It should be pretty close I guess. You would need to experiment with exact timings. The problem is that it takes 3x RAM to run 3 instances at the same time.

Is there a way to add "2 card backdoor flushdraw" (such as in flopzilla) in the range explorer?

If not, I think that'd make the range explorer perfect.

Thanks

Hi Punter. I have a tree for heads up play (with ante) at 25bb deep. For the SB--i.e. the IP player--AA, KK, and QQ are displayed as pure raises (2bb sizing). See first image.

However, I was alarmed to see that when I checked the EV of the various actions, calling was being shown as *significantly* superior to raising, despite the fact that the combos are all *pure raising*. It almost seems like a glitch--whatever it is, I cannot figure it out. See second image.

Also note that the exploitability isn't particularly high--0.48bb/100 translates to roughly 2.8% of the pot in exploitability, per hand. See third image.

Any ideas? Thank you.

Yeah, I agree this one is the most important out of backdoor draws.
For now you can add those by first checking the checkboxes you want and then clicking "selection" 13x13 area. There just choose the suit and add all the suited hands of that suit. It's more clicking and not that convenient but you can make Range Explorer to show you stats for the range with a backdoor included.

Yeah it looks like something went wrong there. I will need the tree config to take a look at re-run it here.

Just messaged you the tree config. Let me know if you need anything else. Thank you.

for some reason i only get 89% CPU utilization when running a simulation, I remember getting 100% a couple months ago on the same PC (older i7 4core cpu) I have since reinstalled windows and updated pio, power saving settings are modified for max power, this is on all cores

also tried a different solver - that one uses 100% constantly and ran the same sim 20% faster (but uses a lot more ram so I would really prefer to use pio)

is there something i can do?

Are you having any heating issues? CPU's will throttle at a certain temp. If you are, applying some new thermal paste to your CPU should help. Running less processes would help too

throttling due to temperature lowers the clock, the cpu should still be utilized at 100% it just runs slower - anyway that is not the problem with my system

Pio rarely uses 100% of the CPU because of how multithreading is implemented. 89% is great. It doesn't mean it could be 11% faster, it couldn't.

We will have significant performance improvements in the very near future.

Hi, when im running a sim with ICM factored in running until 0.1 and ive noticed that the number jumps back up.

For example it was at 0.461 then went to 0.452 but then jumped back to 0.464

Yeah, it may happen.
Notice that with ICM there are two problems:

1)It's not a zero-sum game so there is more than one equilibrium which makes it difficult for the solver

2)It's not trivial to scale ICM values to usual accuracy in %. Usually you should strive to run the tree for as long as you would run a normal one. The solution is probably quite good by then.

Another question about weights for flops.

Do they remain the same when separated from the main list of subsets?

I have run the 184 flop subset with two betsizes with the goal of choosing one size per flop.

So when I re-run the 184 flops divided into two different groups, should I still use the same weights as when they were all in the same list?

Weights in a subset only make sense in combination. If you take a subset of a subset you will not get any kind of sensible results for aggregated EV/EQ.
All the individual solutions (for particular flops) are going to be correct though. It's just aggregation that doesn't work anymore.

I ran a script with 1755 flops and some of the trees accuracy is higher than the .50 I wanted, I think because I set timeout to 6000 seconds. I'd like to run those again but I'm curious if there's a way to see which ones have poor accuracy without opening each tree individually.

Thank you

Unfortunately no way to do it right now.

OK thanks. Can I use a subset of a subset to get averages of frequencies?

I'm confused about how raise sizing works. In this sim I have given IP the option to raise to 2.75x, but when the sim plays out IP only raises for the minimum.

The idea of weights is to give average frequencies/EV/equity possibly close to all 1755 boards. Once you split the subset it will not work.


You forgot an "x" there. 2.75 alone means 2.75% which is rounded to minimum size raise.

Oops, thanks! Stupid mistake.

About the weights, I know I'm being slow, but I still don't understand them.

I have a folder with all the K high boards from the 184 flop subset solved for SRP BTN vs BB.

When I run an aggregate report for various nodes, it gives me the average frequencies.

It looks like these frequencies are taken from the sims, so I don't really understand, are they not accurate because they are not weighted?

Can I use an aggregate report from this folder to find out, for example, how often in position folds to a check raise?

Some questions:

1. I tried to open some trees solved and saved with PioSolver Pro (small save without river) with PioViewer Free, but it gives me the following error:




2. Can you explain what the "3bet-call" order criteria are in "preflop hands order"? (Position, Rake, Stack, ecc.)




3. In the faq I read that the PRO version makes a maximum of 16 threads, soon I should change the cpu. With a Ryzen 5950x (16 core / 32 threads) I would have approximately the same performance as with the 5800x (8 core / 16 threads)?

very good program !

Hi Punter

I want to work on my laptop with Pio but the resolution is 1080p and when I open more than two instances of Pio i cant do my work. They overlap and cant see good. Is there a secret way to shrink it or do something about it ?

Thank you!

Frequencies are taken from the script if the script is present. Otherwise it just uses 1 for every board in the folder. You can see what it uses in the aggregation report window (assuming you are on the newest version).
The weights are constructed with one idea in mind: to approximate EV and EQ for the whole game if you use the subset with those weights. Once you split the subset they don't fulfill their purposes anymore.

You need a commercial version to open saves.

Unfortunately I really don't remember. Those were created as examples very long time ago from solving a lot of flops or preflop sims. You can create your own orders if you want.

No. It will be about 2x as fast on the 16core machine as it has twice as many cores.
Speed-up for hyperthreading (ability to run 2 hardware threads on the same core) is negligible and varies from 0 to around 25% depending on the application. It's around 15% for Pio from what I remmeber right now and will likely be lower in the future.
tldr: only physical cores count.

You can shrink the window if you want but not the fonts etc.

Can you re upload the piocloud solutions ? or this is not supported by piosolver ?

https://piocloud.******.com/download-links.html

I am sorry I haven't checked them in a while.
I will figure out what happened to them. We certainly want to continue offering them for free. It may take a few days to re-upload somewhere else or fix the links. Feel free to pm me if that doesn't happen till next Tuesday.
I remember this happened in the past due to download limits on Dropbox. Anyway, please give me some time to figur this out.