Does PIOsolver/equilibrium solvers take into account equity realization/equity denial in regards to effecting the EV of a certain hand combo, say we are on a earlier street?

So this is a 3! pot spot where i ran nash solver and it said we x-raise range/50% and never call etc.




---

The EV of x-raising OTF with our TP combos (being Jx) are equal to the EV of calling OTF, but the solver recommends us to x-raise only and i'm not sure if equity realization is taken into account when generating EV's?

Punter,

wonder if you might think about adding a feature?

I used multiple bet sizings often, and many times see the solver scatter a small percentage of one size throughout the solution.

Would be really nice I could "right click" the "Check" box, and the solver would show the frequencies of hands that were NOT checked?

It would be easy to see, and pretty intuitive i think.

Either that or perhaps we could get a box to check where all choices "Ticked" will display in the solver? So we could have a Tick box for any bet, and for check/fold and we could choose which of several we could display?

Not sure I am explaining myself well, but hope this is clear.

which

Tbh we are doing poker software project not poker strategy research. I have some intuitions because I solved a lot of trees but at this point there are a lot more people who solved way more trees than me and are way better players as well. It's a good idea to try to find them and seek their opinion. I can answer math/theory questions reliably but not so much ones which depend on judgement, especially that often on those many people disagree.

This is support thread for PIO, not other software, you should ask authors about their software.

I don't understand what you mean. Equity realization is quite a useless concept theoretically (it's just EV/EQ and we have both when we are solving) so taking it into account when solving doesn't make sense. What "equity denial" is I do not know.

You are pasting screens from some other software. This is a support thread for Pio, I don't know what you expect tbh.
Feel free to ask theoretical questions but I would appreciate if you could describe things in simpler terms. It's first time I see a term like "equity denial". Also please do not ask about software other than Pio in this thread.

So let's say it's 2% 48% 48%, you want to remove 2%, should it be 50%-50% now, or just 48%-48%?. I am just trying to understand what you mean here.

Quote:
I used multiple bet sizings often, and many times see the solver scatter a small percentage of one size throughout the solution.

Would be really nice I could "right click" the "Check" box, and the solver would show the frequencies of hands that were NOT checked?Quote:
So let's say it's 2% 48% 48%, you want to remove 2%, should it be 50%-50% now, or just 48%-48%?. I am just trying to understand what you mean here.


Well, I would like to choose which action to take, but in a format that might be easier.

For me, I think the easiest is to be able to "right click" on Check if I am OOP for example, and now the solver will rather than show the hands I check, instead show the hands I DO NOT check.

This way I do not have to keep adding various percentages together to get bet frequencies, or subtract Checking frequencies from 100%....

Think that a visual representation of hands I DO not check is pretty cool. So with a Fold also. Right click Fold, and the solver shows all the spots you do not fold,

Ex: if the solver shows 2% bet size small 75% bet size big, 23% check -- if I right click check it will show 77%

If the solver shows 15% call 45% raise small 3% raise big 37% fold, I could right click fold and the display shows the 63% of hands I continue with...(the ones I Don't fold)

And for further clarity I am not talking about the initial screen shot that shows all three choices in different colors, i am thinking of the grid AFTER you have clicked on each specific option which displays in "all yellow".

which

https://www.dropbox.com/s/6iftl4glxy3nk03/pio.png?dl=0

this is screen shot of what i would like, as a "non check"

and there really does not need to have any other stuff with it, since all the individual info is always available to access from the main screen.

Just a sum of the NON check/fold range in a visual sense...

which

PS hope this is clearer now

Yes, it's clear and a nice idea

Not going to quote all the pictures, but wp sir. I would say you could get this speed as well spending 2k$ maybe even a bit less. If anyone looking for making anything similar

(The RAM speed does not make much differents with pio, definitely not needed 1000w psu, and the gpu depends about gamings)

The new ryzen threadripper seems sooo beast!

Enojy!

