Twisted variance on Full Tilt Poker - set mining

    • QubaBuba
      QubaBuba
      Bronze
      Joined: 04.10.2010 Posts: 152
      After playing 50.000 hands on Full Tilt (30k on NL2 and 20k on NL5 rush) I started to recognize few concrete situations which are happening more (or less) often than they should.

      First interesting thing I focused on was set mining. There is theoretical probability of 11.76% that you should hit the flop. That should make one set per 7.5 misses in the longterm. But not on FTP.



      According to this table I did see the flop with 1964 pocket pairs and only in 163 cases the third card hit the flop.

      That makes probability of 8.3% after almost 2000 attempts. That is odd. You might say it's variance / bad luck / etc.., but it's almost 2000 sample, the deviation should not be that high:

      There's thing called Binominal distribution (http://en.wikipedia.org/wiki/Binomial_distribution) which allowed me to count probability of the low percentage occurence. With 1964 tries the equation would be really really large so I used Java's Big numbers math library (which is btw really big helper). In short: the chances of 8.3% or less after 1964 tries are: 0.000037%.

      Which is something like flipping a coin 25 times and hitting only 1 head and 24 tails.

      Few more possible FTP abnormalities are in my mind right now and I'll share them when I have some data to build on.

      My opinion:
      I don't know whether the variance is bended only here or also on the higher limits, I don't even know whether it's bended to all the players on these limits. But it's obvious that the lower limits (or maybe only some people there) are the biggest cash cow (most depositing candidates) for the site. It's in their interest to invoke the biggest gambling emotions they could. It works just like the gambling gaming machines in bars. And finally, let's be honest - if you were the site owner, wouldn't you want to maximize your profit, by not really obvious way, too?

      To paraphrase Shakespeare, something is rotten on the site of Full Tilt. :)
  • 51 replies
    • antstruk
      antstruk
      Bronze
      Joined: 30.08.2010 Posts: 494
      It's because you didnt play your pocket pair to the river 100% of the time. If you did you would have had a lager percent of hits.
    • QubaBuba
      QubaBuba
      Bronze
      Joined: 04.10.2010 Posts: 152
      It's not. The percentage is for hitting the flop. The river percentage would be higher.

      EDIT: Theoretical hitting by the river would be 19% or so. But it doesn't really matter in this case.
    • OZSA
      OZSA
      Bronze
      Joined: 18.05.2009 Posts: 804
      I played 1.5year on FT, now I play for a month at stars, same thing happends... days with 100 pocker pair and only 2-3 set (which you all even loose)... I just accepted the fact its rigged, its easier to deal with variance/cold deck..:)
    • tokyoaces
      tokyoaces
      Bronze
      Joined: 01.04.2009 Posts: 1,883
      My Results:

      Ring: 144 sets out of 1195 hands = 12.0%
      Rush: 182 sets out of 1564 hands = 11.6%

      I'm guessing there is an error in your report.

      .
    • jonnyquest
      jonnyquest
      Bronze
      Joined: 21.09.2009 Posts: 285
      If you played 50k hands you should have been dealt approximately 2950 pocket pairs and you saw the flop 1964 times or 67% of the time. Is that high or am I folding too many pocket pairs?
    • EmanuelC16
      EmanuelC16
      Bronze
      Joined: 02.01.2010 Posts: 13,897
      LOL@2k = sample! :f_biggrin:
    • fembot26
      fembot26
      Bronze
      Joined: 25.08.2010 Posts: 72
      QubaBuba, I've just carried out the same calculation with the binomial distribution (using Octave rather than Java); for some reason I obtain .000039093%, but this is very close to your own result.

      For those concerned about the supposedly small sample-size, the number of samples is actually a part of the binomial formula - the probability of obtaining these results on 1964 hands is about .00004%. In other words, if you considered all randomly-selected sets of 1964 hands with seen flops, only about 4 in 10 million of them would give 163 or fewer flopped sets.

      So if PokerStrategy had ten million members (has it?) and every one of them carried out this investigation, then four members would report such a low number of flopped sets. I'd say it's a statistically significant result.
    • EmanuelC16
      EmanuelC16
      Bronze
      Joined: 02.01.2010 Posts: 13,897
      Originally posted by fembot26
      QubaBuba, I've just carried out the same calculation with the binomial distribution (using Octave rather than Java); for some reason I obtain .000039093%, but this is very close to your own result.

      For those concerned about the supposedly small sample-size, the number of samples is actually a part of the binomial formula - the probability of obtaining these results on 1964 hands is about .00004%. In other words, if you considered all randomly-selected sets of 1964 hands with seen flops, only about 4 in 10 million of them would give 163 or fewer flopped sets.

