Hey I had a random question for support or anyone else that could answer.

I have Pio edge and have been using a computer with 128gb ram to make preflop trees. I decided that I would spring for a better machine with 512gb ram to get better accuracy on more realistic trees.

I am using a website that designs workstations and allows you to select the parts. I was wondering what kind of an impact the CPU would have?

I have options from:
Six-Core Intel® Xeon® Processor E5-2603 v4 1.70GHz 15MB Cache (85W)
All the way to:
Twenty-Two-Core Intel® Xeon® Processor E5-2699A v4 2.40GHz 55MB Cache (145W)
With a bunch of steps in between. The price jumps are fairly significant so I wanted to see what the difference in the speed of the simulation time would be.

Thx in advance for any advice.

Is there anyway of copying a range from the browser in PIO?

I mean copying it for flopzilla or for another tree build?

Is there anyway of rounding these ranges in the preflop browser?

I get the following error message:

https://gyazo.com/8c817d0bd113d7fb4da1565ff2104c16

I want to round them so I can export to CREV or Flopzilla. You can export / copy ranges that are in formats CREV or Flopzilla accept (as if you don't know lol) but I need to do that with these ranges instead of manually typing them into these external programs.

Is there a solution if you are a PIO Pro user and only connecting to the preflop browser with already built preflop solutions?

Every time I hear trees bigger than 128GB are needed I am getting suspicious. Usually it's caused by doing counterproductive things like adding bet sizes in postflop play (that has 0 influence on preflop ranges in all the cases I've seen) or otherwise causing the size of postflop trees to blow up.
I am not saying it's a bad idea to run bigger trees but it's good to remember that the goal should be to get good preflop ranges and for that postflop play needs to realistic but may be quite simplistic.

As long as they are modern Intel CPUs it's more or less proportional to number of physical cores multiplied by frequency they run on (not turbo one, the base one). There is slight performance penalty if you get really a lot of cores but Pio scales pretty well for realistic setups (last time I checked it was x15.2 for 16cores and x22.5 for 24 cores in comparison to 1 thread performance).

Make sure to check if those are single or dualbox setups (dualbox setup has 2 CPUs so the number of cores doubles). Anyway, multiply number of cores by base frequency and the speed is more or less proportional to that number.
Make sure to shop around, there are many server providers and you may be able to find some good deals.

You can copy it by right clicking on 13x13 area (you can choose if you want to copy a stategy, range at given point, EVs etc.). I am not sure if the tools you mentioned accept our format - I doubt it. We do accept their format though (so you can paste the range from there to our range selector and it will work).

There is no way to round the strategies in the browser as of now unless you still have a whole tree so the solver can round them and calculate new EVs. You can also generate preflop charts (there is an option to round things there) if it's a preflop solution.

Make sure that:
-you are using 1.9.2 solver (it says at the bottom: connected to...)
-you are using the newest viewer, here:

https://www.dropbox.com/s/gesnfhgrbo...iewer.exe?dl=0 (download and substitute the file in your Pio folder).

This is a small bugfix from few days ago which shouldn't influence the exact thing you are mentioning but it's always best to have the newest version.

Tbh I don't know, last time I've heard Flopzilla was accepting 5 weights or something like that and then it was using quite a silly format as well. My advice is to nag the author of those tools to improve on the situation and/or use our range explorer (which to be fair is quite simplistic as of now).

I am not sure what you mean here, you can read already done preflop solution with the preflop-browser which is shipped with basic/pro version - yes. I probably misunderstood your question so feel free to elaborate.

I would like to somehow convert .cfr to preflop charts... .is it possible ?

Flopzilla doesn't take Pio ranges, but Power Equilab does, so the free version of Equilab may also.

Thanks for feedback. Is there a vid explaining about the preflop chart generation?

So I generate preflop charts from current solution with an option to round my strategy before the charts are sent to the output directory. What do I do with them text files,? How can I view them in Flopzilla or CREV? Can I view them in PIO OOP or IP ranges when building a new tree? I tried copy and paste and that does not work. Not sure how I am meant to convert them ranges and why there are so many text files for one conversion?

Any feedback on this appreciated.

I get it now. I have found a way to do what I want. I could not delete previous question so please ignore for now.

Cheers, gl.

I was thinking of buying a new computer with one of these AMD Ryzen cpus. I keep reading about how these 8 core AMD's are actually decent this time. I don't know much about computers. Is there anything you can tell from the benchmarks that would make these less efficient for pio use?

It looks promising but we simply don't know. We need someone to run Pio benchmark on it

Hello there,

when I want to the file this is what I got:



What am I supposed to do?

I already tried using connecting to "preflop-browser" and also to "PioSOLVER-pro19". None of it worked.

Thanks for your assistance.