Are there available more of these ready preflop hand ranges? (In addition to those that come already with PIO)

E.g. SB flat v EP/CO/BTN open? Would be cool if there was some public/semi-public database for PIOusers where you could upload/download these and do just minor tweaks to them for sims, and not having to build all from scratch.

I agree, I am considering getting one myself although I wonder if 2018 isn't going to bring even bigger CPUs at reasonable prices (getting greedy)

Not publicly as far as I know. Calculating those takes a lot of server time and money so it's understandable that not everyone is willing to share. For example our friend sells some here:
https://piocloud.******.com/

That being said our policy is that you can do w/e you want with the solutions you created yourself so if you want to start some solution sharing place you get get it going by uploading some

Hey,
in simple words (I've only spent a few dozen minutes playing with free version before), please explain me how big of a difference will be noticable between solving different spots on these 3 CPUs:
-i5-750 (yes, the one that came out in 2009)
-ryzen 5 1600
-ryzen 7 1700
I guess with piosolver basic it will be close to none since it uses up to 4 threads? If so, let's assume we're talking about pro version.
Thanks in advance for help!

It's 6 threads these days.

I have no idea about i5-750 because I don't have access to CPUs from that era.
As to Ryzens: Ryzen cores are about as fast as Haswell cores for Pio and speed is proportional (more or less) to numbre of cores * frequency they run on so that means that Ryzen 1700 gets 3Ghz * 8 = 24 and Ryzen 1600 gets 6 cores * 3.2Ghz = 19.2 so without overclocking 1700 will be about 25% faster.

Making a wild guess about i5-750: it has 4 cores clocked at 3.4Ghz which makes it 13.6, the cores are maybe 50% slower (cause old CPU, no hyprethreading which is worth about 15% and I am guessing). With those assumptions Ryzen 1700 is going to be 3x-3.5x faster. That might be significantly off though. You would need Pio benchmark time for Kaby Lake CPU to have a better idea.

1) Are there any publicly available common ranges available for download? For example LP 3B vs EP. 3B button vs MP, Full ring, etc? Basically I am looking for more than the dozen or so that comes pre-installed and I rather not spend time trying to create these if they are already available somewhere.

2) I remember coming across a list of pre-defined flops that approximate/represent common spots. Where can I get this list?

Thanks

See the bottom of this post:
https://forumserver.twoplustwo.com/s...postcount=3134

You likely mean a subset of flops which if used to representing the whole game gives good EV approximations. Those can be found here:

https://www.piosolver.com/blogs/news/piosolver-1-9-2 (point 3)

and the ones we now think are the best are listed here:
https://pastebin.com/AtyU84qC

Hi,

Let's say Piosolver was playing 6max 100bb with standard Preflop 6max ranges (Preflop OPR : 20% UTG & MP, 25% CO, 45% BTN, 50% SB, etc...). With no limp strategy (OPR sizing between 2.2 and 3x).

Of course Pio does not deal with multiway pots, but either you can take these off the experiment or consider Pio would solve them as a HU spot, whatever.

I wonder what would be an estimate of Pio's stats (frequencies) for :
-Flop CB (eventually distinguish SRP and 3BP)
-Flop Raise IP
-Flop Check Raise
-Flop Donk
-Flop bet vs Missed CB
-Turn CB
-River CB

Please, don't insta throw my question away because Pio is not made for 6max. Even if your answer is a vague estimate, I am very very curious of what these numbers might be.

It might also be helpful when trying to detect GTO bots.

ty

There are two tools in Pio created to answer those questions specifically:

1)Multifile aggregation reports:

With those you can solve a chosen spot (for example BTN vs BB) on many flops (maybe choosing from our subsets created to represent the whole game) and then choose a spot to run "aggregation" in. This will produce EVs/EQs and strategies for hands on all flopsas well as an average of those across all flops.
There is a bit more about it here:

https://www.youtube.com/watch?v=5Ozjfnatwms (question number 5)

2)Average line frequencies displayed when you browse the tree

You need a preflop solver to run the preflop soluton and then when you browse it you will see average line frequencies in the right upper corner (it's possible to switch it on/off in Tools->Configuration->Data presentation other for performance reasons).

