Hi, I want to test if 1 betsize for flop its better to 2betsizes on limp pot, i did my first tree with 1 size and then i did with 2 sizes and the results were the same EV and MES, so how can i know which strategy is better ?

It's very often the case that adding additional sizes don't change EV much or at all. In that case both strategies are as good. Usually you will get a (usually small) difference though if you add strategically important options (like turn or river overbets for example).

I think I figured out but I would appreciate a confirmation: is it that the vertical axis is representing the % of time that a given combo reached that node (blue in my picture) and that the orizontal axis is giving the frequency with which we should take a given color-coded action? (red in my picture)
https://drive.google.com/open?id=13N...fB4bRMKGZxem2u

This is true for the main view. Let me go through some options of what you can display:

1)Strategy:
The numbers as well as graphical representation describe frequencies every action is taken with

2)EV:
Same but EV of a hand at this point is displayed

3)Range:
How often we have given hand in given spot. If a hand starts with 1.0 weight an then bets 50% of the time, it will be in range with weight 0.5

4)Strategy+EV:
The graphical representation (X axis) is for strategy and the numbers show EV

Independently of that there is a checkbox which says: "square size proportional to weight" this checkbox makes it so Y axis represents range, it's here:

https://gyazo.com/615bbfd8d8fc4694b07e4ac9c5c477a7

On your screenshot it looks like you are using preflop chart tools for postflop range which is not going to work (or something else didn't work properly as it gives an error about overselected combos, if you are using that for preflop ranges and getting errors, please report to support@piosolver.com)
Your description would be correct for the main 13x13 view though.

I don't fully get how does MES work.
Lets say we build a tree. If both players play balanced, there should be no exploiting possible ?
If we lock some nodes and run again, shouldn't new result be equal with MES ?

I feel that I lack some theoretical understanding here ...

Yes, in perfect equilibrium EV of MES is the same as EV of the strategies. Sadly nailing exact equilibrium is not feasible (for anything but the simplest of trees) so there is some exploitability. This is defined by how much MES would win vs current strategy per hand.

Yes, the way we calculate MES is to respect node locks (so max exploitive strategy has to follow strategies locked and can only use the max EV line in not locked nodes), again though: the solution is not perfect so there is some exploitability.

What is the top end of hardware requirements? E.g. 6 bet sizes, 3 raise sizes, 6 max ranges, how much RAM is required and about how large will the tree be and what kind of processing speed would be needed for a timely computation? Thanks.

Also could you give any guidelines about running PIO on cloud ?

Which parameters would be best ? AWS, Google Cloud or monthly server (not $/h like AWS ) ?

Faster for scripting is the new version of Pio or the old one?

Those are huge trees, you are unlikely to be able to assemble a machine which solves those at all (and surely not within reasonable time). You can use the free version to estimate size of various trees. I would do that for you but there are so many variables (exact sizings, cap used, stacks, starting pot, all-in threshold etc.) that I am unlikely to input the kind of tree you have in mind.
Dedicated servers are more convenient and cheaper. If you want some experimentation then cloud instances are ok as well although they are not really suited for our licensing system as you may get different hardware between VM invocations (which triggers our licensing system). As basic/pro licenses are in principle per machine (moving allow within reasonable limits) this causes problems.
You will always be able to get your license back but most cloud instances aren't a viable long term solution in case of basic/pro licenses and a bit of pain with the edge license.

As to hardware: you need enough RAM to be able to build trees you want, you can estimate the size using estimate button in treebuilding and calculation tab.
As to CPUs: as long as they are modern Intel (or Ryzen generation AMD) CPUs the speed is more or less proportional to number of cores multiplied by based frequency they run on (so 16 cores total @ 2.6Ghz gives a score of 16 * 2.6). You can easily compare the performance using this metric. With some multi core VPS insteances (cloud providers) the solver won't be able to use all the cores though (reasons unknown for now, it's a very rare problem).

The new one has some bugs fixed and performance improvement for some trees/hardware setups although sadly some trees converge less reliably due to floating number accuracy issues. I am improving the situation but right now there are some trees which just don't converge as well to very good accuracies (if you are using 0.3% or more you are unlikely to ever be affected). If you find such tree please send it to us.

Hi, I have a question not related to PIOSolver but to Jesolver.

Is it still active? I mailed support about a week ago about buying it and I still haven't got a response. Can someone who bought a license explain to me what they did to buy one? Thanks in advance

Sadly Oskari's interest in the project comes and goes these days. Jesolver is fast and stable but you are mostly on your own when it comes to support. If you want it, keep trying to contact him

dumb question but if you run PIO for long time, (exploitable 1% -> 0.1%), does the .cfr file grow?
or does it just adjust the numbers.
is the final file size created during first calculation?

