hey guys!

Did some simple aggrigation report on piosolver and encounted bug in the results as it is on picture below (checked with friend he has same bug)

This is not a bug.
The numbers shown in the viewer represent what % of the range take a given action while the number in the report represent how often this action happens.
The difference between those two numbers it not obvious at first. They are different because of card removal effect from the opponent's range (what the opponent has influences how often we have various hands). In your example opponents range is such that we fold more often in reality than simple range analysis indicates which makes sense (the opponent already called a bet so their range "removes" more of our non-folding hands).

We have chosen to show % of the range instead of real frequency by default because we think it's a more useful number for general analysis. Otherwise there would be a lot of confusion in spots like for example:
-range: AA and KK
-AA calls 100%, KK folds 100%
-total call: 45%, total fold 55%

which happens on regular basis.
You can however change the number which is shown to real frequency by going to Tools->Configuration->Data presentation on 13x13 grid->show total strategy as real frequency.

So much to learn...

Is it possible to change available bet size for each specific Turn card? (...and each River card?)

For example, if the original tree allowed Turn bet sizes of 33,66,100...could I alter the tree to change available bet size to allow only 33 on specific Turns, and 66 or 100 on others?

Conceptually I don't think that make much sense but I guess you could go through and node lock each specific turn card. Why would you do that I don't know.

I would like to calculate a single "optimal" bet size for each Turn card (if I was to use only one bet size per different Turn card) and the resulting GTO strategy (...for the entire group of Turn cards with their specific bet sizes).

Hello, im running preflop sims with ante. Should i deduct the ante from the effective stack ?

It's not possible as explained here:
https://forumserver.twoplustwo.com/s...postcount=5690 (last question)

Yes. Deduct from the effective stack and add them to dead money in the pot.

Hi,

Where could I found something about the maths behind Piosolver?

Thank you

Hello, I emailed PIO support few days ago about my license, still didn't get any response.

Thank you.

Over estimated memory capacity and ran to many scripts at once causing some of them to stall in the middle stages, will I be able to run these once the current scripts finish and have more memory or will I have to start from scratch as its been a few days already? Thanks.

figured it out already, to resume script, simply Tools > Load Script.

There isn't anything specific to our algorithm you can find.
You can find information about solving for equilibrium in general by googling for "poker equilibrium" and about specific algorithms by for example googling for "counterfactual regret minimization".

Maybe your email was missed for some reason or maybe you sent it on Thursday evening just before our national holiday/long weekend.
One way or another please pm me your email address and I will check what's going on there.

I'm having trouble understanding the main grid's Bet/Check frequencies. [link to related images at bottom of post]

On the following Flop [Td6c5c], for example:

1. OOP checks
2. IP Checks KTs 29.9%
3. IP specific combo frequencies Checks: 44.2%/44.2%/1.4%/x

On the Turn [Td6c5c][As]:

1. OOP checks
2. Main grid: IP Bet KTs 40.6%
3. Specific Combos IP KTs Bet: 14.0%/9.6%/98.1%/x

On this Turn, IP is betting 11.8% of remaining combos:
(0.442)(0.140)+(0.442)(0.096)+(0.014)(0.981) = ~0.118

Why does the main grid display 40.6% on the Turn? If this somehow has to do with card removal (?), how do I get it to show "Bet: 11.8%" on the Turn?

ThanX!

Couldn't get the image embed to work (kept showing 404 not found)...here are the related images:

https://imgur.com/a/VChYf0u

The default is just average across all combos with every one on the group being weighted equally. If you would like to weight them by frequency you have a given hand in range then check "bars width proportional to weight" checkbox, here:

https://gyazo.com/0ef3cc3039108e4972c18e1b8b77e6b0

Please notice that averages on the squares on 13x13 grid have little value postflop as they combine values for hands which are often strategically completely different.
In your example it's of little value to know what to do with "KTs" on average when one of those combos is a very powerful hand while others are quite weak.

If i want to check only flop frequencies can i do that without filling turn and river options or would that change the whole strategy?If i want to only check out flop cbet frequency do i need to fill in turn and river options aka bets,raises,donk bets?

[QUOTE=punter11235;55538569]Yes, you can use node-locking and combo-locking specifically for that.
Choose the decision point you want to edit and then click ctrl+N to bring up node-locking interface.

I edit his folding range where he folds a lot more now and his fold frequency increased but our strategy didnt change,we were still cbeting and checking at the same frequency even tough he folds like 7-8% more to the bet.Why is that?

Thank you for the replies.

Yes, you need to fill turn and river with sensible options. It doesn't need to be very detailed if you just want flop frequencies but it must be sensible with no important strategic options missing.

Did you recalc the tree?
Is the node-lock preflop or did you just node-lock one of the flops?

hello,

I'd like to run a few aggregated reports on multiple runouts, and it seems very slow on my PC.
Like one is currently running for 5 hours now (Turn spot, on turn saves file), so maybe I am doing something wrong, I don't know.
I only export such a report if the spot is saved in the cfr file though (I never export a River spot if it's a Turn save for example), but it doesn't seem enough to get a decent speed.


