I've answered it in the middle of the previous post - it's not possible right now

That's not correct. It supports CPUs with more cores if you have them but it doesn't matter on normal ones. As the basic version is compiled to use 6 threads these days it doesn't matter on laptops or desktops with quad core CPUs (it only matter on hexa core and bigger CPUs).[/QUOTE]

My processor only has 4 cores and can do only 4 threads at a time (i5-4460). So I would not be able to use even Basic version? Or it's fine and can still use Basic or Pro version? Please help, still don't get it

I heard there is a skype group, is it possible to join?

Why does PioSOLVER tell me that I only have 9.8 GB of RAM out of my 32 available when PIO SOLVER is the only thing I have open?

https://gyazo.com/c3010abb65f468edb1ca9d7dbe2a6902

Is there a way of bypassing the first memory check? I have a good processor and 32GB ram, I have ran sims with this size before and am sure it would be fine once it kicked in.

Thanks in advance.

Hello,

I just bought pro version and run my first script, but wanted accuracy doesn't work. I put in the value like on the FAQ video "0,35", but its calculating very fast one flop after another and every tree accuracy is around 3% instead of 0,35%. How to fix this?

Thanks in advance!

Settings: https://gyazo.com/4997991a71fbe531703f7536eda2a1ca
Running script: https://gyazo.com/c521c1a8e5b2108750f04a21c160b230

You don't need to worry, it will work with both basic and pro versions.

Yes, pm here or email support@piosolver.com to get a link.

This is what Windows is reporting so likely something is taking your RAM. Maybe you have another Pio open and a big tree built there.

Yes but make sure to check if you really have the RAM available as builiding a big tree otherwise will result in swap file usage and that will freeze your computer and you will likely need a reset. To check what is going on open task manager and sort processes by RAM usage.

To bypass the memory check:
-ctrl+b to bring up arbitrary solver command
-type:
ignore_mem_check on
and hit enter
-build your tree

I strongly recommend against using that option.

We use a dot "." as a decimal separator, not a comma.

Notice that the exploitability in the script window is given in chips so if you put 0.35 there and the starting pot is 1050 (like in your example) it will calculate till 3.675 chips per hand exploitability.

"Yes but make sure to check if you really have the RAM available as builiding a big tree otherwise will result in swap file usage and that will freeze your computer and you will likely need a reset. To check what is going on open task manager and sort processes by RAM usage.

To bypass the memory check:
-ctrl+b to bring up arbitrary solver command
-type:
ignore_mem_check on
and hit enter
-build your tree"

Does this work for scripts with large batch of flops? I have the memory.

I am running a batch now and I have noticed stopped, timeout reached and then noticed I have it to stop calculation when 600 seconds have passed. This seems like to me that my data is going to be flawed now? Is that correct? Should I get rid of this option of 600 seconds?

When generating a script there is an option to ignore the memory check, here:
https://gyazo.com/2192cc125f32e28aa03f24705d50339f

It's a good idea to combine accuracy and some sort of timeout for the rare cases where solver can't reach required accuracy on a difficult tree. Maybe you want to make it longer than 10 minutes though if you are solving bigger trees which take longer to reach desired accuracy.

[QUOTE=punter11235;52135367]When generating a script there is an option to ignore the memory check, here:
https://gyazo.com/2192cc125f32e28aa03f24705d50339f

But I thought this was after the first tree is built? I have tried to ignore the mem check and it never works. I know I have 32gb ram and an octo core ryzen processor so I should be fine but every time I build (just build not run) a big tree, PIO takes almost half my memory usage before I run it so then when I try and run a batch job via a script I'm told I don't have them memory and it does not run.

Unless 32GB ram is not enough to run tress 15-20GB?

Thanks for your time.

