Straddles: Math and Theory - Page 2 - Poker Theory

quesuerte · 05-31-2012, 12:00 PM

Quote:

Originally Posted by dgiharris

Okay boys and girls, I've updated my models making them a little more complicated and truer to life and am incorporating Hero's equity vs Villain's equity if called

Hero will straddle UTG and then when the action gets back to him, Hero will shove with ATC. For the purposes of the model, Hero is considered short stacked so we don't get into your awesome post flop skills and plays which would be impossible to model.

For this round of modeling Villains will call if and only if they have AT-AK, KQ, TT-AA and they will fold all other hands. Each villain has a "chance" of being dealt the above hand so hero will face anywhere from 0 to 6 villains per squeeze at a 9 handed table.

The model steps through various squeeze amounts. Obviously, villains aren't likely to fold for 10bb if there are 6 villains limping, however, the math doesn't change as far as profitability is concerned. However, what you can do is use your "intuition" to pick the bet sizing that you would squeeze based on the situation and then match that up to the model and see where it falls EV wise.

All bets are made in terms of big blinds. I used Poker Stove to determine Hero equity in the event of 1-6 callers. As this thread progresses I will "widen" the calling ranges of Villains and show the results

CONCLUSIONS:
If we are at a weak-tight table in which villains will only call our shoves with AT-AK, KQ, TT-AA, and we are short stacked, its profitable to shove fairly wide if we can isolate a lot of dead money. So we need 4, 5, and 6 limpers for routine squeezes to be profitable. The more limpers the better obviously. Similarly, we want to squeeze the "least" amount possible that will generate fold equity. We don't want to squeeze 50bb in situations in which 25bb would generate the same FE.

For the next rounds of modeling I'll tighten up Hero's range as well as widening our Villains ranges.

EDIT: doesn't "only" apply if we are short stacked, i'm just unable to model post flop play. So if you are 100bb you can still use these results as a baseline and structure to work around and then just incorporate your post flop game...

DGI, not sure what you are actually trying to show here.

This is not really showing anything about the profitability of straddling or not, only some maths about the profitability or otherwise of plays when it is limped to you in the straddle, which as others have pointed out should be the same from the big blind.

Even given the scenario is so unrealistic as to be fairly useless IMO (I know this is poker theory so I have no problem with that, just wondering if you believe the results to have any practical use.) there seems to be one huge flaw in your process. You never mention a limping range. If it is 100% then I don't see why you need a simulation to tell you that if people are limping super wide and calling raises super tight it will be profitable to raise ATC. If they are not limping 100% then that will affect their calling frequency. Obv if somebody calling a raise with 3% of total hands but only limps with 4% he's calling 75% of the time.

dgiharris · 05-31-2012, 04:23 PM

I admit that so far, this model is really about "squeezing" moreso than Straddling, but this aspect is the first step.

Second, I disagree with it being useless. Whenever I'm playing, Someone ALWAYS straddles and the MOST COMMON situation that occurs is that someone straddles and the entire table limps.

So, in my mind this seemed the first step in modeling "Straddles". Since I ROUTINELY squeeze straddled pots with a pretty wide range and find my squeezing to be profitable, I also had an interest in modeling the squeeze aspect of straddles first.

I guess the question I have for you is simple?

How often do you squeeze straddled pots?

The reason imo that "straddled" pots are different is because villains are likely to INCREASE their limp range because straddled pots illicit more action and fun. Fish love straddled pots and they all want to limp and get a piece of it. Thus, this situation is DIFFERENT than the normal pots.

So basically, I think this is far from useless.

Similarly, if something I said above is wrong, I'm all ears.

quesuerte · 05-31-2012, 04:58 PM

So you are giving the limpers 100% range?

I appreciate the need to simplify problems but in this case I find the over simplification so great that the model can't tell you anything that logic wouldn't.

IE wide limping ranges and tight calling ranges make stealing profitable.

