Coaching: Fixed Limit [Advanced] with Boomer2k10

just stop already please

Just me or everybody?

Even if Avataren gets pissed off, I think I should correct one thing where I made stupid mistake.

Originally posted by YohanN7

...bluff surprisingly much because you
can give up some of your bluffs (and naturally not value hands)
at will on later streets whereas your poor opponent must be prepared to
call down all the way...

Just as player A has hands he bluffs one, two or three times with, player B must have hands that he will bluff catch one, two or three times with. This is a sign of internal consistency of the model.

My bad. How silly of me to think that the disadvantaged player B must call down all of his hands. This clearly is not the case.

Anyway, that was just an attempt to find an intuitive explanation why A can bluff so much more than on an even playing field. This must be because A has some essential advantage over B. But maybe it is not possible to isolate the reason beyond the massive range asymmetry that is very advantageous for A.

Originally posted by Turuntor
Even if Avataren gets pissed off, I think I should correct one thing where I made stupid mistake.

Originally posted by YohanN7

...bluff surprisingly much because you
can give up some of your bluffs (and naturally not value hands)
at will on later streets whereas your poor opponent must be prepared to
call down all the way...

Just as player A has hands he bluffs one, two or three times with, player B must have hands that he will bluff catch one, two or three times with. This is a sign of internal consistency of the model.

My bad. How silly of me to think that the disadvantaged player B must call down all of his hands. This clearly is not the case.

Anyway, that was just an attempt to find an intuitive explanation why A can bluff so much more than on an even playing field. This must be because A has some essential advantage over B. But maybe it is not possible to isolate the reason beyond the massive range asymmetry that is very advantageous for A.

I don't see any essential advantage for A over B, or any asymmetry. What I see is that GTO probably (remember, it's only a model we are discussing) advocates looser and wilder play in general when heads up than I previously thought. Bluff more than 1-street balance dictates - and call more. It also seems to provide a justification for going ballistic in heads up play on the flop, as well as an explanation of the well-established fact that "maniacs" aren't as easy prey as one might think when pots are contested two-handed.

Originally posted by YohanN7

Originally posted by Turuntor
Even if Avataren gets pissed off, I think I should correct one thing where I made stupid mistake.

Originally posted by YohanN7

...bluff surprisingly much because you
can give up some of your bluffs (and naturally not value hands)
at will on later streets whereas your poor opponent must be prepared to
call down all the way...

Just as player A has hands he bluffs one, two or three times with, player B must have hands that he will bluff catch one, two or three times with. This is a sign of internal consistency of the model.

My bad. How silly of me to think that the disadvantaged player B must call down all of his hands. This clearly is not the case.

Anyway, that was just an attempt to find an intuitive explanation why A can bluff so much more than on an even playing field. This must be because A has some essential advantage over B. But maybe it is not possible to isolate the reason beyond the massive range asymmetry that is very advantageous for A.

I don't see any essential advantage for A over B, or any asymmetry. What I see is that GTO probably (remember, it's only a model we are discussing) advocates looser and wilder play in general when heads up than I previously thought. Bluff more than 1-street balance dictates - and call more. It also seems to provide a justification for going ballistic in heads up play on the flop, as well as an explanation of the well-established fact that "maniacs" aren't as easy prey as one might think when pots are contested two-handed.

Then you don't see it and thats fine with me, not that keen on arguing.

Perhaps I could see an asymmetry if it is pointed out to me. All of this is almost entirely new to me. (Even if I have an annoying ability to sound like I think I know it all when I don't. ;))

Originally posted by YohanN7
Perhaps I could see an asymmetry if it is pointed out to me. All of this is almost entirely new to me. (Even if I have an annoying ability to sound like I think I know it all when I don't. ;))

Please accept my friend invitation in comtool. I think of my comments to your posts are not of general interests, I'll send a pm instead trying to clarify some of my points. Not so that I would know it all but some of the points come quite directly from literature and can be considered cold facts, not that much arguing about them.

Wow what a lot of commenting going on, unfortunately I dont have time to read all a them right now, but will do it later.