Hi Punter, could you help me with things? What is a good accurancy for calculating preflop? Does it depend on the "stack to pot ratio"?
Theoretically, if I calculate a pre-flop with 10x stacks with certain ranges and put a 28% bet on the flop, turn and river then calculate a new preflop (always with stacks 10x) but with 50% size on flop turn and river, Does equilibrium change or remain equal?
I noticed that in the "preflop_subset" folders there are 2 types of files, for example in the 184 folder there are 72_THREE_EQ_2__529 and 72_THREE_EQ_2__1072, what is the difference? Is one better than the other or is the same thing? Thanks in advance, I wish you a good day.

I keep getting this error when trying to activate https://gyazo.com/326f5acea5826a533064d80d92a08f82

I downloaded 7-zip but when I right click there are no options to unpack or extract all. I downloaded file viewer as well and opened (and I think extracted and or archived) the file the turboactivate.dat but that didn't help either.

I also get this error when trying to extract https://gyazo.com/0db21d1a0ebb38940203778998e6ad4f

I should also mention that I got PIO to work yesterday but when I woke up today I need to reactivate and it wont activate anymore

I have a guestion about "Remove lines" feature.
When i try to solve a board on turn (i put a 4 card board) and want to remove OOP players ability to check-raise turn, and also remove check-raise from river no matter what the turn line was.

First I put in:
Check, Bet, Raise
This works! and remove checkraise on turn.

When I add to remove river check raise for oop started with putting in this
Check, Check, Check, Bet, Raise
This doesnt remove checkraise on river, infact it seem cause the IP player only being able to check behind on river.

Is there something i am missing or is this a bug?

I hope that mentioned "adjust payoffs at end nodes" feature (for ICM/knockout/skill adjustment) is on the way for next update.

Is there an internal limit to the length of lines? I would not actually make a tree like this however I noticed the following (with effective stacks of 500):

add_line 0 0 0 20 50 95 163 265 265 265 265
ERROR: add_line Specified line is too long. Some actions were not included in the tree.

Am I mistaken or is this technically a valid line? Flop: c:c Turn: c:b20:b50:b95:b163:b265:c River: c:c

If I do something shorter like add_line: 0 0 0 20 50 95 95 95 95 that works.

In this spot i am wondering how can IP call still have 6 combos after a check shove from oop on the river if oop only xs nut blocker + nuts.

XS OOP range


CALL IP range

Hello Punter,

I have ran some simulations via scripts and have made some aggregation reports.

My reports looks like this:
https://gyazo.com/dbe204b61dc9a0f074257fbd922c8391

I have 2 questions about the numbers highlighted in red;

1. IP and OOP EV is shown in 5 and 6 digit numbers, what's the difference between those, and why use different numbers?

2. IP EQR is shown is 4, 5 and 6 digit numbers, why?

Compared to OOP EQR it looks like it has to be a percentage.
Can you explain a bit more about the Equity realisation factor?

As far as I understand this is a percentage(?) that's based on how good our range is compared to our position. If this is a percentage, then IP's percentage is always above 100(%). Is this because IP has the advantage to respond to OOP's action and can take a free card (if checked) or raise and push OOP of his hand, and in this way IP can add equity advantage to his realisation percentage?

Thanks in advance!

.

I pasted instructions above. What happened when you try following the steps?
You can also add:
ignore_mem_check on

at the top of the script

Pio just reports what Windows tells it. Another thing to check would be if you are not running some crippled version of Windows (like home basic), some of them have limits on how much RAM can be used (if you go to control panel->system->system you will see that information).

There are 3 possible explanation for that:

1)The tree is around 15GB
2)Your system can't really use 32GB of RAM
3)Other applications take a lot of RAM and there is little left

There is the checkbox in the script generation which will take care of that (this is really a Windows problem as Pio frees the memory but Windows doesn't see it as freed as it has delays when reporting free RAM, if you check the checkbox and generate the script after that it will not run the check after the first tree).

It's enough.

That's a judgement call, it makes sense to aim at 2bb-3bb/100 in single raised pot in my view.