Sorry if that sounds over critical, that's not my intention. I realize you are trying to take this further and that you need to "walk before you run" so to speak. I was just surprised by your surprise at the results.

avramia · 05-31-2012, 05:01 PM

For me it's enough to know that straddle is not profitable.
Every decent poker player loses money only in BB and SB. Adding UTG as another losing spot is a very bad idea.
Regarding the math... just multiply ~2 times the loses from BB .

dgiharris · 05-31-2012, 07:32 PM

Quote:

Originally Posted by quesuerte

So you are giving the limpers 100% range?

I appreciate the need to simplify problems but in this case I find the over simplification so great that the model can't tell you anything that logic wouldn't.

IE wide limping ranges and tight calling ranges make stealing profitable.

Sorry if that sounds over critical, that's not my intention. I realize you are trying to take this further and that you need to "walk before you run" so to speak. I was just surprised by your surprise at the results.

Yes and no.

Of course it would be profitable, but what my models do is they identify the exact conditions from where it goes from profitable to unprofitable and that is something deductive logic cannot do.

Similarly, if you aren't surprised by my latest round of results then I don't know what to say, you are a genius. Those results say that if you are at a table with TWO wild players that the EV in squeezing 10bb is the same as squeezing 50bb. To me, at least, that is pretty surprising looking at those curves overlap like they did to the point of being almost identical.

Did you look at the last batch of results, I noticed you quoted one of my earlier models. Similarly, I intend to play around with Hero's ranges and table conditions. And again, yes, I have an intuitive feel for whether something is profitable or not. However, that intuitive feel can't tell me HOW profitable or what conditions result in the MOST profit.

So basically, I'm trying to quantify all those things that intuitive logic would deduce. And i'm sure before this is over I will uncover some more suprises.

dgiharris · 05-31-2012, 07:37 PM

Quote:

Originally Posted by avramia

For me it's enough to know that straddle is not profitable.
Every decent poker player loses money only in BB and SB. Adding UTG as another losing spot is a very bad idea.
Regarding the math... just multiply ~2 times the loses from BB .

No, its not so simple. UTG straddle cuts all effective stacks involved by half. So if the table is 100bb, if you straddle, you cut eff stacks to 50bb. Similarly, players are more likely to widen their limping range with a straddle than a normal pot. Then factor in that there are a certain breed of players that love squeezing straddles (I haven't modeled this yet but plan to) and then adjusting to squeezes in a straddled pot or anticipating a squeeze and limping with strong hands...

Basically, if you regard straddles as "nothing special" then you are wrong. THey alter the table dynamics and cut effective stacks in half.

Nichlemn · 05-31-2012, 08:40 PM

Quote:

Originally Posted by dgiharris

No, its not so simple. UTG straddle cuts all effective stacks involved by half. So if the table is 100bb, if you straddle, you cut eff stacks to 50bb. Similarly, players are more likely to widen their limping range with a straddle than a normal pot. Then factor in that there are a certain breed of players that love squeezing straddles (I haven't modeled this yet but plan to) and then adjusting to squeezes in a straddled pot or anticipating a squeeze and limping with strong hands...

Basically, if you regard straddles as "nothing special" then you are wrong. THey alter the table dynamics and cut effective stacks in half.

Any bet lowers effective stacks, that doesn't make random blind bets profitable. The only way straddling can be profitable is if a) it's constantly profitably for you to squeeze ATC from the BB b) people respond to straddles in weird, inconsistent ways.

quesuerte · 05-31-2012, 09:20 PM

What are the limping ranges?

dgiharris · 05-31-2012, 09:35 PM

Quote:

Originally Posted by quesuerte

What are the limping ranges?

RIght now, the limping ranges are irrelevant, so you can think of them as 100%.

The reason they are irrelevant is because the limp range has no correlation with the squeeze calling range. If there was a correlation between a limp range and a squeeze calling range, then the limp range would matter.

