This is how every tree starts after being built. You need to solve it after that by clicking "go". I recommend watching a quick start video explaining the process, here:

https://www.youtube.com/watch?v=JqGQoQKbCB8

and specifically 30 seconds starting from 1:40

Another possibility is that sometimes the view doesn't refresh after solving. Just click on yellow "root" rectangle at the top or go back an forth to refresh it.

Hi, just a suggestion:

It would be nice to have different boxes for raise sizes based on how many raises have already occurred. That way the solver will no longer bother with low-frequency raise sizes. Seems like a good way to trim down the game tree size.

The problem here is that there is always a trade-off between simple to use interface and a powerful one. For example making full tree construction possible (as we do for preflop) would be quite unusable for postflop. We tried to add all the frequent options useful to trim the tree but it's true that if you want something vey specific you need to edit the script yourself.
Make sure that setting cap or "don't 3bet" option don't solve the problem for you.

This is one of the things we want to improve in further releases. Probably the main interface will stay simple but there will be an option to edit specific lines.

Hi,

i want to run an aggregation report on a subset of flops, all with different weights.

ideally i want to break up the running of a script into 2 sessions

so say i want to analyse a 49 flop subset. i create the script for 49 flops.

i want to run the first 25 flops in one batch, so create a script to do that

and then the next 24 flops in another script [so my comp isnt running for too long / when i am out]

can i then add all the flops into one folder with the original script for 49 flops and get a correct aggregation report?

hope i've worded this clearly.

hi there,

I am a new user of the Edge version. I've got quite a big config for a personal computer, with i7-5930k (3.5GhZ, 6-core <-> "12 CPU"), 64GB of RAM (DDR4 @2133). I started yesterday with some tournament (smaller stacks) SB vs. BB preflop solving.

The first mistake I made was to save a full tree after setting it up. A 5BB-deep simple situation in HU (limp/minraise/shove) generated a 12GB file.

For this first complete computation, I didn't check the oop-must-check checkbox; but I will definitely try it soon as it _really_reduces_ the size of the RAM for a tree.

Currently, I am struggling finding how-to information about the software. This thread is already 62-page long, some videos show how v1.4 or v1.6 worked, overall I have to copy/paste whatever relevant info I can to build my own digest manual. Anything I missed there?

Please help me with this simple question : I setup a preflop tree, estimated the size so it can fit in the RAM, then I built it (ahem, PioSolver did actually), then I start the solver. Then, I can monitor the RAM and CPU usage, but I have no clue about the eventual convergence of the solution. The question is : if I stop the engine, click calculate, browse and realize that this is not close to equilibrium, can I click start again without losing the computation? Start/Stop/Calculate/Look where I stand as many times for the same problem?

Thank you for your help.

Look here:
https://www.youtube.com/watch?v=VhG_YsEvfwU (questions 6 and 7)

Consider just constructing one script, you can stop it and run it again later (it will skip over already done flops if you use PioViewer to generate it).
There is also an option to insert a break between flops to give your CPU time to cool down. Yet another thing you could do is to use less threads than maximum supported, in that case add "set_threads n" at the top of the scrit (no quotes, n = number of threads to use).

If those aren't enough to solve your case just generate the flop for all 49 flops, then cut it in half (after 25th flop) and run it. Then when you are back, remove the 25 flop script and run the full 49 flop script, it will nicely skip over all done so far.

The skipping mechanism is implemented by "skip_if_done" command and is automatically added to scripts by PioViewer when you run generate script function.

Notice that if you intend to run aggregation report you need a full 49 flop script in there as the viewer reads weights for flops from that script.

thank you, extremely helpful

There is a simpler way I didn't think about:

1)generate a script for all 49 flops
2)insert "exit" (without quotes) somewhere in the middle after dump_tree (say after 25th)
3)run this script
4)later remove "exit" from the script
5)run again

thank you.

when you say insert this would suffice:

Yes. That's correct. You can also insert it just after dump_tree but it doesn't matter. "Exit" termintes the script so the window will disappear but don't worry about it, the saves are there.
Notice that this script wouldn't skip correctly over already done cases in version before 1.8 because of spaces in file names but it does work now.

One small bug I encountered:

to insert the exit command I had to copy paste into notepad, and when i tried to paste back into the script generator box it wouldn't paste the whole script (I think it pasted 74 of a 95 flop subset).

That really shouldn't be needed. You can edit the script cirectly after clicking "generate script". Are you at 1.8.2 version?

I'm not at my comp, but I only upgraded to pro 4 days ago.

It allows me to type in the script generation box, but I could not start a new line by pressing 'return'.