Yes, if there is forced 3bet (or even 4bet) you won't reach those accuracy and there is really little reason to aim for that as the goal should be to get good preflop play but the exploitability will be mainly postflop anyway.

It changes but experience shows that as long as the betting structure postflop is reasonable the preflop solutions are more or less the same. That's why it makes sense to simplify postflop play as much as possible when trying to get good preflop ranges approximation.

Ignore the names, they are just 2 different subsets. We've recently run another set of benchmarks and our view is that the best subsets are the ones listed here:

https://pastebin.com/AtyU84qC

Please follow the installation instructions in your email (you don't need 7zip for anything). Specifically ones in the 4 minutes installation video, here:

https://www.youtube.com/watch?v=ahEu...ature=youtu.be

Before I get into it make sure to download 1.9.2.5 version of the viewer, here:
https://www.dropbox.com/s/6p13pzi7j3...iewer.exe?dl=0

(download and substitute in your Pio folder)

there was a bug in remove_line in one of the previous versions so I want to make sure you are running the newest one first.

Yes, 40 actions.

Hard to say without seeing the whole case (feel free to email) but you don't need to include checks to the end (they are added automatically) so the last bet is enough.

Ranges and overall frequencies are showed without including card removal. This is the proper way for ranges because otherwise it would be incorrect to say that you get AA:1 on the flop (even if you always play it) as that would depend on exact range of your opponent. That's why we show what a given player carries to given point in play as a range not what they really have at given point.

When it comes to overall frequencies you can toggle between "what a player thing they have" and "what they really have" kind of frequencies. It's described here:

https://www.youtube.com/watch?v=zLGpcZavxeQ (16:10, see the video description).

It looks like 3 decimal places to me

Looks like 3 decimal places again

EQR is just EV/EQ they have to be normalized to be the same units before division though (normally EV is in chips and EQ in % of the pot).

While you may think about EQR in various ways the definition is EV/EQ (so how much a hand wins in comparison to how much it would win if it was forced check-down).

I think there's a bug in the Script generation process. I chose 0.35% of pot accuracy and then clicked "generate script" but when I looked at the script it says "set_accuracy 0.206500".

Here's the screenshot:

Also, I think it would be nice in the browser to be able to see the % of frequencies (I know you can see it using coloring, but sometimes it's harder to imagine the %) for each hand without hovering over it.

I pasted instructions above. What happened when you try following the steps?
You can also add:
ignore_mem_check on

at the top of the script


It said ok or success and then when I try and run my batch job it says I don't have enough memory. I just presumed this memory check bypass was for 1 tree and does not work for batch jobs via scripts because it never bypasses when I try and run after the script generation and I click run this script? Always has memory issues because PIO is using the memory ever since I built the tree.

Pio just reports what Windows tells it. Another thing to check would be if you are not running some crippled version of Windows (like home basic), some of them have limits on how much RAM can be used (if you go to control panel->system->system you will see that information).


I have Windows 10 Home edition 64bit and have been advised it can use up to 128GB of RAM. I also have seen where there is a message in system via control panel where it will tell you if you have RAM restrictions and I have no message, plus I have seen my PC use up to 26GB of ram so am fairly sure its not a restriction thing for Windows, obviously if it was I would need to buy Windows 10 Pro.


There are 3 possible explanation for that:

1)The tree is around 15GB
2)Your system can't really use 32GB of RAM
3)Other applications take a lot of RAM and there is little left

Yes my tree was around 15GB, what's the issue with that? And how can I resolve it? My motherboard allows up to 64GB RAM and as explained above Windows 10 Home will allow up to 128GB of RAM? There are no other applications taking a lot of RAM, I know this for sure, its just PIO after the build that takes it because when I go to task manager after or during the build it is taking a high majority of it.



There is the checkbox in the script generation which will take care of that (this is really a Windows problem as Pio frees the memory but Windows doesn't see it as freed as it has delays when reporting free RAM, if you check the checkbox and generate the script after that it will not run the check after the first tree).