Another reason is that when you model something you have to isolate a variable, something has to be the "independant variable".

In future models, I can give limpers a range and then give them a squeeze calling range but in those models I would then have to "fix" the bet sizing for the model to be at all meaningful.

But so far, I'm fine fixing/isolating the number of limpers and then altering all the other "stuff" to see how everything impacts everything else.

Let me reverse your question.

Why would the limp range matter for the purposes of modeling the impact of squeezes???

Quote:

Originally Posted by Nichlemn

Any bet lowers effective stacks, that doesn't make random blind bets profitable. The only way straddling can be profitable is if a) it's constantly profitably for you to squeeze ATC from the BB b) people respond to straddles in weird, inconsistent ways.

Well, after the end of this thread, hopefully there will be more math to support or refute your assumptions one way or the other.

Similarly, I'd also hope to gain some new insights into the various situations straddled pots present (i.e. like squeezes) and then incorporate those learnings into my game.

I'm not trying to say Straddling is +EV, I'm just trying to model all the dynamics that take place in Straddled pots and identify the various +EV and -EV lines and their potential consequences/opportunities

quesuerte · 06-01-2012, 03:05 AM

Quote:

Originally Posted by dgiharris

RIght now, the limping ranges are irrelevant, so you can think of them as 100%.

The reason they are irrelevant is because the limp range has no correlation with the squeeze calling range. If there was a correlation between a limp range and a squeeze calling range, then the limp range would matter.

Another reason is that when you model something you have to isolate a variable, something has to be the "independant variable".

In future models, I can give limpers a range and then give them a squeeze calling range but in those models I would then have to "fix" the bet sizing for the model to be at all meaningful.

But so far, I'm fine fixing/isolating the number of limpers and then altering all the other "stuff" to see how everything impacts everything else.

Let me reverse your question.

Why would the limp range matter for the purposes of modeling the impact of squeezes???

Well, after the end of this thread, hopefully there will be more math to support or refute your assumptions one way or the other.

Similarly, I'd also hope to gain some new insights into the various situations straddled pots present (i.e. like squeezes) and then incorporate those learnings into my game.

I'm not trying to say Straddling is +EV, I'm just trying to model all the dynamics that take place in Straddled pots and identify the various +EV and -EV lines and their potential consequences/opportunities

Of course the limp range matters. You are calculating fold equity based on how often the limpers call. This is one of the reasons it is so unrealistic.

If villain is calling a "squeeze" with TT+ AQ+ (or whatever assumption you make) then your fold equity is a direct result of his limping range.

If he only limps 88+ AJ+ then you have virtually 0 FE. This is one of the reasons your results seem to indicate you can raise so profitably when in real life this is never the case.

dgiharris · 06-01-2012, 05:48 AM

Quote:

Originally Posted by quesuerte

Of course the limp range matters. You are calculating fold equity based on how often the limpers call. This is one of the reasons it is so unrealistic.

If villain is calling a "squeeze" with TT+ AQ+ (or whatever assumption you make) then your fold equity is a direct result of his limping range.

If he only limps 88+ AJ+ then you have virtually 0 FE. This is one of the reasons your results seem to indicate you can raise so profitably when in real life this is never the case.

If I were trying to do everything in one go using analytical equations then maybe you'd have a point. But i'm not. I'm empirically building the model one piece at a time with an independent variable set against various dependent variables. I will go through different independent variables and eventually I will incorporating limp ranges into whatever scenario I decide to model.

Look, lets say I want to model stealing the blinds from the BTN. Do I need the limp ranges of UTG thru CO??? Do I need the limp range of the SB and BB??? No, I only need to create a customized scenario assumming everyone folded and I'm on the BTN and then I only need the calling ranges of the SB and BB in relation to my raise. So in this scenario, why would I need the limp ranges of UTG thru CO??? I can just assume they fold because that is the scenario i'm trying to model.

I'm basically doing the same thing here so far...

