Just upgraded to Pro Elite. Liking what I am seeing already

This (FGS) takes into account the fact that we are UTG and we'll be posting the BB in the next hand, and therefore has us shoving wider. I don't know how it works exactly, but I'd assume that it doesn't take into accounts things like that if we obtain the chip lead then we're going to be able to apply more pressure.

Interview with the HRC creator in which he talks about FGS

I don't think any program does this Sauce stuff. It can be improved, as Chidwick said, using the equity of 77 vs shoving range and constructing the scenarios weighted by the likelihood of each one happening. I think this solves the only thing that is missing on it.

Hey Stephen,

"The one flaw I see in this video (if we let the burger analogy slide) is the second bullet point on the "case study: calling with 77" page. Assuming we are still in the tournament with an 18bb stack every time we call entirely negates the worst part about calling the shove. In reality 42.7% of the time we will have busted the tournament the hand before and have zero +$EV shoves and 57.3% of the time we will be afforded this extra profitable opportunity the next hand, whereas if we fold we have a 100% chance of making it to the 23% profitable shove range the following hand. This has to be taken into account. "

The fact that we're moving from blind-->button in the subsequent hand is independent of the fact that a bigger stacksize increases our EV in subsequent hands (although both of these things occur together in my example).  We could re run an importantly similar example in a situation where we moved from UTG to UTG+1 at a 9 handed table, and the stacksize point still holds, although it always has to be modified by the fact that our EV is greater than 0 at the poker table provided we aren't in the blinds.  That we have a +EV opportunity next hand just means we need to set the threshold for shoving in the first hand relative to our opportunity cost of missing subsequent hands.  If we meet our opportunity cost, then it's still better for us to gamble for the big stack provided that our opponents are all playing ICM tight, or nearly so.  I realize this is kind of a tricky point, though, but my intention was to make the example as simple as possible.

I'll try and explain this another way because this first paragraph is reading pretty densely for me, heh.  Ok, so within the ICM model we know that we aren't considering our EV in subsequent hands, and we all agree that this is an important assumption of ICM that isn't true of real poker (and crucially, in my hand example, we move from the worst position in the BB to the best OTB in the next hand).  My point is just that ICM's ignorance of our position on the next hand is independent of its ignorance about our changing stacksize.  So, to keep things independent, I measured the EV of our changing stacksize on the next hand by staying inside the ICM model, and measuring the number of +prize pool EV ICM shoves we have with the bigger stack versus the smaller stack.  So, by keeping ICM's assumptions constant in both hand A and hand B, and measuring EVs from inside the ICM model, I thought we could zero in on the thing we want to measure which was ICM's ideas about the way stacksize influences the freq of +Prize EV shoves.

Another thing to think about is that if we are trying to find an equilibrium strategy that involves calling slightly -$EV hands to get the chiplead we also have to slightly diminish the value of having that chiplead since we will then get called slightly wider by the remaining players once we have it since they are playing the same strategy.

I agree, and I mentioned this in the video.  My point is just that the equilibrium strategy is looser than ICM and tighter than CEV, I don't know exactly where.

Stephen,

So you disagree with calling 77s in that spot?  I think it depends on table dynamics and how good the opponents are.  So against two really good players I would gamble there, against players that aren't that good not worth the gamble.  

If we are all evenly matched players I like the call.  Against the button's 43.5% shove we have around 56% equity.  So 56% of time we will have a stranglehold on the match and probably win.  

According to ICM...  if we win that hand our chances of winning the tournament go up to 50%. (5btn 14sb(us) 9bb).

In such scenario when you have a medium stack very few hands are actually just slightly -EV, most are very -EV, so this doesn't change much in calling ranges. ICM pressure is very high and (big stack push range % / medium stack call range %) ratio is simply huge.

Could you PLEASE make a non satty SNG vid, discussing your views on ICM and and ChipEV, would be really really interesting to watch.

enzyne, 

I think your points will be easier to understand if you distinguish the times when you're talking within the ICM model from the times you're talking outside of it.

