For example: If i saw that the current hand was a bomb pot ($100) I would count that $100 as the big blind for that hand so now if mike were to win a 1k pot that pot would only be considered 10bbs in total. Also i msged you all the numbers privately.

You make a good point, but is that really how it works tho? If postle were to show up with a std d of 140 lifetime you are saying its incorrect to use that std when calculating a small hand sample?

LoL Dneg's vlog title today:

"Using CTO to defend against GTO in the mini main event"

I appreciate you showing your work but how did you account for the bomb pots and straddles?

Let's say Mike loses the 1/3 blind for -1. The next hand is a $30 dollar straddle and he wins $300.00, are you calling that +100bb or +10bb?

Bombpots can be different in some are run out without further betting or its just limiting the preflop action. But lets assume early in the game, $30 each, nine handed and they only limit the preflop betting. If he wins a $270 pot on the flop, are you saying he won 9bbs or 90bbs?

I don't think it really matters. He cheated more in "bigger games" (ie deeper and when straddled). He wasn't good enough at PLO to even be profitable when knowing other people's cards. The difference if he won at 200 or 900 or whatever is pretty negligible imo.

If the hand is a straddle or a bomb pot the big blind becomes whatever that number is. I don't get what you mean there is only one way to do it. Doing it this way can only benefit Mike anyway dude.

Good question tbh.. I think it matters that the variance calculator and its inputs not being meant as a tool for comparing one players former performance to his performance *now*, but as a way to figure out how likely a certain good/bad run is compared to the data used as input.

I can't formulate my exact thoughts here properly, but when the session(s) we're looking into has a wildly different std dev than both the players previous std dev (Postle not cheating Std dev) and what a reasonable actual normal play std dev (even when erring on the high side like you did) would be, the results aren't applicable/comparable I think.

Again, I can't fully iron out my thoughts here, nor am I even sure of if I'm right, I lack in intelligence/statistical knowledge

edit: Basically I think what I'm trying to say is "Ofc 150bb/100 std dev assumption means it's mathematically impossible to win 1600bb/170 - this doesn't prove he cheated, it can also be that the std deviation of his play wasn't 150bb/100"

Pretty sure you just proved my point. Obvoiusly his std deviation isnt 150bb/100 in the sample thats because hes cheating. His std is some absurd number that doesnt make any sense.

I'm no statistician; but can't this lead to some rather significant consistency/distortion issues? Theoretically; each hand could have a different BB, or at least BB is not a consistent "thing" if you will. To use an analogy, in one hand a BB is a "car tire" and in another hand the BB is "a set of car tires." If we simply add up 4 car tires plus 8 sets of car tires for a session win of 12 BB's; what is that 12 BB really: it's not 12 car tires and it's not 12 sets of car tires.

Simple example:
- Assume a 2 hand session.
- Hand 1, non-bomb pot, non-straddled hand, MP wins $90 (or 30BB)
- Hand 2, $30 straddle, MP wins $550 (or 18.333BB)

The session win is 48.333bb if we just add. But; is 550+90 actually 48.333bb for the session when bb means something different in each hand.

This is my last time explaining this.
Examples:
Hand 1: $100 bomb pot postle takes it down on the river for a total profit of $1500(15bb)
Hand 2: The exact same thing happens except its a $30 straddle
postle wins $1500(50bb)
Hand 3: same action as before except no stradde blinds are 5/5/10
Postle wins: $1500(150bb)
I also did the same thing for pots he lost
Example: postle loses $100 in bomb pot (-1bb)

Yeah but anyone can get a super high std deviation, especially in a small sample, all you have to do is play a couple of 1000bb pots and you're there

The first person to show me a hand sample of 174 hands un edited with a std dev similar to postles in the sample I showed gets to defend him in court.(Losing sessions doesnt count)

I just listened to the UTG Podcast with guest Andrew Milner, an aussie developer who made Poker GFX, who I must say is a top bloke who does a fair dinkum job of answering questions for the layman.

