Doesn't seem to be working. It's the first time I've opened the file location so I'm not sure why the ranges have randomly gone missing (haven't updated GTO+ lately but using latest version). It appears every time I close GTO+ the ranges disappear.

Sent from my G8441 using Tapatalk

Thank you that did the job.
But do I need to do that (merge files) every time I want to play vs the solutions?

Yes, it appears to be in a specific spot I believe, as far as I can tell although admittedly I haven't been taking notes on crashes. I just fire up the program again and just brush it off. I'm not sure how much a save file will help you ?Because it happens with more than one (163 flop) database file. As far as I can remember, and I don't want to say 'only' or 'always', but it tends to happen while going through a hand history, reviewing the solution, and then, and only sometimes, when I click the two cards icon to go back to playing against the solution, the program will hang up for a second or two before it closes out. The only other programs I usually have open while using GTO+ is Brave web browser running 1-3 tabs on average, one of them being a youtube music thing. System overhead at this point as far as RAM goes is around 45% of 16GB. CPU usage is miniscule. I'm not tech savvy enough to know how to use event viewer to help me either, or if it would even help diagnose. Maybe I should read up on it so the next time it happens I am prepared? Just spit balling here. Any ideas?

Edit: I should also point out that my system is 24/7 stable otherwise. No overclocks are enabled expect for XMP on the memory ( I supposed a tad concerning).

It sounds like some sort of bug that hasn't been detected yet. I haven't personally experienced this, but most likely we'll run across it somewhere in 2021 and fix it. Without knowing any specifics it may be difficult to track down right now.

If you want to play against a set of files, then merging them all together would indeed be the way to approach this.

Can you send one of the files that you're trying to restore to support?

Okay, no worries. I appreciate the software as-is anyway.

Another question however, if you will:
In a sub-set, I am told they are designed to approximate "all" possible flops in order to approximate expected-value. My question is, if a sub-set says that OOP is 'donk'-leading say 3.06% on average across 163 flops, does this average percentage approximate the meta 1,000 whatever strategically unique flops ? Or are sub-sets not designed to calculate these sorts of averages? Is this 3.06% only an average across these particular 163 flops? You advice would be helpful. Thank you in advance (so I don't have to muddy the comments later on).

It's an average across the 163 flops, however, I would expect this average to be close to the average over the extensive set of 1755 flops.

Hi scylla,
If I'm interested in exploring flop spot over decent db can I build a tree that ends on the flop or the only way is to manually remove all the turn and river branches?

hi trying to open a database i ran last year but have since updated the program and i get the error msg 'sorry the trial verson can only load savefiles that wre created by another trial version'? i never even had a trial version i just purchased it from the get go so bit confused. any work arounds? thx

A tree will always need branches on the turn/river in order to calculate what happens when those lines are reached. So it's not possible to make a tree end on the flop.

You can simply request a new key for GTO+ here: www.gtoplus.com/request-a-key
It will be sent to you within a few minutes of your request.

Is there any merit in having a merged database of actual hands that we want to review and then play against the solution to correct some leaks ?

Say I want to check my bluffing frequency on the river in 3 Barrels spots and I import in a file X flops won/lost.

Can I lock the runouts ? Should I ?

Idea would be to either review how I played certain formations (same tree) over a period of time or review a few formations (different trees in different files) on a particular session.

Alternatively, is it possible to play against the solution to review a database of flops with different trees which have been solved separately ?

Any thoughts ?

Tia

There's many ways to use the "Play against the solution" feature.
However, I can't really provide any advice in what would work for you personally.

Hi scylla,
I ran a sim during the night which finished successfully, I saved it and trying to open it now but when the software loading the file there is a pop up says "unexpected file format." and crashes, every time I try to open this file this message pops up.
what can I do and how should I send the file to support?

I have two servers:
1. (Released 2019)3960x (24 Cores 48 Threads) 3.8GHz Base Speed
2. (Released 2012) Dual E5-2670 (16 cores 32 Threads Total) 2.6GHz Base Speed

Both finished solving at 1% dEV with the exact same sim with times:
379.08 seconds and 398.18 seconds respectively with a surprising difference of
just 4.9%.

What do you think could be the cause of this? It is quite unlikely that this Intel CPU IPC is higher than this new Threadripper since this CPU is almost a decade old already.

