Also, when will the improved nodelock functionality be out?

In how far is PioSOLVER better/worse than crEV in modelling a certain strategy that is not equibilibrium?

AKA is Pio capable of nodelocking a strat, in the same way that crEV lets u model 2 strategies and run a max exploit?

AKA what things Pio is not capable of doing, that crEV can?

TY

1. When PIO calculates strategy, does it take into account whole tree, or is it doing it
street by street? If we set complex parameters for OOP player and only basic ones for IP, how good strategy do we get compared to complex parameters for both players?
2. If we come to the river with 0.x combos, is PIO's strategy OTR still accurate?

Yes, or better yet just save configs. It's better to node lock on fresh trees anyway.

We don't give ETAs or make promises when it comes to features. Those never work out anyway. My view is that a business working in a way ours does (that is selling copies of software we first produce for a small price (in comparison to development costs)) isn't really suited for giving ETAs, making promises etc. For that you would need your own employees. The reason for that is that there are too many things which influence what we choose to work on at given time and it's too easy to overpromise and be too optimistic.

Questions about comparison to other software out there don't really belong in this thread. For the record: I don't know, I don't have crev license and I have never used it.
We don't have much functionality to model non equilibirum strategies at this point though.

Again, this doesn't belong here. I don't know answers to those questions and with the amount of interesting things there are still to implement/work on in poker I have 0 interest in finding out as there will always be something new/useful to do.

I am not sure how street by street approach could work unless we are talking about big simplifications. Pio takes the whole tree into account. It means that if you change bet sizes on the river it may (and likely will to some degree) influence flop strategies.

That's a difficult question. One would need to do a lot of research to answer it. If you are smart about removing IP's options (that is remove those which don't influence overall EV much) you can get pretty good approximation. It's not easy to decide which options are safe to remove though.

Yes.

What do u mean with save configs? Don't do saves just save parameters of the tree and set it up+calculate again?
Why is it better on fresh trees?

Sorry about comparison, it's just that Pio is a great program, but w/o a great nodelock it's more about finding truth for the sake of it rather than for practical implementation. So I would politely suggest too focus all attention to nodelocking functionality;-)

Verstuurd vanaf mijn XT1635-02 met Tapatalk

You can save a config itself (just a few kb, enough to re-create the tree), I mean those:
https://gyazo.com/083e1d2c0c945deb46eb86e29e105cfe

You can also save trees with the results, there you have a choice between full/small and micro saves (all the information, flop+turns or flops only).

The solver's algorithm works in a such a way that it might be difficult for it to adjust from already well solved tree to something new (which you introduce by node locking) so it's more reliable to node-lock on a fresh tree and solve. You can do that on already solved or pre-solved tree (by pre-solved I mean one which you just let run for a short while) and it will usually work.

That is for sure one interesting direction to take. The thing is we think there are more interesting things right now (it will be soon clear what they are). We would of course like to do more tools for creating/analyzing strategies used by real life opponents but it has to wait for a while.

How do I calculate raise size percentage? Is it percentage of total pot after the bet is made?

Yes. You can also use x syntax, for example: "3x", "2.5x" (notice that's a dot, not a comma). Like here:

https://gyazo.com/e7f909ad29cddb1f8be94e0f53836792

Dear support,

I have a question about river spots.
So I have saved postflop trees only with flop & turn an I'm loading them with my own program. However If I run a command: show_strategy xyz for a river spot, it displays the results. Is that calculated on the fly? How accurate is that? Do I need to give any other command, to get proper results?

Thank you

Yes, this is recalculated on the fly. This happens automatically every time you ask about river node.

The accuracy is given by set_recalc_accuracy command. The defaults are 0.1% of the pot on the turn and 0.05% on the river. This is in general way more accurate than you would get by solving the whole tree even to very good accuracy (the reason for that is that spending 200-300ms on a river is not a lot if you only need to spend it on those you ask for but it's prohobitive if you need to do it on all rivers in the tree).

No, although you may want to play around with set_recalc_accuracy to find a balance between speed and accuracy. We think the defaults are pretty good though.

Hello,
I have a Piosolver edge.
I want to make a simple 3bet tree in HU hyperturbo;
for example: SB(button) min raise 90% of his range and I want to calculate what should I 3bet AI and 3bet NAI.

1. Is 8gb RAM and i7-4720HQ enough to do it?
2. If yes how can I do it?

Regards

Assuming you can call the minraise the config looks like this:
https://pastebin.com/5fAHMw5s

(you can copy the text and then use Tools->paste tree building config) to load it in your viewer. It looks like this after loading:

https://gyazo.com/fdc923e60441630ae538788a17869e55

You would be able to run this tree on reasonable subset of flops (say 42) below 32GB o RAM but certainly not having 8GB.
If you remove the option to call then you would be able to run it, then the tree would look like this:

https://pastebin.com/bwAL58KZ

In general runnin the preflop solve on a laptop is a bad idea. You need either a dedicated server or cloud instance to run it there in such case.