Hmmm... I typed and retyped several posts but they were tl;dr

I list the assumptions and parameters of the scenario, if you don't see the utility of it then fair enough.

Basically, I don't think you understand what I'm doing, maybe its because you would approach it a different way.

quesuerte · 06-01-2012, 06:47 AM

We seem to be going in circles.

Would you accept that the ranges of the players who have already limped will effect there folding frequency?

V1 and V2 both call your squeeze with top 5% hands

V1 limps 100%
V2 limps 50%

Do you think they are both calling your squeeze with the same frequency.

I understand what you have done. You have shown how many random hands need to be in the pot before your chances of running into a monster negates the times you steal. This is true of all positions and not relevant to straddles.

DrTJO · 06-01-2012, 10:12 AM

Quote:

Originally Posted by quesuerte

We seem to be going in circles.

Would you accept that the ranges of the players who have already limped will effect there folding frequency?

Isn't there a fundamental difference between limping in a straddled-pot and limping in a non-straddled pot?

Intuitively, from experience, we suspect this to the case. For reasons relating to the social dimension of "gambling", players will more likely limp with a wider range simply because UTG has decided he's "donating" 2bb blind. There is an unspoken etiquette---social pressure---to likewise donate.

For example, a tight passive in MP may play J9o in a straddled pot, when normally these cards are instantly mucked. Of course being a tight passive, he's not playing 36o, because he doesn't really gamble like others. J9o has greater EV than normal as a starting hand, because the others are blind---it almost plays like QJ in the TP's mind.

However, the TP isn't going to call your squeeze with J9o, in the same way he wouldn't call your B/B raise with QJ. But say he limps in a straddled pot with AJ? Does this hand play more like AK on account of the straddle? Does this mean he might call your squeeze somewhat "lighter" than normal?

I'm not sure about this, but I sense the tight-passive might still muck AJ because he's MP in this scenario. But he may call with this hand if he were C/O or B/N.

MSchu18 · 06-01-2012, 11:59 AM

In to read this later... looks like good theory OP.

dgiharris · 06-01-2012, 05:33 PM

Quote:

Originally Posted by quesuerte

We seem to be going in circles.

Would you accept that the ranges of the players who have already limped will effect there folding frequency?

V1 and V2 both call your squeeze with top 5% hands

V1 limps 100%
V2 limps 50%

Do you think they are both calling your squeeze with the same frequency.

I understand what you have done. You have shown how many random hands need to be in the pot before your chances of running into a monster negates the times you steal. This is true of all positions and not relevant to straddles.

For some reason that I cannot possibly fathom, in your head you somehow link limps with folding frequency???

The two are mutually exclusive. Me limping 100% of the time has ABSOLUTELY 100% NOTHING to do with my fold frequency PERIOD!!!!

oh wait, I think I see what you are getting at....

If I limp 50% of the time that "must" mean my range is wide and therefore i'd fold to squeezes a higher percentage of the time to someone who limps 10% of the time which means he limps with a stronger range and thus will call a 3bet a higher percentage of the time???

Even though that is logical, that isn't necessarily true. What if I decide to fold 100% of all pocket pairs because that is how I play and I raise 100% of the time whenever I have a 2 (lucky deucese never lose right) and I limp 100% of the time when I have a 3. How would you account for that?

Well, my models can BECAUSE I INPUT THE RANGES DIRECTLY and let the chips fall where they may.

This graphic illustrates where I think we may be having our disagreeance. As near as I can tell, you are trying to link limping frequency with 3bet calling frequency by using deductive logic however that is not accurate as shown by the ranges I specified above.

Not to say that deductive logic couldn't be used to link limping and 3bet calling frequency. You "could" do it that way however, as i've shown, all it takes is ANY unconventional circumstances for that model to blow up. It would be a lot more accurate to just identify the limping, calling, 3betting, 4betting, etc ranges DIRECTLY.

Which is what i'm doing.