We could create a workaround for this, but I think that it may be better to keep the ability to edit trees in one central place in the program.

Oki Scy, im understand that.
That twist may not be so relevant.

I'm sorry for always bothering you, but since I know you read and take note of the things that can be improved, I take the opportunity to bring ideas from what I see in other GTO programs that could be implemented here.

My go to solver is GTO+ ​​by far, but it would lack the filter by suit, the multiple filter selector and the display mode bar. (I know, I'm asking a lot )

With those better ones, no one with half a brain would choose another solver.


In this image I upload another "idea" or + filters that are used for river nodes. I think it will very usefull


https://gyazo.com/938e51644270b32fdd1ba12f850c274f

FB = Flush blocker
SB = Straight Blocker
FB+Made = Flushblocker + made hand (good for looking into thin value)
SD+Made = Straight Blocker + Made hand (the same)

Hi Scylla,

I was building a tree with a few betsizings and typed in this syntax for OOP on the turn: "33,75p,75c,150p,150c" indicating that I wanted to build in a 33% sizing regardless of the situation (probe, donk, or continuation all ok), and a 75% and 150% sizing in only the probe and continuation lines.

However, when I solved and examined the tree, the only thing that appeared after flop went check-check was that the OOP player could check, bet 75%, or bet 150%. The 33% sizing was missing. Why did it do this? Is it because I specified probe and continuation for the other 2 sizings, and once it saw that, it needed specification for the 33% as well? Is ""33c,33p,33d,75p,75c,150p,150c" the proper syntax instead?

Ok, I will consider it for future releases.

Yes, the second syntax is indeed what it needed.
Once custom input is given with 75p,150p it will be assumed that you're looking for a specific setting for probe bets.

Got it. Guess I have a large number of trees to re-run, then. Was trying to look for optimal ways to speed up the process and one thing that occurred to me was that I wasn't optimizing my desktop for these sims. For instance, I saw a post that said you could take your CPU cores, multiply that by 2, and put it in the number of threads box, i.e. I have 6 cores so input "12" for threads when I'm processing a database. Is that correct? I also have a RTX 2060 GPU - is there a way to involve that as well?

With 6 cores, most likely the optimal number of threads will indeed be 12.
This is however not always the case; for example a processor can also have 16 cores and 24 threads.
You can test the optimal number of threads by checking with the task manager at which point CPU usage hits ~100%.

GTO+ works on the CPU and not the GPU.

Hi guys!

I need to create a subset to play against solution BB defend (cc 29%) vs BTN (rfi 43%) (ranges for low stakes)

Just finished calculating subset 163 flop and got a strange result. A nit on the button. Total checks flop IP 60% and have such a low total total cbet 40%. It doesn't match up with the way real people play.

My settings was:
target dev 0.1%
for a player in position cbet flop 33,50,75. Turn 50,75 and river 50, 75.
For a player out of position 0 donks, turn 30p, 70p, 150p and river 30c, 66c, 1000c, 33p,66p,100p


Can you explain to me why BTN checks flop IP 60% on all flops and have such a low total cbet 40%.
Because of the number of bet sizings for a player in position? What do I need to change to raise the cbet frequency to 60-80%

Thank you!

Hi Scylla,
In my site %rake is taken from effective pot. They explain the effective pot = size bet (call) + current pot. Examples player A bet 500 into pot 1000 and win so effective pot is 1500. Is it the same way that %rake from pot in GtO+ calculated?

I checked and I have 6 cores and 6 logical processors (I guess these are the threads). Sure enough, at 6 threads my CPU is running at 100% util. I'm not a tech genius or anything, so I was wondering - if I upgraded the CPU to one with say, 2x the number of threads, does that correspond with a 2x faster solve time? Or does the processing speed not scale linearly with the number of threads?

Hi scylla,

Am I right to say that currently if I want to check turn lines from a database level, I need to click report and find it on each individual flop and aggregate them manually?


If so, are there plans for aggregated report for turn/river in database mode? This would help to check turn probe frequencies, what to do vs probes frequencies as well.

I can't really tell much from the screenshot.
It's not possible to see which tree has been built, which ranges have been used, etc.
If you want me to take a closer look, then please send a savefile to support.
Store it with "File->Save: Convert to Basic storage" to bring down its size.

The problem with turn/river aggregated data is that it's an average of an average (averaged over all turns, averaged over all flops).
Also, all measuring points are reached with completely different frequencies.
In one flop/turn combination, a point may be reached 60% of the time, whereas it's reached 0% of the time in another.
There isn't really that much that can be done with this data, and it's very easy to misinterpret it.
We can consider adding it for future releases, although personally I think that it's better to leave it out entirely.

In regular rake calculation, the 500 would be returned to the player prior to determining the pot.
This is also how GTO+ determines the pot.
However, you're saying that they explain "the effective pot = size bet (call) + current pot".
So, it says "(call)".
I think this means that if the pot is 1000, player A bets 500 and player B calls, then the pot becomes 1000 + 500 + 500 = 2000.

The processing speed scales linearly with the number of physical cores, and their processing speed (gHz).
So a CPU with 4 cores of 3.2 gHz comes down to a 12.8 gHz system.
I would focus on cores, and not threads.

twiceing threads does not twice performance

Correct me if I'm wrong, but my plans was to look over the averages to figure out where will differ the most, then I will drill into individual tress to look at.

For example, SRP, BB v UTG and BB v UTG, if I can see the average of BB turn probe line, I'd imagine there'd be some tress that probe more and others that probe less, then I can drill into the trees where I can probe more and the turns that I should do most with. Right now, if I want to find a turn to study on a board, I need to manually select one a few and guess for the others by proxy of these few. I just think that maybe with that average of average, I can have a different way to find other spots that I might miss.

But I can see your point as well and I'm not sure if it would make too much difference to my growth if I can see the report vs if I can't. But as I type this out, I think eventually I will have to find out for myself and I love to be able to do it on GTO+. Cheers

Hi Scylla,

when I play against the solution the results of the session are not saved in "session filters".
is there a way to save the results?

Ok, I will consider it for future releases.

Hands will only be saved when they have been played until showdown.
Are you perhaps using F1 or F2 to skip to the next hand in the middle of a hand?

Solving speed increases linearly with the number of cores; not the number of threads.
This means that an identical CPU, with 2x the cores, will lead to 2x the solving speed.
The number of threads can just be used to set which % of CPU you want to assign to the solver.

well, i tried the same test but now with a core affinity. still, no speed doubling. maybe AMD specific though.

That's rather expected though. If you are limiting a 5900x to 12T/6C, the cores that are used can be boosting higher/longer before hitting thermal limits, as half of the chip just sits idle. So the per-core/per-thread output is going to be a bit higher.

As plexiq points out, this is indeed dependent on how this test is run.
When going from 3 cores to 6 cores, the solving speed drops from 450 to 235, which is close to 2x faster.
But when going from 6 cores to 12 cores, 100% of the computers capabilities are now in use.
I would not expect this test to work perfectly when performed on a single computer.

No. I do not skip to the next hand. I play until showdown.