Hi, I saw your post on 2p2, had a couple questions:
1. Which sites/games does the data come from?
2. For PF info, how do you account for the fact that hand which reach showdown are a biased sample? Eg, a fish's trash range will on average fold more often than their strong hands, and so a raw sample of just shown-down hands would underestimate their frequency in his range.
Hi !
1. GG, PS.com, 888, Ipoker, WPN. I'll be adding more and more hands as it goes.
2. PF debiaising is quite interesting. So we know good hands showdown more often than bad ones. So we use our GTO solutions to find the probability of showdown for each combo in each specific node of the game tree, and we apply it to what we observe in the hand histories. We've run tests and the debiased ranges fold/call/raise percentages land very close to the real stats we observe. Usually within 2%. So it's not perfect, but it helps a lot visualizing ranges.
Jonah · 18:22
Got it. In essence, use GTO as your "oracle" (can always see hole cards) and use that to take showdown ranges in real online data (no holecards on folded/mucked) and backwards infer the PF ranges.
My big question would be, for truly fishy players, whose PF ranges are of most interest, the ranges are so much wider than GTO that you simply won't have GTO data to compare it to, because GTO would never play those hands. Eg, people limping A2o or 87o or 53s up front, etc.... How do you deal with that?
Yes so at the moment we apply the average showdown weight on the hands that are not in the GTO range. Which is a bit wrong. We under represent trash hands in Fish ranges. It is conservative but not the best approach. So we are looking at applying the closest weight. Let's say GTO has 98o, we will apply this weight to the fish 64o. And if GTO doesn't have a strong hand, like AA for example, that fish love to slowplay, then we would apply the weight of let's say JJ if it's in the GTO range. We are working on it so we should have it updated soon.
But if you look at the debiased ranges right now, at best a better version would change the frequencies of combos, but you would still see a quite similar range.
Wednesday 7 Oct
kaytor · 17:24
Hi 🙂
I play MTTs exclusively, could you detail what is supported for MTTs and what is not supported at the moment ?
I saw that GTO comparison is not available yet, do you have approximate time frame for this ?
Would you be my beta testor ? I need feedback for the MTTs. I'm specialized in Cash games, and know very little about MTTs. So I could give you free access for a few days and in exchange you tell me what you would like to see. What do you say ?
Hey guys ! I've been quiet for some time as I was fixing some bugs and rebuilt the whole database.
40Million new fresh cash games hands were imported (PS, 888 and ipoker).
PLO is around the corner, I should release it next week.
And I'm planning to import more MTT hands. Let me know if you have any preferences
Many updates are coming to the app ! Details will be published in #announcements
No trials yet but I will think about offering one. Part of the app is available for free up to NL10. Including MDA Postflop.
Meanwhile I can manually give you access for you to try things out. Send me a dm with your email
You asked for it so we've released a "light" version of MDA Postflop that is much more familiar to what some of you know. It's more practical when you use always the same scope.
And a better way to show how sizings can impact your decision
Hi we have just added the player stats breakdown in MDA. I believe it's one of the most useful breakdowns.
More details : https://www.pokalab.com/community/g/pokalab-community?post=851bf448-717a-4249-89d0-519c9296a480
Thursday 12 Nov
enemi · 20:24
how do you get river bluff frequencies for WPN? I think the bluffs are mucked in river
Yes, you're right. Each room handles showdowns differently:
• GGPoker & Winamax: every hand is revealed
• PokerStars: the second player to show can muck a losing hand
• WPN: around 70% of losing hands are mucked
So we built a correction that rebuilds the mucked hands, based on the rooms where every hand is shown.
We know it works because we tested it:
• On PokerStars rivers, where the real hands are known, the corrected numbers come within ~1 point of the real ones.
• Each room hides a different share of hands, yet after the correction they all end up close to each other and close to GG. On every street