Btw, we recently determined the best subsets out of the ones we are shipping, they are listed here:

https://pastebin.com/AtyU84qC

Is PioSolver portable?
I mean, can I use it on my work PC without admin rights to install it?

In general it should work without admin rights. I recommend trying the free version first.
I am not sure why you call it "portable" though. The license is not intended as floating one although you can move it from machine to machine within reasonable limits for purposes like getting a new computer, some experiments, long travelling etc.

Hello,

I didn't get the difference between basic and pro version.

1 - fully scriptable. What is it mean ? is it useful for what ?

2- uses up to 4 threads (basic) , Uses up to 12 CPU cores (12 hardware threads) (pro). I didn't get it too, is the computer capacity ?

3 - I was using free version to understand better the program, and it solve only turn, but I realize that river strategy will depend on what I did on the streets before, so when you give me the turn strategy, what was the assumption made on flop ?

4 - what is node locking ?

5 - Can I change my strategy based on what I think villain is doing ? for example: the program can solve a optimal strategy based on a player that it isn't playing optimal?

Hello,
Is there a way to manually pause a script while running it and restart it later?
Thanks in advance.

Verstuurd vanaf mijn ALE-L21 met Tapatalk

The main usecase is automating solving on many flops. You can for example schedule solving on 100 different flops when you are asleep our away form home and then browse the solutions or run aggregation report on them.

Basic version is compile to work with 6 threads these days. It only matters on hexa core CPUs or bigger ones.

The free version solves one flop these days (Qs Jh 2h) so you can experiment with that as well.

You can tell the solve that you want it to play certain hands at give decision point in certain way. For example you want AA to always raise etc. Then you can solve with those assumptions.

Yes, this is what node locking is for.

You can't. You can kill it and restart it though. It will automatically skip over already done trees (as long as you don't move the saves from original location) and start from the first unsolved flop.

Hello,

I mailed support@piosolver.com, but got no response so far.

I dont have access to my old piosolver since i formatted and now using another computer, and i would need to reset it so i can reinstall it

Waiting for your help,
Regards

We got your email around 30 minutes ago. It's Saturday night so it's natural you are not getting an instant response
I've just answered all the emails so your issue should be solved already

Actually i sent the email on 16th April, but i dont mind i was not in a hurry.
I still dont have an answer though,

my order is #5669, you should see my email

A few quick questions:

1. Any idea if RAM frequency (e.g. DDR 2133 vs 3200) matters at all in terms of computation speed or is pio purely CPU limited?

2. Since the upper-tier Ryzen's have 16 threads, have you thought about a Pro version that can use all of them? Jumping up to edge to take advantage of this seems like overkill given that with a 64GB RAM limit on these machines you still wouldn't be able to do much preflop work.

Sorry, one more: is it possible to estimate how big a preflop tree would be with the free version? I'm just trying to get a sense of what the limits are.

I see, we got a very similar request yesterday.
In your case the reason you didn't get an email from us is that your email server refused to accept an email sent from us as well as the following tries.

I suggest either using a different email server or pm'ing me here.
I need your registration key and in case of pro/edge license information if it was the earlier or the latter license.

It's not purely CPU limited but mainly so. Last time we checked it there was little difference once RAM was clocked at 1600. It's very unlikely it changes much in the future.

Yes, the pro version will support 16 threads in the next release. I am not changing it now because the difference between running 12 and 16 threads on 8 core CPU is minimal (likely less than 5%). The reason for that is that the way hyperthreading works you don't get direct speed up from running the extra threads. Switching between them is faster which makes some application faster (like Pio) but this is limited to the savings in context switches. In case of Pio there is about 15% speed-up due to hyperthreading (so going from 8 to 16 threads). Running 12 threads will be almost as good as 16.

I agree with the first part. I don't agree with the second. I see the tendency to overblow preflop trees but you can really do a lot with 64GB. The tricks is to simplify postflop play as much as possible (it doesn't influence preflop play much or at all if you do it right).

No, not right now. It's possible with basic/pro version though.
We will have some preflop functionality in the free version in the future to allow some testing (I don't like trials so those won't happen, I prefer free vesions with enough functionality). This is not high priority at the moment though.

Bad news

Does anyone have an idea how to get a rough computing time estimate when running huge scripts?

Is there any emphasis gonna be on creating learning paths with PioSOLVER by the PioSOLVER team or is this application work of PioSOLVER gonna be left over to external parties? Theres obv alot of value in simplifiying Pio stuff into streamlined strategy development in collabo with u guys, however I dont know if u guys are mostly programmers, and dont wanna get involved in the practical implementation of Pio or u are open for this type of expansion?

Are u guys just hiding in ur labcaves in Poland being all mystikal or is Pio gonna be going in this direction in the future?

I think its win-win.