenever ICM gives you -$EV result it is never worth pushing no matter what the blind situation is or how much FE I will lose?

I am trying to get some insight on ways to modify ICM for the blinds.

[Izverg04]

I doubt anyone will give you a credible answer. I'd say 0.01-0.015 of the prize pool is the typical range if you expect a large increase of the FE but I am just pulling that number out of you-know-what.

[Moonsugar]

[ QUOTE ]
Y

A weird example that should still intuitively make sense: It's an hour into an SNG with 300/600 blinds and, through an amazing coincidence involving eight way all ins preflop, the chip stacks look like this:

SB: 7000
You (BB): 2992
8 other people: 1 chip apiece

It's folded to the SB, who goes all in with two cards or possibly one card. You have aces. What do you do?

In this situation, I don't even have to look at the ICM to know it will say to fold,

[/ QUOTE ]

You should have looked. ICM says call. And to use your favorite pharase: IT IS EASY.

Can't wait for the book, it should be as enlightening as all your other drivel. LOL.

[Phil Van Sexton]

[ QUOTE ]
I am trying to get some insight on ways to modify ICM for the blinds.

[/ QUOTE ]

I was working on this a few weeks ago myself.

I wrote a program to compare ICM to actual results in the tournaments I have, and then grouping the results into categories so I might see where the ICM was most wrong.

One report that I did was to compare ICM to actual results for the various positions in 4-handed hands. This contains about 10,000 hands from 400 tournaments (I think). I computed these results for all 4-handed hands, for all 4 players (not just me).

****NOTE: These numbers are really worthless due to small sample size, but I figured that I would post them and let you draw your own conclusions. It may also motivate me to finish this ICM project, and maybe get me some volunteers to send me hand histories to help my sample.


According to this, the ICM overvalues your $EV by 0.007 when you are about to post the BB. This is 0.7% or $7 in a 100/9 tournament.

Even though your question was about UTG, the ICM numbers are really "what will my $EV be AFTER this hand if fold or win". Therefore, you should be looking at the BB number in these charts.

http://www.geocities.com/icmcalc/icm2.gif

Broken down by blind level. I believe this shows the effect of my small sample size.

http://www.geocities.com/icmcalc/icm1.gif

[adanthar]

Being sleepy this morning, I managed to botch the ICM calc three times, and on the fourth time around, I admit that the ICM does say to call aces. Of course, it also says to fold kings in a not particularly close fashion (61.17 vs. 59.32 in a $100 SNG). So, if your strategy in this spot is to hope to get exactly aces, the ICM works great!

Talk about forest for the trees again...yeah, time for the ignore list.

[microbet]

When you do an ICM calc against a range of calling hands, you pick what you think is the most likely range and go with that.

I think in some cases you would get a vastly different answer if you had the flexibility to put the opponent on a weighted range of calling ranges.

I think this is relevant to the idea in your post that you will sometimes have more than enough FE because the range is too tight and sometimes be +$EV because you are ahead of your opponents range.

[Moonsugar]

Thanks. I thought I was going crazy and no one would ever understand what I was trying to ask. This info is helpful. Will send my data to you if you want.

I guess I suck at english.

[Moonsugar]

Ignore me. Thats fine.

Go ahead and keep trying to win @ 2+2 message boarding. I concede that anyways.

I will try to better my poker.

[Moonsugar]

Did you break down the effect by stack size ranges? It should be larger for smaller stacks. Would be curious to see ranges of 3x, 4x, 5x, 6x, 7x, 8x.

[Phil Van Sexton]

[ QUOTE ]
Did you break down the effect by stack size ranges? It should be larger for smaller stacks. Would be curious to see ranges of 3x, 4x, 5x, 6x, 7x, 8x.

[/ QUOTE ]

I did break it down, but again the sample size problem is magnified when the data is broken down.

You can see the 2 "big stack" categories are out of place, but the 3 "small stack" categories behave as you describe.

http://www.geocities.com/icmcalc/icm3.gif

[Pokerscott]

[ QUOTE ]

Typical Party SnG payout, 6 people left, you are UTG with 1000 other stack sizes are: 1400 , 600, 1000, 1800, 2200 (bb). Blinds are 100/200. Blinds will increase to 150/300 in 9 hands. Opponents are about as good as you.



[/ QUOTE ]

The bold is why I don't think you should give up much EV to preserve fold equity.

If everyone has FE then no one is advantaged because everyone is at the same skill level.