@Boomer

Originally posted by Boomer2k10

Wrong. Strong draws like FD + Overcards can be up to 30-33% if you include drawing to a hand that can value bet.

Ofc but you almost never bluff these, or you are doing something wrong, because your bluff range is built up bottom-up, you fold the worst x% of hands, and bluff the bottom x+y%, and that makes draw+overcard combos contained by your calling range. The strongest draw you usually have is plain a flushdraw and that has 19% EQ

What you're assuming here is that I'm going to want to bet all my bluffs on the river and that is not correct under almost any circumstances

Why not ? Why would you want to give up bluffs ? You have to give up value hands for sure, but that doesnt mean you have to give up bluffs by any means. Bluffs are used to make your opponents fold and not to see a showdown, you only have to give up them if you dont have nuff v-bets to support. What matters is the bluff:value ratio only, and not the # of combos you have, and you can(and have to) adjust this ratio by the # of v-bet combos and not the # of bluff combos.

I can agree tho that you have to consider that your bluffs have equity, but I dont see why it would justify massive overbluffing (like 43% vs 16% on the flop) in pre-river play, and why cant this problem be solved by carefully choosing your bluff hands so that their equity spreads out across possible river cards.

This gives us a total of 2.75 bluff combos into a total number of bets of 8.75 (Adding the bluff combos to the river hands) = 31% of our range.

And this is the point where it gets cloudy for me. Where the hell does the number 8.75 come from ? The pot never gets that big.

Again now on the flop all 8.75 combos are value on the turn so let's work out how much we need on the flop

Why should I assume that we have value combos on previous streets according to this ratio ?

Say we have 100 combos on the turn from which 31 combos are bluffs, and the river is a blank card that doesnt hit your range at all, or it hits 3 combos, or 10, or whatever # of combos. How does this backwards calculation work in these cases ? Becasue this model assumes that you hit every single river with the same frequency and this couldnt be more far off from reality.

And how would you know at the time that it was +EV to call on the turn unless I told you my exact hand in which case it turns into an exploitative game anyway? Or you knew my exact range on every single possible board?

This is completely wrong, and this is not the first time I see that you dont understand what exploitation means. Explotiation doesnt mean that you tell your hand to your opponent, or "you knew my exact range on every single possible board". If you play unexploitable there is NO WAY that your opponent can take advantage on you, yes you can even tell your exact ranges to your opponent, because there isnt a strategy, not even a single one, which he could find profitable against your play.

And coincidentally your EV may be 0 (fold) but mine is WAY higher because you're over-folding and thus any bet I make on the river is automatically +EV, especailly if you're making an extreme adjustment.

You're talking about single hand examples which aren't realistic in poker, it is a range on range situation. You cannot make every single decision with every single hand your opponent has 0 EV, sometimes he has AA.

What I am playing against is your range not your exact hand.

The way you're describing it here your EV can never be below 0 if you fold all the time. Pretty sure you know how that turns out.

The one who takes only one decision into account is you. Its not ony that river bet that could happen, your opponent can donk on blank rivers for example, and then youll be on the river trying to fold/call in balanced way with your due to the turn overbluff unbalanced river range. Or he can x/r you right on the turn and the same thing happens.Or overcall as I described previously.

IMO balance is not something you can add or substract to your play, it's rather like a dish, once its overdone, or oversalted, or whatever way its screwed up its screwed up, there is no way you can fix it. And if you overbluff at some point that screws up you ranges and your entire game will be unbalanced on later streets. I dont mean that every action you make will be unbalanced, perhaps you can work out a balanced river c-bet range but then your x/b range will be one where the unbalance appeares and you are going to lose money against balanced lines no matter what.

And I still think a turn overbluff can be exploited with an overcall, I show you the below figure I hope it makes it clear:

The first part shows the x/c, x/f lines vs a balanced turn c-bet range, this calling range has at least 0EV against every possible line (providing that you played unexploitable so far ofc). If someone overbluffs the turn, the EV of the line automatically increases no matter what you do on the river. Your opponent doesnt have to overcall because he auto profits from the money you are donating him by your overbluff. It is just a plus for him that he can move some hands from his x/f range itno his x/c range with a profit.