As others have mentioned before it is definitely worth a listen, especially if you are interested in the RFID tech, Poker GFX software, and any graphics changes that may have taken place. Here is a link to the podcast both the interviewer and Andrew do an amazing job: https://podcasts.apple.com/us/podcas...=1000453274898

1:17:00 In depth conversation about RFID, Poker GFX, its potential exploits and how to mitigate them
2:22:00 The 86o graphics change hand
Side note: Postle actually had 86o not 96 https://www.twitch.tv/videos/365129696?t=02h30m44s

So we know that:
1) Cards can be miss-registered but not misread
2) If a card is pitched to the wrong player, and the RFID reader picks it up, the player's who's card was pitched incorrectly will not display cards at all. The software prevents the same card displaying twice in the event of a duplicate RFID read within the same hand.
3) There is a persistence error where prev cards can be read as current cards.

The podcast has new information as to why graphics might be wrong. Andrew, the creator of Poker GFX, lists 3 scenarios that causes graphical errors. The first 2 are listed above as 1 and 2 respectively, and the 3rd is as follows. This explains why there are so many graphical errors relative to other streams. It explains hands like when the Kc was showing up on two different peoples hands preflop. He also goes on to say that when card bleeding occurs, a popup notification is sent to the operator to let them know. Which for me, explains the extremely weird hand where Postle folds TT to J4o which turned out to be JJ. https://www.twitch.tv/videos/306445878?t=5300s

Andrew goes on to say that all 3 of these possibilities are red herrings and that there would never be a reason that the graphics should ever change in real time unless the cards were tabled. I would like to add to this point in that there is no way for Taylor or anyone else but MP, to know the exact hand, down to the suit, what MP held in real time. Taylor then goes to the booth and confirms 30 minutes later that the graphics change was correct, meaning he magically guessed MP's cards and was correct after confirming with MP himself. Even if card bleeding occurred in this hand and a popup notified Taylor, he or someone else would have had to make the split second decision to change the cards, and they would have had to have been right, as they confirm the change 30 min later in the delayed stream. There is no reason to change it. MP is known for crazy plays. MP is known for soul reads. There are graphical errors all the time. Why change it? There must be a motive.

Anyway, it makes more sense now why Stones was experiencing so many graphical errors. They were just lazy with the RFID table. It's possible they didn't want to fix it because it made for a good excuse for Mike, but they made a huge mistake changing the graphics here. I emailed the creator of Poker GFX with more specific questions regarding what could affect graphics and will share it with you guys if he decides to respond.

I don't think you understand what mathematically impossible means. There is an infinite amount of mathematical impossibilities, this is not one of them. It should be impossible for the universe to even exist, as there should have been equal parts matter to anti-matter, yet here we all are. Great effort and something to submit for review from your peers, but its a very small sample size so it is somewhat statistically insignificant. I've seen someone sit down with $600 at 10/20 NLHE and run it up to $24k in 5 hrs. Crazy **** can happen in small sample sizes.

if you were trying to explain the situation to a layman this would be true and yet you'd be misleading them to emphasize the fact that it's not impossible. in some peoples minds any amount of doubt is tantamount to reasonable doubt because human beings are notoriously poor at weighing the proportionality of low probability events.

It's very difficult to assign a stdv over such a small number of hands in general because one of the core assumptions is that the sample stdv is equivalent to stdv of all possible hands that could have occurred.

Which is one of the reasons why we'd be using baseline estimates of what the stdv tends to be in those types of games.

his sample stdv being so astronomically high could be a consequent of a unique play style but is also something you'd expect to happen as a consequent of him cheating and so it's not really an effective way to estimate the probability of that kind of a run happening by chance.

the difficulty is establishing to non poker players why peoples stdv tend not to be that large absent of cheating (or playing in a manner that would show enormous losses). it's very easy for people to see poker as some kind of an art form where some just have super genius iqs that let them play in ways that would befuddle mere mortals and see the accusers as being bitter/jilted losers looking for someone to blame for their losses.

If $4500 session win is equivalent to winning 215BB, that means the average BB for the session was 4500/215=20.93. But you're saying that hand 1 had a $100BB, hand 2 had a $30BB and hand 3 had a $10BB: that's an average BB for the 3 hand session of (100+30+10)/3 = 46.67. There is a massive difference saying the session averaged a bb of 20.93 vs 46.67. In other words, the hypothetical 4500 profit can either be 215BB's or (4500/46.67)=96.42BB's: that's a rather large discrepancy.

know what you're saying, but seems the big problem here is PokerGFX software wasn't really conceived for commercial use. For that reason it was a cheap entry into low-budget streamed poker, unlike the bespoke versions used for other streams. In the particular case of Stones, their lack of protocols generally made an even softer target. It is possible of course, but cheating would be very unlikely for any stream absent of both of these factors.

