Is there a way to turn a small save into a complete tree, for node locking, or do i have to compute the whole tree again?

Yes, we invite you to the Discord. While it's not that active yet as it's less familiar than Skype we think it's much better for many reasons. We also have almost 1500 members in our channel now although most are too shy to post

Short answer:
There isn't but building a new tree from scratch, locking and solving from there is better

Longer answer:
There theoretically is but you probably don't want it. You can call "rebuild_forgotten_streets" command from solver console (ctrl+b to bring it up). It will take a while but it will rebuild the tree (with unsolved rivers of course), you can node-lock now and solve. It will likely not be faster than solving from scratch though.

Hey, I started getting this error message:

"Setting and locking strategy doesn't make sense on incomplete tree. Do you want run plugin anyway?"

and

Pugin Tree: Round strategies failed
ERROR: round_up_to can't do that on incomplete tree

Why didn't I get these error messages before? Weird. Anyway, I can't make my tree complete, because I have BTN always raising certain range, so he can't fold those hands. How the hell can I node lock now and round strategies?

BTW, after rounding, should I still run the PIO again?

I still do not understand some things. I will put it with an example:
- Starting pot 75
- Bet 25%
- Raise 50%

Bet = 75 * 0.25 = 18.75 rounded 19
Raise = (75 +19 +19) * 0.5 +19 = 75.5 rounded 76 (if I understood the banking rounding well), but PioSolver shows 75

Yeah, apparently it's rounded before addition for some reason. I will investigate why it's happening.
Again, if rounding in bet sizes matters to you when creating a tree you are likely using too small initial pot/stack sizes.

Hi,

What are the differences between the flop subset that you recommend on your website (2015) and those downloaded with the solver (flop subset 2016)
For example : the 184 flops subset are not the same. Is the more recent one more precise ?

Thanks

The ones we ship now are newer and better. They were recalculated I think in 2017. 2016 in the name doesn't mean anything

Hi,

I'm looking to simplify a complete tree in order to keep only 2 sizings on turns. The idea would be to force the solver to use sizing A on some turns, sizing B on others. Basically it is like adding/removing lines but with the possibility of specifying turns. Is it doable ?

This is right now not doable. Making the betting structure the same on all the cards allow many optimizations so that was the design choice. It's possible in the future we make it possible although it will make things slower.

Hi,

I have two questions for you guys,

1) I notice that the solver likes to bet quite rarely and prefers a small sizing @ monotone flops. Do you know why ?

2) In a SRP situation, with a starting pot of 55 and an effective stack of 975, I propose to the solver two differents simulations OTF :

- bet 33% pot OR check

- force to bet 33% with full range

I made an agg. report with the results below


I was asking myself from how many EV loosing we should consider give up the simplified strategy (force bet 33% full range) ?

For example :

a) 776s simulation 1 => IP EV = 32.53
776s simulation 2 => IP EV = 31.79

That's a 0.74 chips difference, with a starting stack of 1000 chips that means 0.07 bb right ?

Is it really that relevant or not ?

b) A32m => IP EV = 31.40
A32m => IP EV = 30.01

That's a 1.39 chips difference, with a starting stack of 1000 chips that means 0.14 bb right ?

Is this amount more relevant or is it still negligible ?

We can bet more frequently if we employ a small bet size and it looks like it's more profitable to force junk to fold every time than to get value vs middle of the range on flops where not much is happening. Why is that though I do not know. It just turns out to be more profitable strategy.

When having strategies which are close in value it's worthwhile to think which one is more likely to induce mistakes. It seems small bet strategy is successful because people do a lot of mistakes vs that like folding too much and raising too much. You can often simplify without losing much EV but it only makes sense if your opponent is likely to make mistakes against your simplified strategy. Simplifying for the sake of simplifying just makes your opponents' life easier.

it's 0.074bb/hand or 7.4bb/100 hands. Average rake in postflop spot in 5/10 games is 15bb/100 so I would say 7.4bb/100 is not nothing.