Well, try it again:
http://i.imgur.com/9QOZxCA.gifv

report if it doesn't work

yes i am running 1.8.2

just tried it again after restarting and got the same result.

cannot start a new line in the script and when i try to copy and paste from notepad into the range generation box only 73 of 95 flops paste over - its as if there is a limit on # of characters allowed. here are the last 4 lines when i attempt to C+P over.

trying to C+P the remaining script doesnt work

Please, add this feature
https://gyazo.com/7b4f554a8c0a7fbae597d9ab9b6cce24

Because many times I can't remember if I set a 55%-60%-65% bet, I asked it in the skype group twice and a lot of people were happy with it

hey,

can i force piosolver to make a strategy when it can choose a card 100% or 0% for an action (raise, fold, call)?

for example, normally it choose K2s 50-50% K3S 50-50% K4s 50-50% K5s 50-50% for a raise or fold a given spot. In my example it should choose K2s 100% and K5s 100% for raise maybe, and K3s K4s 0% or something like it.

I hope its not a stupid question

The two strategies are not as similar as you think, unless the other player has no 2x-5x in his range.

i dont think its similar. but i wanna force pio to choose 100% or 0% raise fold call in every hand. maybe it would choose this maybe not
so it will be exploitable for sure

other question:

i made 2 preflop trees for 8.5bb hu

first tree oop vs limp can izoraise 2x, izo shove, checkbehind
second tree oop vs limp can izoshove or checkbehind, CANT izoraise 2x

everything else the same

Obviously first tree is better for OOP

In first tree OOP EV = -0,1
Second tree OOP EV = -0,2

My question is, how big this difference is? I dont understand what this ev numbers mean
Should i definitely use strategy 1 or this 0,1 difference is like nothing?

Yup, it's on the list somwhere. On the other hand there are many things on the list

Yeah, basically this. It's not that easy to dumb down (or like some people prefer to call it "make more human") a strategy without making it significantly worse. It's even more difficult to keep that constraint during solving while maintaining something close to GTO.

While this is theoretically possible to solve with such constraints it's not easy to make it work. All known algorithms for finding equilibrium are based on possibility of making small steps.
I am sure we will get there at some point but it's more of a research problem than a development one and there is a lot of things we would like to do before attempting that.


If you want any significant comments about trees you create you need to attach the configs. Otherwise it's not clear what you mean or if you have constructed them correctly.

Assuming you constructed everything correctly it just means an option to iso-raise to 2x is worth 0.1 chips per hand. How much that is depends how big the blinds are. If there were 5/10 blinds for example it's a 1bb/100 difference.

how do the solutions of your programm compare to pokersnowie?

dont wanna spend 1k just to find out it gives very similiar advice...

I have no idea as Poker Snowie is not a GTO solver. I am not into business of analysing software based on not clear model without specific claims about accuracy. Your best bet is to ask around, I am sure many people use Poker Snowie and one of the GTO solvers.

I understand you are running a business. But for a potential customer there are many products out there who all claim to play GTO poker. I wanna compare/verify the accuracy of their results before committing to one product.

maybe give me/this thread the chance to compare your solutions to your competitiors in some tricky preflop spot.

There aren't that many. I can tell you more about our competitors who in fact solve for GTO solutions. Poker Snowie isn't one of them (at least last time I've checked their own claims).

But there is nothing to verify when you don't have any kind of claim about accuracy.
Out of the people who in fact solve for GTO it seems it's a consensus that every available solver leads to the same results. Where they vary are set of features, speed, memory requirements etc.

Again, you can ask me about comparisons of specific features but it's a better idea for you to ask around.

Our solver tells you how good the solution is in terms of exploitability across the whole tree. When it comes to preflop trees you can usually get to 1bb/100, sometimes to 2bb/100 on big trees with smaller subsets.

I can tell you that I played around a bit with other solvers and it seems the solutions are the same. Ours is verified using various algorithms implemented by us, other programmer who made UPI compatible solver (Oskari, who first got his work on solving limit holdem published in Science and then programmed by far the fastest solver on the market). We also have river solver based on completely different approach which gives the same results as well as two implementations of EV/MES which confirm the solutions are correct.

From all that I can guarantee you that what our solver claims is in fact correct and that probably no one else run as many tests as we did. I can also tell you that our competitors who I can consider serious have correct solvers as well as far as I was able to test.

As to Snowie:
-not a GTO solver
-not specific claims about accuracy

There isn't even anything I can test there as their goal is different than ours. Your best bet is to ask around and listen to what serious players say.