Great effort Xenicide! didn't think you had it in you building up more of these would be great. Also think breaking down MP's river play, as discussed above (and possibly his pre-flop play, particularly in certain positions), will prove illuminating too.

It should most clearly illustrate....

But we can't just change the std dev to something much lower just because we're assuming it's possible only by cheating, or else we've made our desired conclusion one of the assumptions of our analysis. We have to show that even given his high-variance style of play, it is still unlikely that the mean (his win rate) is possible without cheating -- that so many positive results (winning sessions) could happen within that sample size by chance (i.e., without cheating).

And our best predictor of the std dev of the population (his long-term results) is the std dev of the sample. We have to assume that his long-term style of play would be the same or at least similar to the way he's playing in the sample, so the std dev would be similar. So I don't think it makes sense to apply a smaller std dev of a more conservative style of play to him for this analysis.

What? why are you grouping hands together and trying to make an average? thats not how you look at it each hand is different. I didnt add any numbers up in terms of $, every hand was added in terms of big blinds so total profit at the end is not measured in $ its measured in bbs

Facts.

Thank you for your effort and work.

are we really going to debate whether or not 150 is a high std dev for a winning 9 max player?
Is this high enough for you? still not convinced hes cheating?
It takes 250bb/100 std dev just to make him 1 in a million with a 60bb/100 w/r

Based on your post (#8895); you are representing the session's profit as a summation of BB's won/lost. But you also stated that BB is different depending on if the pot was normal pot, bomb pot or straddle pot. I'll go back to my tire analogy: if a big blind is one car tire in one instance and one set of car tires in another instance; adding up number of BB's when BB is something different each time presents issues in consistency. Saying someone is 14 BB's tall when BB can mean 4 inches, or 12 inches or 6 inches all at the same time (i.e. same session) doesn't really tell people anything other than someone is 14 BB's tall: you can't relate it/compare it to anything when BB is not static. If you are going to translate a session's profit into # of BB's won; you probably should average out what the session BB was as opposed having each hand (potentially) having a different BB and simply calculating session profit to be a summation of non-consistent BB's.

Here's a rather simply example:
- 2 hand session
- Hand 1, $100 straddle, MP wins 1BB
- Hand 2, no straddle/no bomb, MP loses 1BB

If session profit is the summation of BB's won/lost, MP made zero BB's, was he even for the session? Hand 1 he wins $100 and Hand 2 he loses $3 for session profit of $97. So now, because BB is different in each hand; zero BB profit is equivalent to $97 profit? A person need not be a statistician to realize something isn't right here.

I also heard Poker GFX was not intended for commercial use and believed it, until I listened to the podcast. It simply isn't true. It is intended for multiple use cases, like a swiss army knife, that doesn't mean it isn't intended for commercial use. Their are many measures and safety precautions built into the software, but when the person you have to protect against is the guy actually running the damn thing, that is where the problem occurs. What this really shows is that there needs to be stricter rules and regulations in place for all live streams. I recommend giving the podcast a listen since they go over all of this in depth.

I agree, it is a great effort by Xenicide, but I also think he shouldn't throw around phrases like "mathematically impossible", he shouldn't be so adverse to critique or suggestions, and he shouldn't imply his findings or stats alone will prove Postle guilty in a court of law. Other than that, there's no harm in crunching numbers to see what comes out.

You are telling me if I play 1000 hands with my friend dave at 1/2NL
and beat him for 10bb/100
Tomorrow we play again except this time we raise the stakes to 2/5
I beat him for 10bb/100 again
You are saying the two of these things are not the same thing?
sorry dude im not debating this anymore.

ah ok, fair enough, haven't listened to the pod. Think I heard that from Berkey? If i was looking for similar behaviour, i'd still start at PokerGFX powered streams though...


Agree 100%