The key question becomes how much of a disadvantage does it become when you don't have FE and everyone else does (i.e. you are the short stack). In my experience, good players generally don't push against a really short stack in the BB. The short stack tends to call (an if really short, then has to call). The other players also may want to keep the shorty around if in a bubble situation.

Anyway, with players of equal (and decent) skill, I wouldn't drop a ton of EV to preserve future FE. However, if the other players are worse end-game players than you (i.e. tight), then I think having fold equity is worth a lot and thus worth an in the moment -EV play to take a shot at.

Pokerscott

[Moonsugar]

That is killer. Too bad you don't have enough hands. Very interesting stuff. Besides stack sizes I think the 2 other most important variables will be when blinds change and how many players are left in the game.

[Moonsugar]

Not that it matters now cause you are ignoring me, but if PVS' data for 1-5bb is accurate then you were correct on the QJ poll. The players would have to be really loose to make that push incorrect. (They can still be theoretically loose enough to make the push incorrect, but in reality that will rarely occur.)

You were right, I was wrong. Hopefully somone besides me learned something from our debate.

[Moonsugar]

If you look at some of PVS replies I think he shows at least that you should consider pushing sometimes even though ICM models the push as -EV. I think we are a far way from quantifying how much ICM is wrong but I think the shorter your stack, the fewer the players, and the sooner the increase in the blinds the more you should be inclined to push.

[nokona13]

I'm assuming this graph is the difference between the prize pool percentage predicted by ICM and the actual prize won, averaged? If so, did you try to take into account that the ~2,000 hands you used from each tournament are highly dependent on each other?

[microbet]

I know this is a small sample size, but I hope Eastbay will take a look. He was doing some position based corrections to ICM and I think he was doing the greatest correction to the UTG position, but you are showing UTG to be the most inline with ICM.

I haven't started yet, but I'm going to do that project about collecting hand data and would certainly share data for other projects.

[adanthar]

Meh. I can still see titles of posts.

Look, among other reasons, I post here because I was awful a year and a half ago and learned enough from this site to get to this point. It's fine to question my play - there are a half dozen posters I can name off the top of my head just in this one subforum that are equal to or better than me, and it's no contest in the high stakes NL or shorthanded forums. It's not fine to be obnoxious about it.

Having said that: look at the overall picture and don't try to quantify it, either with or without an improved ICM. The ICM is a useful toy I played with for 5 minutes once, saw the flaws of and left alone because even if you could make a perfect version of it (you can't, microbet is absolutely right about different ranges of calling ranges and that takes too much user input) you'd never be able to make a decision based on it within 30 seconds.

I'll be happy to see a version of this that says to make a call with kings, queens and maybe JJ-88, AK-AQ in the example I put up. It could still be off in the other direction but it would be pretty useful to draw a comparison with the original.

Anyway, apology accepted.

[Phil Van Sexton]

[ QUOTE ]
I'm assuming this graph is the difference between the prize pool percentage predicted by ICM and the actual prize won, averaged?

[/ QUOTE ]

Yes.

[ QUOTE ]
If so, did you try to take into account that the ~2,000 hands you used from each tournament are highly dependent on each other?

[/ QUOTE ]

no. I've heard this argument before, I think, but I still don't understand what this means.

My goal was just to write a program that could report on a bunch of data using different filters. My filter for these charts was "use every 4-handed hand". I could certainly run it for with other filters if/when I get more data.

btw - This is only for 4 handed hands, so it's only like 40 per tournament.

Again, I just posted this to get some discussion going. The sample is too small to be useful yet.

[Phil Van Sexton]

[ QUOTE ]
I know this is a small sample size, but I hope Eastbay will take a look. He was doing some position based corrections to ICM and I think he was doing the greatest correction to the UTG position, but you are showing UTG to be the most inline with ICM.

I haven't started yet, but I'm going to do that project about collecting hand data and would certainly share data for other projects.

[/ QUOTE ]

I've already talked with Eastbay about this a couple times. I think he's knows there's no point in getting into the numbers just yet, especially since I may never finish this.

I don't remember his UTG discussion, but it could just semantics. As I said earlier, my BB number is your ICM before posting or being dealt cards in the BB. You would use the my BB numbers to determine if you should fold after being dealt QJo when your are UTG.

Since you will be BB after folding or doubling up, you would use the BB value in your calcs when you are currently UTG. Maybe he named it differently. It's confusing either way.