I calculated one spot like that for you using ~43% opening range vs BB. I've included two flop sizes (35% and 65%) and one turn/river size (75%).
Donks OOP weren't included (in single raised pot) because those are never chosen anyway. I've used 74 flop subset from our list.
The tree is 75GB big and took about 4 hours to solve to 2bb/100 on my 16 core Xeon server.

Here is what the solver shows for SRP situation:

https://gyazo.com/fb105be52ff4bdc37b54c574b4582379

Total cbet% is 62.46% here. Notice that I've chosen one of the flops (A52r) but it doesn't matter as the frequencies on the right are across runouts (that is all flops/turn cards/river cards if applicable).

Isn't applicable for SRP. In 3bet pots it's usually 0%

Depends what sizing was used on the flop. In our tree it's 16.8% vs smaller size:
https://gyazo.com/340fe5ccd5218534d96a3e45868a594c

and 8.79% against bigger size:
https://gyazo.com/0d115a64db8afcad63f38d1f7d04de65

More or less 0%, that's why it's safe to remove from option to make the tree smaller/easier to solve.
Here you go:
https://gyazo.com/09345ef9aa6812e61e21969d32af9917

https://gyazo.com/e4cb6d6b80236bdd11be7952c4de8d0a

https://gyazo.com/6b7ab7d5743a7eebbafd53be3054c9b5

I will add one more with a short story. I remember when I started playing poker more than 10 years ago (notice it was before trackers, HUDs, solvers and even decent equity calculators) some people discovered "a leak" that most players have: they fold 60%+ if it goes CB-call/check-check/bet river. So it "made sense" call CB very light and then just bluff every river. It of course makes sense why one would fold a lot in this line (it costs money to get there so can fold more than pot odds suggest and the opponent still can't profit automatically) but it wasn't obvious back then (1 - alpha argument and all that).
It's nice to see that the solver confirms though:

If cbet was big (65%):
https://gyazo.com/9422ddeb9638b6673ebc2d2d437cc012

If the cbet was small:
https://gyazo.com/449b8d2c0d44f97a8b143116d3711133

I hope I provided something interesting. They may change with bet sizings/ranges used so it's hard to answer in general. One way or another that's one of the topics we wanted our users to be able to research themselves that's why we added the tools to do it

I have an old notebook. It takes very long time, like 15 minutes, to solve single raised pots. Especially wide range vs wide range. Can somebody recommend a computer setup for it to take like 2 minutes? CPU, GPU if important, RAM etc.

After a small research in the FAQ and Google, I couldn't find how many flops should I use to have a decent result to find the preflop of BTN vs BB

Thank you very much for your answer.

I have checked the report.csv and indeed, it gives us the average frequency in the given spot, like for example BTN vs BB. But only in this given spot of course.

In fact my question was for global stats. (I mean not only BTN vs BB or SB vs BB).

In other words, if Piosolver was a 6max player and I was looking at his global stats in my tracker, what do you think these stats would be for :
-Flop CB (eventually distinguish SRP and 3BP)
-Flop Raise IP
-Flop Check Raise
-Flop Donk
-Flop bet vs Missed CB
-Turn CB
-River CB

I've written a bit about hardware here:
https://www.piosolver.com/pages/faq#hardware

It's hard to answer because it's a judgement call. I personally think you don't need that many (~50 for decent overview, no need to go over ~75). I've talked to very smart people who think my standards are too high (they use fewer flops) and to those who think you need much more.
Make sure to use flops from this list:
https://pastebin.com/AtyU84qC

I am usually using 74 flop subset myself.

I really have no idea. Some of those stats make very little sense to begin with. For example "total cbet" contains both SB vs BB and BTN vs BB which makes it more or less useless (unless someone has a huge leak like betting everything).
You can try approximating that by calculating averages for various spots and then averaging them (weighting by probability of the spot occurring).