Im using the newest GTO+ v1.3.3

EDIT: I just tried CPU-z just a simple benchmark, I guess it's pretty obvious who is the winner here. I used the (48override) option for Threadripper so it seems like the problem is in GTO+.

I will be trying another solver just to compare results

Additional info:

Used another solver, there is definitely around 30% increase in performance. Not what I was expecting but the difference was there. With GTO+ (6+) unfortunately, im having the same result in solve times. The difference that I get is from loading from one flop to the other. On my Threadripper, it was clearly visible that switching from Flop 1 to Flop 2 was fast compared to the other solver. But solving wise, it looks similar as far as I can tell.

Hope you can look into this.

Did some tests on my own and here are my results when running two sims tested on 3 systems:

250MB tree at 0.5%

My PC (Ryzen 2600) 18.39s 19.14s
3960x (48 override) 9.71s 9.71s
3960x (32) 8.71s 8.67s
3960x (40 override) 8.95s 8.90s
E52670 (32) 12.64s 12.32s

750MB Tree at 0.5%

My PC (Ryzen 2600) 114.75s 120.17s
3960x (48 override) 45.90s 46.29s
3960x (32) 42.81s 44.21s
3960x (40 override) 43.24s 43.77
E5 2670 (32) 63.76s 64.97s
E5 2670 (24) 72.36s 71.43s

Looks like anything above 32 doesn't really do much more with GTO+ in fact it made the times worse.

I just tried a quick test and ran two 250MB tree sims on my Threadripper and it finished around 15seconds which is a bit faster. But on 750MB the last one finished 99 seconds which is more than twice. I set the threads for these ones at 24 each. Something is also going on the bigger the tree is.

Wish you guys could do more testing with these. I'll prob just get a 3950x or even 3900x server next time since the program doesn't really scale well and not worth it to get the higher core count CPUs.

You can try sending the file to support, however, is it of a managable size?

A reasonable explanation here would be that at a certain point there's a bottleneck in your system, other than your CPU. Even an inifinitely fast CPU does not necessarily mean an infinite solving speed. At a certain point other components in your computer will set the upper limit to speed.

3.5GB and the file originally was supposed to be 6GB. Not managable I would say.
I didn't turned off my PC so the files on the tmp folder are still exist I assume, is there a way to reproduce them without resolving? It was an 184 flops db.

Bottleneck really? I don't mean to sound rude but that doesn't sound a reasonable explanation at all after I did some testing and posted results.

Have you tried testing on systems with more than 32 cores using the override option yourself?

How could you just say my system is bottlenecked when I just explained in my previous post that I used another solver and it was faster the more threads I set it to being 48 the max. While the increase wasn't linear, there definitely was difference every thread I added onto it. With GTO+ anything above 32 using the override option I did not get any additional speed in fact it made things slower.

How can you then explain why it's running same or worse on 48 threads with 100% utilization vs 32 threads around 75% ultilization
But on the other solver, I get similar utilization percentages but there was definitely an increase in the amount of calcs it can do per second.

I understand not getting 50% more speed from 32 threads to 48 threads but not getting any difference seems off to me.

I can not say for an absolute fact why you wouldn't see anymore speedup above 32 threads. This is because there's dozens or even hundreds of moving parts involved, making it difficult to narrow down why a particular effect would happen on a particular computer. That being said, comparing to another solver doesn't have to mean anything. GTO+ is roughly 2x faster than pio, so it's not unthinkable that at 64 threads you would see the same effect (provided a bottleneck is involved). In the end it's not very straightforward to just compare two different algorithms.

I don't immediately see a workaround here. However, in the future, when working on very large databases, it may be better to export it into its separate trees, solve those trees, and then merge them together again after solving. For more details on this approach, go here: https://www.gtoplus.com/processingdatabase/

Have you tested on systems that had more than 32 threads? Did you have better results when it was more than 32? If you did could you please let me know which CPUs you tested these on so I could have a better informed decision when getting another server in the future. I was thinking if whether or not to get a 40 core intel alternative. I really want to get the fastest server I can get my hands on. Do you mind sharing what algorithm GTO+ uses? It's possible it runs better on intel than on AMD CPUs. I have been seeing a lot of threads(forum, no pun intended) online where intel just out performs AMD with even less threads(core count) just because it is more optimized.