That sounds like a lot but again, it's a judgement call. I don't think EV differences are the most important factor when deciding if it's worthwhile to simplify the strategy. See the previous paragraph.

When I run agg reports for the turn and river the fold EV is in reference yo the flop. It is kinda weird looking at -EV calls, and -EV folds. Is there any way to have turn reports reference fold to 0 and river have fold EV 0. Like it is in the GUI? Is this how it is supposed to be or did I mess something up? Thanks.

Yeah, that's the way solver reports it. It's always EV of the whole line. The GUI changes it to the usual convention using the formula in Tools->Configuration->data presentation

I think it's not possible to apply those to aggregation reports right now but I will talk to Kuba (PioViewer developer) to see if we can add it.

Thanks. It is not such a big deal. I more or less only use the turn and river reports for the bet/check/fold frequencies. I just thought I might have accidentally messed up a setting. Thanks.

Got 2950x ryzen, and getting 3.8sec bench, I gotta do something incorrect here.. also tried OCd to 4.2GHz but no difference, anyone has 2950x and having similar spots?

Is there is any hardware I can add to my computer so that piosolver will be able to do more complex trees and run faster? I have an intel computer, 4 core processor 16 gb RAM. Thanks for any help.

1)which version of Pio?
2)energy saving settings off? (full performance)
3)how many cores are working when you look at it in task manager?

You need more RAM for bigger trees and faster CPU to solve them faster. There are CPUs 2 or 3x faster than your on the market (depending on the exact model) but they are pretty pricy and will likely require upgrading other components as well.

Fixed, it was about RAM since the Ryzen 2950x is quad core if should have atleast 4 ram sticks if not all, I used 2 sticks only before.

Without overclocking bench 2.02, and a little overclock to 4.2GHz I got 1.95 which is sick!

Advice for future buyers, definitely can recommend 2950x

I think I am going to get one of those myself. It's more than 4x speed-up in comparison to my Ivy-Bridge quad. I think you are 2nd person who run into RAM issue with Thread Rippers.

That's awesome! What cooler and CPU voltage are you using? I have a Liqtech TR4 360 and a 2950x and at 3.9Ghz (1.26875 vCPU) it runs between 68-74C even with the side of the case off and I'm a little leery of trying to push it further.

I should also add that without an overclock it benches at ~2.3 and at 3.9GHz it's ~2.09.

Are you using distributed or local memory mode? I've tried both and local seems very slightly faster for the benchmark, but I'm not sure how that applies to a longer-running calculation.

Base clocks, I've been trying 4.0GHz, 4.2GHz but since the baseclock is swiching between 3.5-4.40 (with the boost) im ok with it, it seems to be stable.

Using AiO, Kraken x72 (360) - I would get Enermax 360 II if it's in the market but it wasn't.
It runs stable 68C (case closed).

Im using currently the older PC RAM which is 4X8GB 2400 Kingston.

Going to read through.

Hi guys, I'm a piosolver pro user. I'm looking to buy a new setup in order to enhance the speed at which piosolver gives the solutions. I would prefer if i can get solutions for a single flop within 5-10 mins even for a 100bb big tree with multiple bet sizes. Can you suggest me what shall be the specifications of the system that i should buy to get that kind of speed? I plan to run scripts as well.So keeping that in mind how should i proceed?

I don't have much technical knowledge. So you're assistance would be much appreciated.

I just researched buying a new computer for PIOSolver. I got most of my information from reading #hardware channel in the pio Discord. I think posting there would be a good idea as well. I was able to take advantage of some Black Friday sales so you should be able to as well if you make the purchase in the near future. Relevant information for someone to give you suggestions would be:

-budget
-whether or not you plan on upgrading to edge and using preflop functionality in the future

My budget is $2000 - $3000 . Yes i plan on upgrading to edge. I saw the discord hardware section as well but got confused since i felt most of the people are assembling it only. Given i'm beginner at assembling(have used imac all my life) , wasn't sure what my question should be, so i posted on the general forum first? Is there any ready to go system at this price range or do i have to assemble it only. I plan to run scripts as well. So i guess i would be needing as much as ram as possible in my budget.