I would focus on stats in similar spots though, for example CB IP in SRP instead of just CB overall.

Is it a bug?

I just ran a sim with a very basic range :


First question :

I don't understand the difference between these 2 stats in red


Second question :

Hero has only 2 hands here AA and KK :



Here are the stats for raising :
4.09
4.09
4.09
4.09
4.09
4.09
9.14
9.76
9.77
9.87
9.93
9.07

So the average calculation gives 6.84% : it's different from the 2 red statistics. How could this be?

Couple questions about PLOCalc:

1. Does it accept PPT range syntax?
2. I can't have it installed on two computers at once or I just can't use both installations simultaneously?

No, it isn't

One shows strategies for action at this point. The other shows averaged frequencies across all runouts (all turns/rivers following this line).
They could be different on a flop as well. The reason for it is a bit complicated:

There are two ways of calculating frequencies:

1)what a player thinks they are doing
2)what really happens

To illustrate the difference between those two imagine that you have AA/KK in your range and your opponent has only AQ. The board is A7522 and you are calling a bet.
You call with AA and fold with KK, how often do you call?

According to 1) it's 50% because you call half of our combos. According to 2) you are calling only about 14.2% because when your opponent has an ace you get AA way less often (1combo vs 6combos of KK).

This effect is not as strong with bigger ranges but it's still there and it's surely there in your example. We show 1) in big rectangles but 2) in the overall frequencies. This is because we think 1) is more useful immediate information but calculating line frequencies only makes sense as 2).

You can change the settings for that in Tools->Configuration->Data presentation on 13x13 grid->show total strategy as real frequency.

Simple average isn't very interesting because it doesn't take into account how often a given hand is in range at this point. You should weight them somehow. Two ways that make sense is:

1)use current weights of a hand (for example if you had AA 100% on the flop and called 50%, raises 50% you now have AA:50%).
2)weighting by probability of a hand occurring (similar to 2 in the explanation above), this requires taking into account card removal effect of the opponents range.

On your screenshot it's already clear it's not the beginning of the hand (total gets rounded to 0.0 combos) so your average should include weighting.

In the future please attach the config when asking about details in the solutions. It's easier to have the whole tree solved instead making a lot of assumptions looking at selected screenshot. The way to share the config is:

1)use "copy to clipboard" button in the treebuilding and calculation tab
2)paste the text:
-either to pastebin.com and link here
-here but enclose it in [ code ] [ /code ] tabs (without inside spaces)

It doesn't. We will be accepting he simplified version of PPT syntax (the one Monker Solver uses for example), the one which contains combos and weights only.
In general in PLO translating postflop ranges is hopeless (because there are 200k+ hand combos and too many ways to construct a language). If a tool author wants their format to be importable or to export other formats they really should implement a simplistic format for preflop ranges (16432 combos only) preferably containing 16432 numbers in a given order or (less preferably because it takes more space) combo:weight.
We will be accepting (it's currently in dev) PPT format like that.

We will be extending it to 2 activations anyway (today or tomorrow, all the customers up to now will get that as well of course) so it will be 2 computers at the same time.

Hey punter, what is the minimum screen resolution to run piosolver is it 1920x1080 or...?

Thanks.

It should run at lower resolution without problems although you may have problems seeing some information (especially average strategies/ev for hand groups in 13x13 grid). Again, it's the best to test with a free version. The window is resizeable so if you can't see stuff make it bigger/fullscreen and see if it appears.

Very novice question here, but my brain is not working.

On the flop the pot is 5BB, OOP must check and my IP cbet size is 2BB, if I do bet then he can raise by betting 6BB.

Now some EV's are like 10.98 for the nuts, how do they get this? Im bet 2BB to win 5BB how can my expected value be nearly 11BB????

Say on the flop I have the nuts and have say 90% chance to win the hand

then

EV = 0.9*7 + 0.1*(-2)

which is definitely less than 11BB?