Is there a database of GTO stats?

It depends on lot of factors, so its tough to give some definitive GTO numbers. But what you can look into other than obviously preflop charts, is aggregate reports for flops. You can use tools like GTO Wizard or DTO to give you some idea about how often to cbet, what sizings to use the most often etc. But as I already said, there are lot of factors going into it, some stats can be interchangeable pretty well (you can play higher flat, lower 3bet or lower flat, higher 3bet and both strategies can have pretty much the same EV)...

You can run aggregated reports in Pio or generate Pio HHs and import those into a hud

You can see Pio is way below MDF in this line

Mobius Poker has a GTO stat checker with nearly every spot mapped out with GTO frequencies. I have no affiliation with them, but you can buy it here: https://www.mobiuspoker.com/stat-checker

Video here:

Doesn't seem to be anyway to contact the anyone on that site. Do you know if GTO stat checker works with anything besides hand2note?

You have to use hand2note.

Their roadmap says aggregated reports for all spots are coming soon, so until they implement a full stat checker system, that would work well.

I realize this is an old thread, but has anyone actually looked at these stats? They look pretty horrendous to me, I'm almost certain they're nowhere near GTO. Seeing these stats, along with the fact that the bot actually lost money vs. the humans and the researchers had to use some variance reduction calculation to 'prove' that the bot would have been a winner, makes me highly skeptical that this bot was indicative of any major progress in poker AI or GTO poker strategy. (Supposedly the technique they used meant they could get a statistically significant winrate from just 10k hands, and they used it to say the bot was a 5-6bb/100 winner when it was actually a 7bb/100 loser over the 10k hands). I suspect the researchers were just trying to fluff their own nuts, so to speak.

I don't think we can say it is or it is not horrendous (I assume horrendous being far off from true GTO) when we do not know what is the GTO in the first place.

A few things to keep in mind:

There was zero rake, which probably explains preflop.
For some reason they decided Pluribus should not have a small sizing for the flop cbet.

Those are the ones I remember.

10k hands is nothing, I would definitely pay zero attention to the actual win, or loss in this case.

I don't think it's fair to dismiss Pluribus' results just because they used a variance reduction technique.

AIVAT is unbiased, mathematically sound, and the process was agreed to before the match even started (not some ad hoc p-hack). It was peer reviewed by scientists who are for more rigorous than 2+2 degens.

Furthermore, 10k hands is statistically meaningful when you reduce variance by 10x. You can see standard deviations in their paper.

Not really a database, but you can now display player stats for any line in the latest HRC beta.

UI for that feature is still rough though, heh. [That screenshot is for some 25bb ICM hand, fwiw]

Pretty cool, but looks like the stats are aggregate so it includes single raised pot, 3bp etc. all into one? Is there plans to maybe allow it to be filtered a bit?

Yeah, down the line we'll do filtering. You can already select specific lines and check the stats for that specific line, but checking stats over e.g. all 3-bet lines in aggregate isn't directly possible for now.