Yes I am using this checkbox, the issue I have is getting it to run the first one, it always comes back with an error message saying I don't have enough RAM because PIO is using it after the build and before it has started to run. That's why I asked you for the bypass of memory check. I have already ran trees for 24GB of ram but its like I just need to get lucky for it to bypass the memory check. This is the main issue I am having. When the bacth job actually kicks in, everything runs smoothly.

Does the fact my version is PioSolver PRO have anything to do with this? How can I get windows to see the memory as freed after the build and before I run my batch job? As this is the issue here.

Thanks for your time.

The solver uses chips internally. 0.2065 is 0.35% out of 59 which is likely your starting pot.

You can see them if the 13x13 are is big enough to display minimum size font. See here:

https://www.youtube.com/watch?v=5Ozjfnatwms (question 11)

While this is about EV specifically you can choose to display strategy ("strategy" button) and then there will be fractions in 13x13 area. Notice that they are average for all combos in a given group (so for example 4 of them if you look at 76s) so they may misrepresent what is really happening (as some combos may have completely different strategy than others within one group).

It should ignore the check after the first tree and if you add it to the top of the script it will ignore it even for the first tree. For normal tree building estimation actually disposes of the previous tree so there shouldn't be issues either.
Are you running the newest version btw (should be 1.9.2.5 if you go to About in the top menu). It seems there is some kind of misunderstanding. Feel free to pm me so I can add you on Skype, maybe it will be more efficient to do a quick session.

Yes, that's not a problem then.

No issue, you said it takes half the memory so it looks like it's around 15GB.

I am confused. Pio checks for RAM before building the tree (obviously as it needs to know if the tree fits in memory). Maybe you are attempting to both build a the tree and then run a script. It won't work as script generation window opens it's own solver in the background and if the memory is taken it won't be able fit its tree. We have chosen this kind of design so you run a script on the side and still do things in the viewer (which has its own solver open). It looks like you are first building a tree in the viewer, this takes 15GB or so you are then attempting to run a script and then there isn't enough RAM anymore. If that's indeed the case you can do 2 things about it:

1)don't build a tree in the viewer before running the script
2)call "free_tree" command from arbitrary solver command before running the script (again, ctrl+b etc.)

1) is the recommended way as 2) may cause Windows to not see the whole memory as free.

Pio doesn't have any restrictions when it come to amount of RAM it can use.
You can't get Windows to report free memory exactly sadly. The reason for this is Windows' architecture and the way virtual memory (RAM + swap file) works. I am not going to go into that here.

Right now it looks like your problem is going to be solved if you disposed (or don't build at all) a tree in the viewer before running the script. I explained the architecture above, I hope it makes sense why we have chosen to implement it that way.

Hi Punter, thank you for the answer, so confirm me that I have to set it as in the following picture?

https://s16.************/otopmgq6d/s...-_14-43-35.jpg

I wanted to ask you, can you stop a preflop calculation and resume it later?

If I put a flat / 3bet / 4bet blocked range, what accurance do I need to stop the calculation for a good approximation?

When I calculate a preflop tree, usually the CPU works at 90%, there is some option where you can decide whether to use all the cores / threads or not?

http://imgur.com/a/1NvWC

Hey guys, when will you roll out a software that also accommodates PLO?

Yes this is obviously the issue here because I do build the tree before running the script.

This may be a silly question but if I don't build the tree, how does the script know the details for the tree I want to build? Your saying if I put in all the settings in the viewer, then generate script and run script, it will then check for memory and run with the settings that are there? I don't have to tell it my sizes for all streets because it just uses the ones that are in the viewer? I think that's what your saying and this is the mistake I have been making.

Thanks a mil for your time and offer for PM and Skype session, its much appreciated. But I might be ok if I don't have to build the tree in the viewer before generating and running the script.