Ben, ICM based analysis of single hand doesn't really consider bust probability and its implications. They can be achieved through some serious mental modeling only.

When you call there is a ~50% chance that tournament ends right here, and upcoming hands won't give you any opportunity to continue making +EV decisions hand after hand. In another 50% of cases you double up and continue tournament in a new scenario - as a big stack where chips flow in your direction. 

When you shove bust probability is much lower pretty much always, so by continiously pushing you can accumulate +EV hands one by one and improve your tournament result. 

This dynamic is pretty much outside of ICM single-handed scope, but it has to be considered when you are contemplating a ~0% EV call.

For example, in HU sng if opponent pushes first hand and you make a slightly +EV call, tournament can end immediately and you will be negative cause of rake being > than +EV call you made. So you can choose to pass some scenarios which have high probability of bust scenario in order to even have a chance to accumulate enough +EV decisions during tournament to beat rake.

Right, but irrespective of the EV you are analysing the future implications of the call under the assumption we win every time, no?

Stephen,

"Right, but irrespective of the EV you are analysing the future implications of the call under the assumption we win every time, no?"

Yes, but it doesn't matter.  Just imagine for a second we're playing an imaginary poker game where position doesn't matter (like stud, with its ante/bring structure).  In this imaginary game we know calling in hand A is worth 0.  If we fold we get X% +$EV plays on hand B, and if we call we get some additional number of +$EV plays on hand B.  So we should call in hand A.  

The blind structure complicates things slightly and forces us to weight positional considerations against stack considerations, but it's useful (for me at least) to separate these two points in theory.

vulture,

I recorded some of the play at the final table, I'm going to go over that footage.

GameTheory what the heck is that? 

GameTheory what the heck is that? 

http://en.wikipedia.org/wiki/Kinesin

Sauce opinion on it. One of the reasons I bought it was this "foreword" from him. No regrets, amazing book,

zen, don't have HRC now but from Sauce example we shouldn't be looking into only 1 future hand ? What is the result ?

Raphael:

I used 6 future hands in the FGS simulations. The most I could use before calculations became too heavy. My computer is a beast, but trying 9 hands choked it. :-) 

tks

I'll be working with HEMRcalc during the final table review, but I won't be giving a tutorial on how to use it.

Thanks Luke !

Sam,

Thanks a lot for laying it out like that, well done.  (and making my life a lot easier :)

My whole point was that

ICMV(call and win) + PrizepoolEV of having a big stack = ICMV(call and win) 

or at least very close to it. But suppose i'm wrong, and PrizepoolEV >0, and we do gain EV in subsequent hands. In that case you have to compare

A) ICM(V call and lose ) + ICMV(Call and chop) + ICMV(Call and win) + PrizepoolEV of having a big stack

and 

B) ICM(folding 77) + PrizepoolEV of having our old stack.

Only once you establish that A > B, you can say that calling 77 is better than folding.

Thanks for putting it in those words that make it a lot easier to understand, I was having trouble agreeing with what Ben was saying and really siding on Stephens side with the idea that he wasnt accounting for the times we bust. This does leave me wondering though, does this mean that everytime we are put in a neutral EV spot we should always go with it (using this particular 3 handed example)? If not, could you please give an example using these parameters. Thank you

Theories,

"Thanks for putting it in those words that make it a lot easier to understand, I was having trouble agreeing with what Ben was saying and really siding on Stephens side with the idea that he wasnt accounting for the times we bust. This does leave me wondering though, does this mean that everytime we are put in a neutral EV spot we should always go with it (using this particular 3 handed example)? If not, could you please give an example using these parameters. Thank you"

The answer depends on what you mean by the bolded bit.  If by "neutral EV spot" you mean a spot that an ICM model like HRC defines as prize pool neutral, then probably yes.

Sam,

"Given that stack sizes remain similarly distributed the $EV of folding isn't the $EV of our stack after folding but the $EV of our stack minus our expected loss rate times the number of hands we expect to play at this stack size."