@kavboj

The model builds on a couple of key ingredients:

• The complete range, and each of its parts that you bet on the river is for value.
• On preceding streets, you add bluffs to the total river range that you are willing to give up.
• The relative number of additional bluffs are calculated according to 1-street balance (because you give these up).

You could read the second to last post of mine (the one humbly beginning with Eὕρηκα) on the previous page for an explanation of the concepts, whether they are good or bad. An explanation of the (unfortunate) designation of the number 8.75 as "pot size" can be found there too.

If we define overbluffing and overcalling with respect to traditional 1-street bluff/call frequencies, then, yes, the aggressor will be overbluffing.

Can it be exploited by overcalling?

No, it can not. You must overcall on both the flop and the turn (, but not on the river), otherwise you are the one being exploited.

The slightly paradoxical "All bets on the river are for value" is easily explaied. Value is, of course, value. The bluff part is, by definition expected to break even versus all opponents whether they call 100%, 0% or something in between. In reality, people make mistakes. On occasion they call with hands that can't beat a bluff and fold hands with too much value. A GTO river bluff range will show a profit versus any human. Therefore the complete river betting range range can be considered to be for value when we are at the turn.

Take it from there, and you'll get the model.

The model can, as it stands be countered by raising, but not by any form of calling.

Additionally there are a couple of things about your post:

1) Not bluffing overcards and FD's

As discussed, the "bottom up" form of bluffing isn't optimal either in terms of maximizing our turn value or future card distribution so that part of the previous model now has to be assumed to be false unless the board is simply so dry there is no option but to "bottom up" bluffs.

Not to say you still don't bluff utterly hopeless hands but hands with equity you'll have to fold to a raise generally now fall off the bluffing wagon, especially in position.

2) You can tell your exact RANGE to your opponent and he can't do anything but you can't tell your exact HAND which was the point I made.

And I have not said this is a GTO model so therefore I could tell you my range and you could exploit it. For GTO to be true you'd actually bluff a different amount of combos relative to the value-strength of your bluffs and that's where it get's tricky. This is a simple "bluff vs value-bet" model and doesn't take semi-bluffing into account (although that actually would suggest we "bluff" more it is tempered by the fact we can be raised if we bet)

3) You want to give up on bluffs because if you bluffed according to 1-street value you'd somehow have to INCREASE the number of bluffs you make between flop and turn.

Additionally if you always c-bet your bluffs and give up on value hands the relative frequency of your bluffs actually rises by street and since your opponent is getting better and better odds to call you that can't be right either.

4) Regarding your "1-street accusation"

CALLING is a ONE-STREET decision because fold ENDS THE HAND. End of. You fold more than alpha dictates your opponent has a +EV bet with any 2 cards...that's what autoprofit is. BETTING is a multi-street decision because you have to set yourself up for the 0,1 situation on the river where you either have the best hand or you don't.

What you're trying to tell me is that you can make a +0.1BBEV turn decision and give your opponent a +1BBEV autoprofit on the river and you'll be in profit because of that. Can you overfold slightly to the tune of 0.1BBEV and maintain balance? Yes you can but that means balancing your turn overcall and river overfold ranges so either way you come back to a balanced position. Just simply calling according to alpha means you've weakened your overall range through your turn overcall and therefore have more -EV hands in your range vs your standard range.

If you call river according to alpha after overcalling the turn you're paying off the vast majority (86%) of your opponent's range and thus making a -EV decision on the whole between turn and river.

5) What you're arguing about is the %-age values

So ok let's look at this from a strictly nash solution

Between turn and river you have to put 2BB into a 7BB pot and let's assume you're all in so no raising no rubbish and you work out in this situation you should bluff 28% of the time, because that's the Nash solution to the problem.