It would be helpful if you could answer these questions, so that I understand how to optimize this feature :

A- this export feature seems to use very few PC resources, but will it be significantly slower if it's run while Pio is solving other spots ?
B- will it be significantly slower if I launch 2 (or 3, or X) reports simultaneously ?
C- when I run an export, what simultaneous process would be the worst in terms of speed : letting Pio calculate simultaneously other sims or launching simultaneously another analysis export on multiple runouts ?
D- If I launch 2 exports (A & B) at the same time, does it change anything in terms of speed if the 2 spots don't use the same cfr files? (I mean cfr files for A and B are all on the same drive, but in one case they are in 2 different folders, and in the other case the cfr files are used by both reports)
E- even if the spot is always in the cfr saved file, I think it's always like : flop spots faster than turn spots faster than river spots, right ?
F- at one street (whatever : flop, turn river) : will the report be slower if the action is complex (like "X-raise-reraise-reraise" vs "donk") ?
G- is it possible to run these exports on a PC whose RAM is < to the tree of the cfr files (if I run the sims on a big server, but I want to export the analysis on my PC for example) ?
H- Are there other things I should do to speed up this feature? like setting this process in "high priority" in the task manager, or something like that ?
I- what uses to be the limitation which limits the peed of this feature : RAM or CPU? or something else ?
J- if I have exported 1 report like this, and after it is done, I select another spot and launch a second report, will the second report be slower because one was run before? (this sounds a stupid question, but this is what I have just done so I am wondering if this might be the reason why the second one currently takes 5 hours while the 1st one took 2.5 hours).
K- does the time it takes depend significantly on the cfr file size?
L- does the time it takes depend significantly on the nb of combos in both ranges?
M- does the time it takes depend significantly on the nb of bet sizings (tree size)?

I guess I should understand better how it works after that :-D

How would I simulate a $6 rake if seen flop and a $1 flop if not?

will piosolver run faster on a machine with dual processor?

Something is definitely wrong there in such case.
Are you running any 3rd party antivirus? Sometimes those do "file scans" which slow everything that works with multiple files to a crawl.
In general most of the time spent when doing the report (as long as no recalculating is needed) is spent by the loading trees.
I am not able to answer all your specific performance related questions (I really haven't benchmarked it much). It's a good idea to make sure no other software is impacting the performance (especially 3rd party antivirus), perform clean system restart and before running any other potentially resource hungry programs perform a report to see if things improve. If they do you can take it from there.
Let me answer some of your specific questions:

In general it shouldn't be that much slower as most of the time is spent loading/disposing of the trees while solving itself doesn't use resources for that (treebuilding does).

Probably.

Yes because there are many more runouts on the turn than there are on the flop and many more on the river than there are on the turn so there is just much more work to do.

Probably not.

It will work as long as your swap file is big enough but it will be painfully slow. I am talking about the save size though. If the full tree was 10GB, small save is 700mb then you just need that 700mb in RAM.

Don't run other resource hungry programs at the same time.
Make sure the system is in relatively clean state. Make sure no malware/crapware is running (3rd party anti-viruses).

Mainly disk speed and having enough RAM (as long as no recalcing is needed).

It should be the same but maybe it's a good idea to restart the solver between doing those reports.

It's not possible at the moment. It's the first time we hear about rake being taken if the flop is not seen. We might add it in the future.

The question is faster than on what? If you are asking if it's faster than on a machine with only one processor of the same kind then the answer is yes.
The speed in general is proportional to the number of physical cores multiplied by the frequency they run on (assuming those are modern CPUs).

ty very much for the answers and the good support in general !

I have just tested something :
step 1 : I had Piosolver still opened after running my 6 hours export, but export was over
step 2 : I have dragged another cfr file in Pioviewer and it took a lot of time (like 8 or 10 seconds to load a 222MB turn save file).
step 3 : I shut down Pioviewer and restarted it again
step 4 : I dragged a similar cfr file (same characteristics) and it was much faster (less than 1 second)

So it seems that the more files have been uploaded to Pioviewer, the slowest the upload of next cfr file.
Even during 1 same export, it seems that the more files are being uploaded, the slowest the upload time for each additional file (tested with 1000 boards).
I have run this test a few times, and it's always like that.

why is that ?
what can I do to make the speed of Piosolver upload files feature constant ?
can I do something regarding the computer RAM maybe ?
is there a batch file, a command.exe or a script (ahk?) I could run to release the memory or something like that without having to shut down Pio and start it again?
could this command/script be run while one export is being processed and is getting abnormally slow?

I am still running my tests, and it seems that this issue is a much bigger constraint causing slowness than running X analysis reports simultaneously.

I have installed tools to optimize RAM, but it doesn't seem to change anything.

So the key question is : what can cause Pio to slow down the more files are being uploaded?
and how to fix that?