Agreed.  Although, it's very difficult to model the expected number of hands we get at this stack distribution, as well as other similarly bad -CEV stack distributions.  Because it's so hard to model the stack implications of future hands, I think it's better to default to what people's goals seem to be with their stacksize and tournament life.  Statements like "he's trying to move up the ladder" or "he's playing for the win" or any of the gradations between them actually make a lot of sense in this context and figuring out people's intentions can help us fill in the gaps in our models (which typically only work in a vacuum, or over a small number of hands).

I think you are correct. The real question, how exactly to consider POB in such scenario. If we win, we get a lot of push possibilities with low bust probability. If we fold, we continue to get pushed at and at some point we will have to make a call with high bust probability.

Soo.. why not now? If we wait a few hands, our double up will get us back to medium stack and won't grant us chances to make +EV pushes with low bust probability as big stack, thats for sure.

in fact if we isolate the second hand we have a greater increase in prize pool equity in the 9bb/12bb/7bb scenario than we do in the 18bb/3bb/7bb one.

can you elaborate on this?

Pretty please make a non satty SNG vid with your views on them :)

Stephen,

I definitely agree that folding OTB (in a non ante structure) is going to be +$EV for the same reasons you do.

Like I said in my prior post, I was just doing a thought experiment to show that ICM is self contradictory if we compare it against itself over multiple hands.  I'm definitely not making precise recommendations about optimal play (optimal here meaning in a broader than ICM context) in the sng spots I was describing.  For instance, I still don't know whether calling 77 is optimal in the toy sng I described in the video, but I do think the example shows that if our opponents are playing ICM then it's likely a call.  

It's funny, HRC doesn't show the $EV of folding- it just shows how +$EV it is to call or raise. So the ICM models don't allows us to quantify the opportunity cost (relative to folding) of making really thin +$EV jams or calls.  Thanks for pointing this out, and it's another reason why our current models aren't as accurate as we'd like.

Ben, 
ICMIZER does show $EV of folding - and both use $EV_fold to find the results they give. 
If you get a result of +0.03 that is it's EV compared to folding.

Unless there is some anomality in payout structure this statement usually holds true. But of course, if someone has 1 chip and if you bust him from tournament your minimal payout increases, winning this 1 chip from this particular player is very valuable and losing 1 chip to another player is not so valuable.

So busting a player with stack of X chips is not the same as winning X chips from big stack, for example.

Usually, yes.  Take for example an MTT structure fairly early in the money.  Maybe 10% of the prize pool will have been payed out for mincashes already, so the chips we win don't sum to the value of the full prizepool, and so they're worth less.

Limping is a fine strategy, but it won't reduce ICM confrontations.  It will just reduce the size of the pot relative to raising.

Thanks for the answer Ben!

Yes, the think is that by playing smaller pots we put less chips in the pot in average and this is important in a FT. I am talking with an average stack of 40bb or so.

By limping we don't let the oponents that CC with position (BTN, CO, MP) put presure on us by betting often with position and apply pressure to us since we will play with much bigger SPR (stack per pot ratio). (We are going to have to check-fold the flop often in this situation, which will occur often).

You have to take in count that when our stack gets down from 20-25bb (when he would bet turn/river) our BF is huge and we would have to fold very often.

Question 2: Besides of that, do you think the EV of limping in EP-MP has similar EV to Open Raising in a normal tournament or cash scenario? (without ICM considerations)

Interesting that you bring this up. Usually people that limp are thought to be weak players or fishes. It seems some people have been changing their views and limping is a viable strategy.

Interesting topic.

Ben, do you think in a FT where do you have a huge edge with a lot of weak players it would be better to limp in general than to raise first in? This makes the pots size smaller and pays out more your postflop skills (where you are much better than then).

Hey Ben - cool vid. I'm strictly a cash player, and had the same question as Jeff when looking at your examples.