Let's assume this all goes in on the river to give us the easiest solution to the problem: 28%

However:

That WOULD be correct but unless your opponent has you completely dead you can't have 0% equity on the turn. Therefore some of your hands have a "value element" and cannot be counted as total bluffs

The correct adjustment in this circumstance would actually be to bluff MORE combos of hands because your bluffs "have outs"

So we can say right off the start that in terms of combos 30% can't be far out if all the money went in on the turn

Now let's move this on a little:

You're NOT putting all the money in on the turn. You're putting 1/2 of it in on the river.

The river is a strict 100% and 0% situation so we have the solution to that. You bet 14% of your hands (1 in 7) as bluffs. What we add to it is what additional hands we want to bluff on the turn because our opponent is making 2 separate decisions (turn AND river) which we have to make him indifferent about.

Yes the term "pot-size" is unfortunate and I regret using it but it was the only way I could articulate it at the time it's not the size of the pot, it's the ratio of combos we're using. We know we have 6 value combos on the river for every 1 bluff and we are adding bluffs to our range as we go backwards through the model.

Why would you bluff LESS that what Nash is telling you for an all-in bet (with no possibility of improving) when you have both equity AND the fact your opponent has potential RIO working against him on the turn (i.e. leverage concept).

If you want to read about the model, as I said, it's calling Stacking Bluffs and is published in at least 3 major books on poker including Phil Newall's latest one.

Is it a perfect GTO solution? No of course it's not (which is why if I told you my range you can still exploit it) but I never said it was, I said it was a model to work from which is superior to my previous model. There are many other areas to look at when it comes down to this, equity of turn bluffs, future card distribution, board texture, alternate lines, the fact my opponent can raise, narrow range situations etc. but it's definitely closer than "bottom up" bluffing and treating turn and river as 1 street either individually or together.

People seriosuly need to get out of their heads than a human being can put a GTO solution to poker on the table. Whether it's FL, NL SNGs ANYTHING NO-ONE is capable of playing Game Theory Optimally and therefore EVERYONE can be exploited to a degree. What I try to do with these models is minimize that exploitation and open the doors for you to investigate yourselves rather than trying to give a blanket strategy.

Originally posted by YohanN7
If we define overbluffing and overcalling with respect to traditional 1-street bluff/call frequencies, then, yes, the aggressor will be overbluffing.

Can it be exploited by overcalling?

No, it can not. You must overcall on both the flop and the turn (, but not on the river), otherwise you are the one being exploited.

The slightly paradoxical "All bets on the river are for value" is easily explaied. Value is, of course, value. The bluff part is, by definition expected to break even versus all opponents whether they call 100%, 0% or something in between. In reality, people make mistakes. On occasion they call with hands that can't beat a bluff and fold hands with too much value. A GTO river bluff range will show a profit versus any human. Therefore the complete river betting range range can be considered to be for value when we are at the turn.

Take it from there, and you'll get the model.

The model can, as it stands be countered by raising, but not by any form of calling.

Who are you and what have you done with Yohan?

Seriously though man, really nice posts ITT.

Originally posted by Boomer2k10

Originally posted by YohanN7
If we define overbluffing and overcalling with respect to traditional 1-street bluff/call frequencies, then, yes, the aggressor will be overbluffing.

Can it be exploited by overcalling?

No, it can not. You must overcall on both the flop and the turn (, but not on the river), otherwise you are the one being exploited.

The slightly paradoxical "All bets on the river are for value" is easily explaied. Value is, of course, value. The bluff part is, by definition expected to break even versus all opponents whether they call 100%, 0% or something in between. In reality, people make mistakes. On occasion they call with hands that can't beat a bluff and fold hands with too much value. A GTO river bluff range will show a profit versus any human. Therefore the complete river betting range range can be considered to be for value when we are at the turn.

Take it from there, and you'll get the model.

The model can, as it stands be countered by raising, but not by any form of calling.

Who are you and what have you done with Yohan?

Seriously though man, really nice posts ITT.

your a big fat liar boomer. its bad to be a liar.

Originally posted by Boomer2k10
If you want to read about the model, as I said, it's calling Stacking Bluffs and is published in at least 3 major books on poker including Phil Newall's latest one.