It looks like the save you have is not a valid save then. Are you sure it's Pio solution? (it should end in .cfr). If it's a small save feel free to upload it somewhere and email us the link so I will take a look but it looks like you are trying to open something which is not a Pio save.

piotr wrote me this on the skype group about amd ryzen new cpu

"it remains to be seen; AVX will probably be worse than on Intel but that's shouldn't be that bad
it might end up being very good or quite bad, we need to wait "

i found this article about that, i don't know if it can be useful

First, y-cruncher is a program that solves digits of pi, as well as some other mathematical constants. It's one of the few applications that can really stress a processor's AVX instruction set. Intel created AVX, and its CPUs implement additional 512-bit AVX calculations that give it an advantage. In this particular case, the 6900K beats the 1800X by 50 percent. But while that might look horrible, there aren't a lot of applications, particularly consumer applications, that leverage AVX in this fashion
Intel does retain two slight advantages relative to Zen: first, the slight memory-op advantage above; second, Skylake can execute two simple 256-bit AVX operations of the same type per cycle, whereas Zen can only do one (dividing it into two 128-bit halves). So Intel might still be able to win, by a substantial margin, on some carefully-selected benchmarks.

http://www.pcgamer.com/the-amd-ryzen-7-review/4/

what is that mean? is it a disaster for pio and so the intel 6900k is way better, or even 6850k (which is 6core instead of 8)?

Yes, Ryzen looks promising but in the past most benchmarks were more charitable for AMD CPUs than performance for the solver. We can only be sure once we see Pio benchmark although my guess is that it's going to be very decent.

Question 1: Is there any way to export the full strategy for every hand, every turn and river for a given flop? I know that the excel file will be huge.

Question 2: Is it possible to open and read the .cfr files in other programs than PIO?

There isn't. The file would in fact be huge, significantly bigger than .cfr save.

No and it's a very bad idea to attempt that. While our save format isn't obfuscated in any way it contains only information necessary to recreate data shown. That means that to make use of it you would need to re-create significant part of solver's logic. The format is going to change in the future as well.

The way to work with the solver (and the saves) from other programs/tools is to use the text interface we provide, here:
https://cdn.shopify.com/s/files/1/07...68544248083387

You get access to all of the solver's functionality this way. There is a way to connect to console program using standard input/output from any programming language. Our PioViewer is one program like that - it's not privileged in any way so you can write your own using the very same communication protocol.

Feature request: In range explorer, option to see strategy in 13x13 grid so that you can easily see strategy only for selected types of hands (e.g. if you only want to see strategy for all types of straight draws)

Yes, this is a good idea. You can kinda get it by looking at those rectangles in the range explorer:

https://gyazo.com/450a17d9833a44a35d86d2884cb789b1

For example if you want to see how a flush draw plays, you can do it like here:
https://gyazo.com/b12ceb9f6a0bbaf727f535ea6702ada5

(flush draw calls 98.99% in this example).

You can also click on those rectangles to see betting/calling/folding range. I realize it's not perfect but you can get the information you want this way.

when comparing 2 strategies, how do you calculate the bb/100 difference, or is % of pot a better metric? with either bb/100 or % of pot, how much would be considered a big difference in strategy? or at least big enough to where one strat is better than the other.

You calculate difference in chips per hand (the way the solver displays the EVs) and then convert it to bb/100. For example if you play 5/10 game and the difference in EV in the solutions is 0.159 chips per hand then that would be 0.0159bb per hand which is 1.59bb/100.

This is hard to say. If you play 5/10 online with rake capped at 3$ then the rake is 15bb/100 for both players in postflop spots (30bb for both players). This puts in perspective how small differences like 1.5bb/100 are. I would say that anything below 1bb/100 basically doesn't matter, above that is a judgement call which I am not comfortable answering (there are many views on that from players better than me).

if 2 strategies give evs of 30 and 31

at 5/10, this difference is .1bb/hand, equaling 10bb/100?

Yes, that should be simple arithmetic

Hello,
I'd like to know in layman's terms how Piosolver derived it's subsets of flops. I found the article here, but couldn't follow what they did when explaining
Also if the method they used is independent of NLHE? Could i use the same subset of flops to study for PLO?

Hey guys, is there any chance to exclude turn / river cards from the solution if we only define the flop (or turn)?

Like if we saw a third player open-folding so we can treat his holecards as dead cards?

It's difficult to describe in simpler terms than the quoted passage. In short we tried to find the best subsets to approximate EV (calculated on all 1755 flops) and equity (of various ranges vs each other) in the full game. The algorithm used, while simple is not something that makes much sense for someone not familiar with this kind of problems.

Those subsets are really not the best for studying. Their sole purpose is to approximate EV and EQ of the full game. Finding a subset which is good for studying (various board textures to see patterns on them) is a different problem more suitable for coaches/experienced players than for programmers.

We don't know if those subsets would work well for approximating EV/EQ for PLO, my intuition is that they would be decent but this is quite a wild guess at this point.

Maybe although this is not that simple as for example isomorphism code would need re-writing if for example only 2 deuces are possible instead of all four. Preflop tables would need recalculating as well which is quite time consuming.