Okay, so clearly we should open less than ICM suggests against players who call too much. How do we quantify this, though? (i.e. in the example where we open 42%, and the blinds can only call 8% according to ICM- let's say one is calling 10%, and the other 35%, or whatever- what's the best way to go about figuring out exactly how much we should loosen up our range- or at least to improve our estimate of this?

I fully agree with your points. I was thinking the same thing about the bluff success threshhold( havent looked the exact numbers tho). 

But also i have 1 more point- if it is true that icm affects different stack sizes in a diff.way (mb thats an illusion) there will be some spots where there is no/strange equilibrium.

Let`s say- 3 players, stacks A-50chips /B-15/C-15 (prizepool 3/2/1) and hand is between players A and C and pot on the river is 10  (C has 10 left) and C is with polarized range. So, to shove on bluff he risks tournament chips and busting so A could overfold. This means C could bluff more and normally(!) there would be equilibrium when C should bluff X% and A should call Y%.

X1 over 33%

Y1 under 50%

But mb A(reasons-even if he loses he is chipleader and some intuitive things like this) is less affected from icm than C, so he would start making some +chipEV calls that are ok in$EV for him but will make C' bluff losing $EV(for C). 

So the equilibrium would be connected to:

X1 / Y1 = Z1        X2 / Y2 = Z2

Z1 > Z2

So this would decrease shortstack's EV (in general in this spot) compared to cash game model or ICM with equal stacks model even when he is the bettor.

Is that true or im dellusioning myself? :)

2D,

Your argument makes perfect sense to me.  There's a similar effect reflected in the HRC outputs I look at in following videos (although these only work for preflop).  I compared a raise/fold tree without ICM to a raise/fold tree with ICM (and in the ICM tree I inputted the stacks and positions from hand 1 of my tournament review where I'm the big stack in the CO), and the equilibrium reflects the fact that the big stack can overbluff by opening much more often and 3betting much more often than the other players, and the once the big stack has VPIP-ed or failed to VPIP the other player's play changes a lot.

I'll definitely check out this effect in a later part of the final table review, I think it should show up in the HRC calculations.  We can find the effect's magnitude by looking at an ICM sim where Z1>Z2 and comparing it to one where Z2>Z1.  I was also informed that CREV has a tournament mode which might take these ICM adjusted equilibria into effect for hu pots.

For some explanation of FGS:

FGS (Future Game Simulations) Calculations Explained

Interview with HoldemResources Calculator Inventor Helmuth Melcher

I'd say it depends on the size of blinds and average stacks? As I understand it, FGS takes into account the next (or next few) hands of play. If the blinds are still a small percentage of stacks, people aren't really going to be jamming/calling off wide. The higher the blinds, the more stacks should be going in the next hands and so the more FGS should be considered.

I'd seriously suggest toying around with FGS on Icmizer with different stack sizes for the example in your question. You'll get more out of that than any general answers people can give you. The real question is exactly how much should you be widening your range in that spot and that obviously depends on your opponent's jamming range and the blinds/stack sizes, and how will people be playing in the next few hands.

Yes it does. Depending how deep you run the situation in FGS this effect will trickle down towards your original callrange going wider and wider

I added FGS to the equilibrium calculation and posted the new SB calling range. I did not freeze the BTN range, although you can.

Hey Ben- you got it

Miam,

I think your argument is right, but I think the approximation you're talking about is the same, or closely related to, the ones that I talked about in the video.  In other words, you're saying correct strategy leads to different finish % then the ICM model predicts so it's incomplete, and I'm describing some of the strategy changes that must be correct if the ICM model is right but if those changes are right then the ICM model is wrong.  

Cmw,

Hold em Resources Calculator has a utility that lets you define your opponent's shoving ranges and then finds the correct strategy to maximize $EV. It sounds like this tool might help you take better advantage of your opponent's play.

I think this kind of argument is circular. Sure, there exists some algorithm that could output really accurate results, but the devil is in the details :p