Is it a perfect GTO solution? No of course it's not (which is why if I told you my range you can still exploit it) but I never said it was, I said it was a model to work from which is superior to my previous model.

Are you really sure about this [e: on the model being an improvement] ? The stacking bluff strategy is derived as the optimal solution (one of the three books you mention, Chen-Ankenman) to a simplified model. But the model assumptions are quite severe:

1) The betting player is clairvoyant and the caller is not. In this case, this essentially means that the betting player sees the cards of his opponent.
2) Hand values are static through the streets: once nuts remain nuts also later.

These strong assumptions make the model solvable, but trusting this strategy to work in real LHE conditions seems very suspicious when the model assumptions are not met even approximately.

To me this model reveals qualitatively the existence of an interesting characteristic of GTO strategies in some situations. It probably is not good for quantitative predictions of optimal frequencies in realistic poker.

My previous model involved heavily polarized ranges which didn't take into account the advantages of not bluffing with the bottom of your range

Additionally it also had the same errors in terms of dealing with raises that this model does and relied on a relativistically simple preflop game in HUHU situations in order to make it work easier

So, in short yes. This model has no more failings than the pervious and corrects many of its flaws so it's superior

It's not like it's a mile away from my previous model anyway, I always recommended at least high 20's bluffing frequencies in terms of %-ages. but it's based on a lot more solid footing.

It is not like this is the end of me building a poker game model, this is the foundations to build on. I'm not saying this is the end of my building, this is where I'm STARTING.

The model also confirms, or at least shines a light on other areas as well such as:

1) 100% flop c-betting is exploitable (The model gives back 43% bluffs and given we miss the flop 65% of the time this is worrying...although range asymettry and board texture may counter this in some parts)
2) You should definitely drop some bluffs from your range on the turn to the river.
3) There may be some merit to alternative lines when oop (donking etc)

The assumptions made for this model are severe and there are other limitations, such as there's no such thing as a semi-bluff, but this is the simplest way we can start. You can't just present a model and say "there done it". Even the most advanced poker AI's solve a simpler game and then expand it.

For me the model's main limitations are:

1) It only applies to calling
2) It doesn't take into account semi-bluffing
3) It doesn't tell us what hands to bluff even if it gives us a relativistic frequency (i.e. we have to figure out future card distribution)

I guess this is going towards more or less useless academic discussion on differing opinions because neither side can't present really compelling arguments to support their opinions. And of course it's so, the full game is not solved yet. But let's still continue for a moment.

Originally posted by Boomer2k10
My previous model involved heavily polarized ranges which didn't take into account the advantages of not bluffing with the bottom of your range

Additionally it also had the same errors in terms of dealing with raises that this model does and relied on a relativistically simple preflop game in HUHU situations in order to make it work easier

So, in short yes. This model has no more failings than the pervious and corrects many of its flaws so it's superior

Superior in that sense, no doubt. But it's totally thinkable that one model takes into account many more things than a simpler model but still gives predictions that are systematically far worse when the model is used outside the conditions it was designed for.

It is not like this is the end of me building a poker game model, this is the foundations to build on. I'm not saying this is the end of my building, this is where I'm STARTING.

I'm of the opinion that it's not wise to use the strategy as a foundation. It's builds on assumptions that are not fulfilled (even nearly) in most reallistic situations. But now I start (or actually continue....) repeating myself. Afterall, it's just my opinion and I guess I have already given my arguments for it.

The model also confirms, or at least shines a light on other areas as well such as:
1) 100% flop c-betting is exploitable (The model gives back 43% bluffs and given we miss the flop 65% of the time this is worrying...although range asymettry and board texture may counter this in some parts)
2) You should definitely drop some bluffs from your range on the turn to the river.
3) There may be some merit to alternative lines when oop (donking etc)

This I can agree with. It's smart to take some qualitative elements of the optimal solution into account and incorporate them into the full strategy in appropriate situations. In some rare cases the ranges may even be close to polarised (although in LHE less so than in NL) and the stacking bluff strategy can be applied as a whole.

