Absolute Soulreading/Rigged thread #3

[Drag]

[ QUOTE ]
Drag, just so you know, Doubledrag's standard deviation was around 260 PTBB/100 in that 190-hand sample, and he won about 470 PTBB. Even if you think the loss rate should be -20 PTBB/100, the result is within 2 standard deviations.

[/ QUOTE ]

In the screenshot he posted, there is 400 hands, with winrate of 288PTBB/100. He won about 1200PTBB in these hands. Even if we take the SD you cited, i.e. 260PTBB/100, we'd get a result which is about 5 SD far from the mean ( -80 PTBB). I don't remember to what amounts 5SD, but I'd guess 10^-3 that I took as a VERY CONSERVATIVE estimate still holds.

As soon as you start to add more hands (and we know that he won about 0.5 million) we'd get even better purely statistical confirmation. Notice that we even don't need to discuss the specifics of individual hands to see that something strange is going on. Individual hands add even more evidence.

[krumeluren]

[ QUOTE ]
Results:
for 100 hand spans using 13400 hands the confidence interval (usually you use 90% or 95%, that means in 90% you will be below this win rate over 100 hands) is:
90%: 103.65bb/100
95%: 124.07bb/100
99%: 164.15bb/100
99.999%: 283.99bb/100
99.99999999999%: 480.77bb/100
or 1:10000000000000 or 1:10E14

[/ QUOTE ]


Interesting analysys. Good work. We need more of this.

If i interpret your results correctly, under the assumptions your model is correct, if you have 100.000 streaks of 100 hands each, one of those will show a win rate of 284bb/100 just by chance? How many hands are there in ppls PT-databases (which would be the population) and what is the probability that every once in a while we will find a maniac with those winnings without cheating?

And what if you increase the standard deviation a little? I mean, this kind of maniac could have a SD of 100 right? That would make it even more probable that such good streak could occur right?

your model also assumes that winnings are normally distributed. Do you think this is a good assumption and in what way do you think violations th normality would affect the estimated probabilities?


EDIT: I just saw that the had a SD of 270ptbb/100 and that it was 190 hands. That will give completely different estimates. It would be very interesting if you could present a new analysis with those paramters.

My guessing is that you will find a very high probabbility that such streak would occur in ppls PT databases, more than once.

[MLG]

oh boy

[ozyman]

Looks like the story ended up being picked up by Steven Levitt of Freakonomics fame on his blog.

How Not to Cheat

The link should be free.

He even has a hot link to 2+2

[BrianBigNFun]

watch the boulder start to roll downhill....

[teddyFBI]

Wow, things are about to get interesting.

Only thing that detracts from the integrity of his little article, though, is the self-serving blurb at the end:

"(Note that I say nearly undetectable, because while that poker site probably never would have detected them, I am working with a different online poker site to develop a set of tools for catching cheaters. Even if these guys were careful, we would catch them.)"

[BrianBigNFun]

adanthar should edit his 1st post of the hot linked thread to welcome everyone to 2+2

[ikestoys]

[ QUOTE ]
Long post, picking and choosing sections...
The hands won't go into PT, but I think it's about 80/60. Your assumption about 15/30NL players reacting to a player's looseness is reasonable, but that's not what happened at this table. His opponents were laying down to c-bets like clockwork.

[/ QUOTE ]

Son, this isn't true
[ QUOTE ]
You might find it suspicious that the c-bets are so effective, but it doesn't make sense that it's DRAG's doing, because he was c-betting like 90% or so when heads up on the flop. So obviously they know he has nothing usually, but still refused to play on.