It doesn't. We don't allocate any new memory while solving.

Yes. The size is determined once the tree is build. The exception to this is node-locking (there is more memory used for every locked node).

Where can i get info about piosolver spin and go preflop charts?

Anyone running PIO pro on a virtual server? Does it work on a VPS?

Would it use up the license by installing it on one?

The site states "Pro and basic versions are not suitable for cloud instances (VPS) but you can run them on dedicated servers where you get a dedicated machine."

^ After searching the thread it looks like I need a dedicated server. Punter, could you please point me to a server with reasonable specs (that can run pre and postflop)? I'm not too familiar with computer hardware..

Here's a list, I don't know which is good value to run pio pro version:
https://www.ovh.ie/dedicated_servers/all_servers.xml

Can the free cloud solutions be viewed on Piosolver Free?


Sent from my iPhone using Tapatalk

If you mean PioCloud solution I made an announcement here:
https://forumserver.twoplustwo.com/s...postcount=3635

That's all the information that is available about the packs.

VPS are not suited for our basic/pro licenses because you get a different machine at every invocation. You can try it 1-2 times but sooner or later you will run out of activations. Dedicated servers are better for the purpose.

The information is accurate.
You won't "lose" your license no matter what (unless we determine you are intentionally and repeatedly breaching the license terms which never happened so far) as we will restore it for you if it gets stuck somewhere but this is not a long term way to use the Pio licenses.

3 most popular providers used by our customers are:
-online.net
-contabo
-ovh

the order is random. We don't have resources to support/analyze offerings of server providers as they are constantly changing. It's a good idea to consult our Skype group about it, there is even one guy who will help you with setting up a server on Contabo.
In general solver's speed is proportional to number of physical cores multiplied by base frequency they run on so you can easily compare performance/price ratio you are getting.

They can't.
The information in the saves is very minimal, most of it is recalculated on the fly and that requires full engine implementation.

Hi there, I have had an Edge subscription for a very long time but only now have a good enough computer (and never rented a server), so, some very basic questions regarding preflop, which I could not find answered in the introductory video or the FAQ's

1) What accuracy would be deemed decent for preflop? For postflop, I understand that it is recommended to solve until 0.25% - 0.5% of the pot. My guess would be that most users solve to less accuracy preflop?! Does anyone know, for example, which accuracy was used with PioCloud?

2) Does it make any difference in a situation like SB vs BB if the initial ranges are "empty", like so

or 100%? I would presume it doesn't matter; I have only solved one situation yet and started with 0% ranges, and only realized several hours into solving that in your video you start with 100% ranges for both players.

3) I solved the example tree from the video with (only) 10 flops.

SB raising J6s and J4s, but not J5s seems strange - am I correct that the reason for that should be the small amount of flops which somehow doesn't include enough 5's, or does it look like I made a mistake?

4) There are always two files for each preflop subset, for instance with 34 flops, the files 4_THREE_EQ_1__288.txt and 72_THREE_EQ_2__883.txt In the tutorial video, it would appear as if the narrator (Piotr or Kuba) chose one of the two files at random - are they just two "equally good" abstractions for each given number of subsets, and I don't need to care which one I happen to pick?

2bb-3bb/100 is good enough. Most of the PioCloud solution are solved to better accuracy than that (maybe all, I am not sure).

[/QUOTE]

Yes, there should be full for most cases (the exception is if you solve something like BTN steal with defined range vs BB).


10 flops is really not a lot. The subset has a lot of biases.

Since 1.10 version we ship only recommended flops (the best out of the ones we shipped before). You can get them by running the installer/updater from here:

https://www.piosolver.com/blogs/news/piosolver-1-10

Since the very beginning, I'm still floored by the speed of your responses

Still, when I update with the updater, even if I do "Force update", it does not appear that the preflop_subsets folder was changed: still 2 files in each subfolder ...

Solved by PM.
The updater won't overwrite the preflop_subsets folder (as we don't want to overwrite content potentially generated by the user if they have their own subsets). The solution is to remove that folder and then run the updater again.

Feature request:
I would like to see which lines each hand takes more often like: BBB,BXB,XXB,XC-X-B,BBX,XBB, XC-XC-XR,...
I guess the first 3 lines any hand takes more often would be more than enough.

Yeah, this is a good idea. At least for the lines where opponents always checks/call it will be easy to compare.