1) It only applies to calling
2) It doesn't take into account semi-bluffing
3) It doesn't tell us what hands to bluff even if it gives us a relativistic frequency (i.e. we have to figure out future card distribution)

I guess I repeat myself once again but still couldn't help to comment:

1) Yes. By design of the model the ranges of the players are such that it makes no sense for the calling player to make anything else but call or fold.

3) True of course. In the sense of model, all bluffs are equivalent in value. None of them ever improves to beat the bluff catchers.

Turuntor, and also Boomer, have listed a number of shortcomings of the Bluff Stacking Model (call it BSM (the bs model ♦)). These are all valid objections, but some of them can be dealt with within the model.

Set aside that raising isn't allowed (or accounted for or whatever). Then the apparent problems with the model are:

• The aggressor is clairvoyant. He knows what's coming on turn and river. This is where the river betting range comes from.
• The aggressor has a holecard camera. This is why he can categorize holdings as bluff or value.

(It is useful to separate these concepts into two instead of just saying that the aggressor is clairvoyant.) Why is it this way? I haven't read the original sources, but it appears that the model is (at the outset) presented like "Suppose the board has the following appearance on the river. Then ...".

Clairvoyance first. Does actual lack of clairvoyance destroy the model? As presented ITT, we begin with the river betting range and then work backwards. But we don't know the board OTR so we can't begin with it. The workaround in this case is to consider all possible turn + river combinations (the flop is known), one by one, and then to take a statistical average. If you want, it's a straightforward (for a computer) EV-calculation. This will yield ranges for the flop and the turn.

Now, for each possible separate turn and river combination there will be a solution where the frequencies of bluffing are precisely those that has been presented ITT. In absolute terms, we don't know how many hands we bet, or which hands we should pick. This is a separate problem.

What is it that says that a statistical average (that tells us what to do on the flop and the turn) will deviate much numerically in terms of bluffing frequencies? Not much. I believe it's perfectly within reach to make the calculations (not necessarily even a simulation, but a true calculations) on a computer to confirm that, for most flops, the optimal bluffing frequencies will be according to what has been given so far.

Next, the holecard camera. How do we know what are bluffs and what are value bets? This too can be handled within the model. First off, versus a range OTR, it is pretty clear cut what is bluff and what is value. Since we are trying to approximate GTO, it is not too far fetched to put the opponent on our own range. (In case the opponent has a different one, we have just made a too pessimistic assessment, either our value range or bluff range will flourish more than the other suffers.)

On the flop and turn, each bluffing combo will have an element of value so the correspondence between bluffing frequencies and the actual fraction of the total range that we bluff with need not be the same. I don't know how to properly quantify bluff relative to value here, but I'm pretty sure that it can be done; at least that there are realistic models for it. This is really a separate problem from the problem at hand.

The conclusion is that it would be to hasty to scrap the model solely on the basis of the lack of clairvoyance and holecard cameras. The asymmetry (between the aggressor and the opponent) introduced by considering the board and the holecards of the opponent isn't really there after the statistical averaging and assigning the appropriate measure of bluff to each semi-bluff.

Once, and if, this sort of reasoning is accepted, then it is clearly beneficial for the discussion (and logically correct) to implicitly assume that we don't need holecard cameras and clairvoyance.

As I see it, the major shortcoming of the model is that it can be countered by raising and (beware now) donking and slowplaying.

Hi guys

Sorry to break the discussion but I'm afraid I will not be able to make it for the coaching tomorrow.

Will see you all in a week

Originally posted by Boomer2k10
Hi guys

Sorry to break the discussion but I'm afraid I will not be able to make it for the coaching tomorrow.

Will see you all in a week

Damned!

I had to divorce my wife to make time for that coaching. Oh, well, there are other coachings (and women)...

Originally posted by YohanN7

Originally posted by Boomer2k10
Hi guys

Sorry to break the discussion but I'm afraid I will not be able to make it for the coaching tomorrow.

Will see you all in a week

Damned!

I had to divorce my wife to make time for that coaching. Oh, well, there are other coachings (and women)...

Well in Sweden at least they're hot