[/ QUOTE ]
ooook
STAGE #753375692: HOLDEM NO LIMIT $30 - 2007-09-04 17:44:57 (ET)
Table: ALHAMBRA (Real Money) Seat #5 is the dealer
Seat 5 - DOUBLEDRAG ($22373 in chips)
Seat 6 - THE GRINDER ($11554 in chips)
Seat 2 - DTON21 ($1760 in chips)
Seat 3 - PTOWNER ($3055 in chips)
THE GRINDER - Posts small blind $15
DTON21 - Posts big blind $30
*** POCKET CARDS ***
Dealt to PTOWNER [5d 10d]
PTOWNER - Raises $90 to $90
DOUBLEDRAG - Raises $240 to $240
THE GRINDER - Folds
DTON21 - Folds
PTOWNER - Calls $150
*** FLOP *** [6d Qc 8d]
PTOWNER - Checks
DOUBLEDRAG - Bets $150
PTOWNER - All-In(Raise) $2815 to $2815
DOUBLEDRAG - Calls $2665
*** TURN *** [6d Qc 8d] [2s]
*** RIVER *** [6d Qc 8d 2s] [5c]
*** SHOW DOWN ***
PTOWNER - Shows [5d 10d] (One pair, fives)
DOUBLEDRAG - Shows [Kd 3c] (king high)
PTOWNER Collects $6153 from main pot
*** SUMMARY ***
Total Pot($6155) | Rake ($2)
Board [6d Qc 8d 2s 5c]
Seat 2: DTON21 (big blind) Folded on the POCKET CARDS
Seat 3: PTOWNER won Total ($6153) All-In HI$6153) with One pair, fives [5d 10d - P:5d,B:5c,B:Qc,P:10d,B:8d]
Seat 5: DOUBLEDRAG (dealer) HI:lost with king high [Kd 3c - P:Kd,B:Qc,B:8d,B:6d,B:5c]
Seat 6: THE GRINDER (small blind) Folded on the POCKET CARDS

[ QUOTE ]
I mention it to describe the nature of the table; they were letting him run over them. Obv I'm not going to do the calculations, but I wouldn't be surprised that he would have still come out a winner in the session if he open-folded any hand that got to the turn. They were folding to practically minbet c-bets over and over!

[/ QUOTE ]

How bout the hand I pushed with K10 and he called with A3?

[ QUOTE ]

He called with TT in the blind, then check-raised the flop when he hit the set. He was flat calling lots of raises. I saw him do it with JJ too against someone without a big hand.

[/ QUOTE ]
this could just as easily because he was trapping sir

[ QUOTE ]

No. The sample is way too small for that. There are only a few times where he calls the river, and he's not doing it with A high, picking off some bluff. For all we know, he folded to some river bluffs. He makes quite a few small river bets with losing hands. He also makes quite a few big river bets, and we see that a couple of those are with losing hands.


[/ QUOTE ]

Son, he was attempting bluffs with teh big river bets... and I'm willing to bet that you have no background in statistics and are a losing poker player, so forgive me if I don't take your assumption that is too small of a sample
[ QUOTE ]
That's not in this 190 hand session.

[/ QUOTE ] Then maybe before you talk about the sample too small you should collect all the data available

[darkcore]

i need more popcorn i guess...

[ QUOTE ]
[ QUOTE ]
Results:
for 100 hand spans using 13400 hands the confidence interval (usually you use 90% or 95%, that means in 90% you will be below this win rate over 100 hands) is:
90%: 103.65bb/100
95%: 124.07bb/100
99%: 164.15bb/100
99.999%: 283.99bb/100
99.99999999999%: 480.77bb/100
or 1:10000000000000 or 1:10E14

[/ QUOTE ]


Interesting analysys. Good work. We need more of this.


[/ QUOTE ]

[FellKnight]

[ QUOTE ]
Wow, things are about to get interesting.

Only thing that detracts from the integrity of his little article, though, is the self-serving blurb at the end:

"(Note that I say nearly undetectable, because while that poker site probably never would have detected them, I am working with a different online poker site to develop a set of tools for catching cheaters. Even if these guys were careful, we would catch them.)"

[/ QUOTE ]

Agreed. If a smart cheater did this (i.e. played normally, only increasing his or her edge by fractional values even by simply folding hands whenever an opponent had a monster), they could never be detected. Unless there is some PT stat I am unaware of "number of times lost to a monster" [img]/images/graemlins/smile.gif[/img]

Fell