Yes, 32 cores should run twice as fast as 16 cores.
This only applies to physical cores though.

This refers to physical cores. The optimal number of threads is usually 2x the number of physical cores. Beyond that point using more threads will not result in more speedup.

Thanks for the quick reply!
So if you have for example 8 physical cores and you can use hyperthreading up to 16 cores (and you set up the solver for 16 cores to work on), it will still take twice as long to solve a tree than when you are using a CPU with 16 cores and you allow the program to use only 16 cores?

A CPU with 16 cores, using 32 threads, will be twice as fast as 8 cores and 16 threads.
A CPU with 16 cores using only 16 threads will run roughly 30% slower than with 32 threads.

Okay, thanks!

Why the tree builders are different? And in wich point they differ from each other?

I remember running this board on old version of GTO+ and getting results that are in line with Pio: https://gyazo.com/c6c663e366bb507e7fedee979d6a5a62 the mixes are completely different from the pio solution.

https://gyazo.com/5f8ef7d2518b93db4f780db2761001f8 all the zero EV calls/raises get pure raised in the GTO+ solution (the 44, 66, AK, etc.)

I have a lot of other sims that have super extreme distortions like this...going back to old version of GTO+ until there's some update.

Are the ranges even the same here? They look different. It's hard to tell because you use "All at 100%" display option, but the pio screenshot has hands like K6s that don't appear at all in the GTO+ screenshot. Also the pio screenshot appears to have 113.6 total combos in the range, while the GTO+ screenshot shows 67.59 total combos.

The ranges are 100% identical. The fact that I have it set to "All it 100%" doesn't affect the actions it's just a display.

Look at the overall frequency. IP raising the same board 42% in GTO+ and 14.6% in Pio should not be a thing. I'm going to run the same board in 1.28 now.

edit: nm I can't load the save from a newer version of GTO+ but it's certainly wrong at least for this subset...I had ran it before on the old version and now the frequencies are completely different despite no change in the ranges or strategy (I ran it again so I could see the turn and river reports).

Cellar door also having similar issues as me.

A9s-A7s,K9s,K6s,AJo,KQo,T8s are all not present in the GTO+ sim, but they are in the Pio sim. As you say, if they were present they would be showed at 100% given the display option, but they are not shown at all.

can you add dark theme ? there is too much white area and it killing my eyes when i try to study solver in night, also can you add option to recalculate all turns? when i use edit tree and add or remove sizings and then i want to look in turn aggregated reports it ask me to recalculate whole flop again and it take lot of time, would be awesome to have button for recalculating all turns or rivers

right now iam missing:
1. dark theme
2. button for recalculating all turns and rivers
3. no draw category (trash hands)
4. hotkeys for moving turn and river cards and also without losing node iam in, lets say iam in node where IP cbet flop turn and now i want to look how he cbet river, when i change river card it will move me back to OOP check and have to click check so i can see strategy, if i can simply just using arrows on keyboard and changing river cards and also staying in same node it would be really awesome and save lot of time and clicking for me
5. sending data to all flopzilla pro i have opened (or at least 2)
6. locking different groups of hands in bet and check range as well same like in entire decision (example: locking TP and flushdraw together i can do it in entire decision but not in specific like bet or check)

These ranges do not appear to be the same. Certain hands are present in the pio range that are not present in the GTO+ range (they should show up, given that the display is at 100%). In total, pio seems to have 16.6+67.3+29.7=113.6 combos, whereas GTO+ has only 67.59 combos. Also I notice that there's a difference in the bet sizing (208.7 as opposed to 208; this indicates that there's likely other differences as well). If you compare two different spots, then it does not seem unreasonable that the results are different. In this case, both the ranges and the tree are different.

I should be able to compare two versions for you though. Would it be possible for you to send a savefile to support? It should allow me to check which version it was originally, and with any luck make a comparison. V128 should be able to load savefiles from all previous versions, except for one that was only up very briefly. We posted a fix within a few hours, but it's possible that you're using that version. So, if you could please send a savefile, then I should be able to take a look.

We decided to build our own tree builder as opposed to reverse engineering that of someone else.
So there are differences.

Our "Basic" tree builder is the same as their "V1" tree builder though.
In this case the results from both solvers are nearly identical.

Ok, I will consider this for future releases.

I 2nd this. I have been looking for this feature too.

Hey Scylla,

I've solved several databases, using the same tree.

Is it possible to merge those solves so that I can play against the solution for both?

Yes, there's two ways of merging files.

Merge manually
If it's a small number of files, then you can merge them manually.

For this:
1) Load a file
2) Go to "Database" mode
3) Click on the "Lock" icon (see screenshot)
4) Load the next file
5) Go to 2)





Merge automatically (v129 required)
If the number of files is very large, then place all the files in a single directory.
After that, enter the code MERGE after "Import flops from file".
This will bring up a menu that will merge the contents of all these files for you into a single database.

I'm trying to solve very simple toy games in order to understand how equilibrium works.
I'm using the trial version, so a single flop is available.. But well, I need only to solve a spot on the river. Is there a way to model a simple bluff catching range vs polar?

The restriction for the trial version only applies to the flop.
You can enter any board that you like if you start the hand at the turn or river.
As for a bluff catching range vs polar, for this you can just enter a medium sized hand/range versus nut+air.

Ty a lot, I'm pretty excited to work on your software.
May I ask you whether do you plan a dark mode?

The ranges are 100% identical. I literally imported them from the same source into both Pio and GTO+.

Here are the ranges used in each sim:


As you can clearly see the suited Ax, T8s, the Kx you mentioned are all present in the GTO+ range. Any difference is just due to display settings. Regardless, even if they weren't present, they would not result in such massive postflop adjustments for the IP player. However, I fully expect to be told that it's 138.68 combos in GTO+ and 138.7 combos in Pio and that's why I am seeing 3x as much raising in GTO+ compared to Pio.

If a one chip difference in sizing and maybe 1-2 extra sizings (I tried to make the sims as close as possible...I think the 209 vs 208 is due to initial pot size) results in a massive increase in raising frequencies or causes certain turns to not be overbet when they are always overbet in Pio then I am pretty skeptical of this software.

In the screenshot you showed earlier IP had 67.59 combos. See the pic below. When making a comparison, please ensure that the two spots that you compare are identical. For the moment only our "Basic" tree builder and the "V1" tree builder from pio create identical trees. In that case the results will indeed be close to the same. If using the "Advanced" tree builder from both programs (with same ranges, pot, etc) then there will be all sorts of subtle differences between the trees. On top of that, when using multiple bet sizes, within any given Nash distance, many different frequency configurations are possible. This is because all bet sizes have almost the exact same EV performance. Even if ranges, pot, bets etc are exactly identical, then there's still not necessarily much to be concluded from the frequencies being different. Many different answers are possible within a non-zero Nash distance. That being said, once again, if you want me to take a closer look, then please send a savefile to support.

Ok, I will consider it for future releases.

Hi scylla, when i'm node locking certain decision in the flop and then re calculating dEV on the turn the EV tab does not appear on the flop, am I doing something wrong? how can I see the EV tab for flop after I node lock a decision in the flop and want to check the EV changes?

hello i found one bug, when i merged many databases that use different sizings on flop and then i go to database and display results in graph there is missing flop filter option which would be very useful, can you add it please

At the moment the flop filter functions are only available in case aggregate reports are possible for the entire database. We can consider expanding this for later releases, but right at this moment, due to the internal workings of the interface, this is the approach that has been used.