Solving speed
When it comes to CPU, the solving speed will essentially be determined by the number of physical cores multiplied by the processing speed. So if a computer has 4 cores of 3.2 gHz, then this will come down to 4*3.2=12.8gHz. Please note that processors are often advertised with the number of virtual cores, whereas you need to look at the actual number of physical cores.

Memory
RAM will have no influence over solving speed, but will only determine the largest size of tree that can be fit into memory. We recommend having about 8GB available, which should be sufficient for any reasonable tree.

Types/brands
I can't really give any advice when it comes to types or brands of CPU. As long as the processor is fairly recent (so not being more than 5 years old) then there should not be any issues. Older processors may provide slower solving speeds due to them not being able to quickly perform operations on large chunks of memory.

Hi, got 2 noob questions:
1. Why does GTO+ output strategies that look less mixed than in pio (for example more pure calling/pure folding particular combos whereas in pio there's always a lot of mixing). Is it due to a different alghoritm being used or just some auto-rounding? If the latter, does it sacrifice much of EV?
2. How significantly inaccurate will the strategies on later streets be if I never recalculate them clicking on that grey-blue circle?

Regards

Hey Scylla, I ended up solving BTNvsBB and ran solves for separate trees like you said.

Flop is AJ6r - Here are the results:

BTN opens 2.5xBB, BB calls. 100BB effective.

Flop is AJ6r, BB should check his whole range.

BTN's Bet Sizing EV: Highest EV is preferred sizing -

33% Sizing - 3.65EV
50% Sizing - 3.65EV
75% Sizing - 3.66EV
100% Sizing - 3.67EV
125% Sizing - 3.72EV
150% Sizing - 3.72EV
175% Sizing - 3.72EV
200% Sizing - 3.72EV
--------->>>225% Sizing - 3.83 EV<<<--------
250% Sizing - 3.81EV
300% Sizing - 3.71EV

225% Cbet Sizing is Highest EV on AhJc6d BTNvsBB


Question about these actual results:

What is the difference in real money between an EV of 3.83 and 3.65?

It will be due to the algorithms being different. Within any given Nash distance (dEV) multiple solutions will exist. You're probably looking at a selection effect though. In other solutions you'll find more mixing for GTO+ as opposed to piosolver. As the Nash difference reaches 0 the difference in solutions between the two programs will disappear.


No EV is sacrificed.
Both solutions will be valid within the given Nash distance (dEV).



The solution that is provided by default is already highly valid strategy, and far beyond what any human would be capable of.
Clicking the circle will allow you to go even further, and solve all the way to 0%.

The difference is 18 cents (or roughly 5%). The difference in the more "reasonable" bet size range of 40% to 100% of the pot is about 2 cents (or 0.5%). The amount of detail that is lost in translating Nash solutions to something that can be used in practical play can be expected to be far larger than that though, so I would recommend just picking a bet size that you feel would be most applicable to your situation and working with that.

5% does seem like a decent amount of EV though right? If you’re playing 1KNL and the pot is 60 bucks OTF. I’ll take that 3 dollar increase in EV bucks all day. And I feel like if you could identify all the situations where an Overbet is preferred over a reasonable bet, you could increase your winrate at least 1-2BB/100?

Or is this identifying Overbet spots on flops a futile effort?

Thanks so much for interacting with your customers. It helps a lot in understanding the practical side of GTO+ and also the theoretical side.

5% does seem larger than what I would expect. Once again, in the range that is considered "reasonable" of betting between roughly 40% of the pot and 100% of the pot the difference between worst and best performancing bet sizes is roughly 0.5%. It's surprising that huge overbets would perform so well, so personally I would try and figure out if this is consistent across ranges/flops/stacksizes, or if there's very specific situations where this applies. I'd thread a bit carefully though, given that this seems to be far away from conventional wisdom.

Thanks scylla! So I am trying to get the most accurate information possible. Can you explain the different options in this picture?

Is this the distance to Nash equilibrium? And if so, is .1% more accurate than .5%?

Yes, "Target dEV" is the Nash distance. In the screenshot it's 2.5 cents. This will mean that, in the current solution, if either OOP or IP were to maximally exploit the small mistakes in their opponent's strategy, they would only gain at most 2.5 cents in EV. A Nash distance of .1% (0.5 cents) is indeed better than .5% (2.5 cents). The exact solution is found at a Nash distance of 0.

Hi Scylla,

Is there a way to combine two databases if they have been calculated on two different computers? (e.g. half of the flops on one pc and the other half on the other)

Thanks Scylla! I'm sorry but I can't help but pick your brain .

I had another question - I've read there are 22,100 possible flops in NLHE. And 1755 are strategically different.

Do you know where I could find these "strategically different" flop sets?

https://www.piosolver.com/blogs/news...the-whole-game


Sent from my iPhone using Tapatalk

Thank you for the link. I don't see the 1755 flops here though? Just the 184 that are supposed to reflect the 1755.

I would really like the 1755

Google 1755 poker flops. There’s a link to a post on 2+2 that has all 1755 flops


Sent from my iPhone using Tapatalk

Not at the moment.
I will however consider it for later releases.

Here's a file with all 1755 flops: www.crevfiles.com/gto/1755.txt

Hi, do you have plans for preflop solutions in the future?

+1

They usually don't discuss plans for future releases.

Thank you!

To put it another way, that 0.18 difference equates to 18BB/100 hands, I think. At least that's how I saw a pro on a training site calculate it.

Damn that seems like a lot of increase to your winrate?

The difference is 18 cents.
However, you're not in this spot 100 hands in a row.
The concept of BB/100 only applies preflop, prior to posting the blinds.

When you say 18 cents, that's assuming blinds are $0.50/$1.00 right? While you won't be in this exact spot 100 times in a row, aren't you supposed to treat any spot as if it were going to happen an infinite number of times?

It does seem like a lot (and it is!) but like scylla mentioned, you only profit 18 cents each time.