      So if PokerStrategy had ten million members (has it?) and every one of them carried out this investigation, then four members would report such a low number of flopped sets. I'd say it's a statistically significant result.
      GOGO proof-online-is-rigged gadget!
    • tokyoaces
      tokyoaces
      Bronze
      Joined: 01.04.2009 Posts: 1,883
      Originally posted by EmanuelC16
      LOL@2k = sample! :f_biggrin:
      A 2k sample is plenty for a standard distribution analysis. Did your statistics class teach you otherwise?

      Anyway, the OP needs to post his methodology and cross-check his data because it seems very flawed.
    • EmanuelC16
      EmanuelC16
      Bronze
      Joined: 02.01.2010 Posts: 13,897
      Originally posted by tokyoaces
      Originally posted by EmanuelC16
      LOL@2k = sample! :f_biggrin:
      A 2k sample is plenty for a standard distribution analysis. Did your statistics class teach you otherwise?

      Anyway, the OP needs to post his methodology and cross-check his data because it seems very flawed.
      I like it how there's always one smart ass who thinks he can prove online poker is rigged and that he knows math better than Cigital and TST people... Rather than doing this analisys he could be improving his game...
    • Hackett77
      Hackett77
      Bronze
      Joined: 02.02.2009 Posts: 372
      Over 900,000 hands - flop set = 11ish %

      over that same sample I can have a flop set % of between about 5% & 25% over 2k hands
    • tokyoaces
      tokyoaces
      Bronze
      Joined: 01.04.2009 Posts: 1,883
      Originally posted by EmanuelC16
      I like it how there's always one smart ass who thinks he can prove online poker is rigged and that he knows math better than Cigital and TST people... Rather than doing this analisys he could be improving his game...
      Good point. :D

      However, you're dispute is still statistically invalid. :f_confused:
    • QubaBuba
      QubaBuba
      Bronze
      Joined: 04.10.2010 Posts: 152
      If you played 50k hands you should have been dealt approximately 2950 pocket pairs and you saw the flop 1964 times or 67% of the time. Is that high or am I folding too many pocket pairs?

      It's exactly 3190 but it's just like you say. Call20 rule. :)

      So if PokerStrategy had ten million members (has it?) and every one of them carried out this investigation, then four members would report such a low number of flopped sets. I'd say it's a statistically significant result.


      I guess very little percentage of all people playing online poker thinks about these statistics. And I refuse to believe that if it were really variance then I would be the one from 2,5m to experience it. :)

      Anyway, the OP needs to post his methodology and cross-check his data because it seems very flawed.

      I did cross check them several times. The numbers are really simple and clear from PokerTracker: number of pocket pairs with "saw flop" filter and number of hit sets on the flop. The only error I could imagine is some kind of a PokerTracker's data mining failure, but that's at least doubtful.

      I like it how there's always one smart ass who thinks he can prove online poker is rigged and that he knows math better than Cigital and TST people... Rather than doing this analisys he could be improving his game...


      So if I am the smart ass you talk about, then what are you? The dumb ass? :) You clearly don't have anything constructive to say but you are so eager to state your opinion. What are you? :)

      Feel free to do the math too, it's simple high school math, really. The results are clear.

      Over 900,000 hands - flop set = 11ish %

      over that same sample I can have a flop set % of between about 5% & 25% over 2k hands


      I said it was 50,000 hands sample and ~2000 pocket pairs I've seen flop with. I am curious whether you could still get these percentages from any 50,000 part of your 900,000 hands.
    • tokyoaces
      tokyoaces
      Bronze
      Joined: 01.04.2009 Posts: 1,883
      Originally posted by QubaBuba
      I did cross check them several times. The numbers are really simple and clear from PokerTracker: number of pocket pairs with "saw flop" filter and number of hit sets on the flop. The only error I could imagine is some kind of a PokerTracker's data mining failure, but that's at least doubtful.
      I'm using PT3 as well. I can check a couple of million hands but I really don't want to since my personal results show FTP is 100% fine.

      How are you setting your "hit sets on the flop" filter?
    • shehanshah
      shehanshah
      Bronze
      Joined: 02.12.2010 Posts: 385
      2000 - 15/04/2011- Me thought online poker is not rigged...

      15/04/2011- Anything is possible X(
    • EmanuelC16
      EmanuelC16
      Bronze
      Joined: 02.01.2010 Posts: 13,897
      Originally posted by QubaBuba
      I like it how there's always one smart ass who thinks he can prove online poker is rigged and that he knows math better than Cigital and TST people... Rather than doing this analisys he could be improving his game...


      So if I am the smart ass you talk about, then what are you? The dumb ass? :) You clearly don't have anything constructive to say but you are so eager to state your opinion. What are you? :)

      Feel free to do the math too, it's simple high school math, really. The results are clear.
      [img]http://t1.gstatic.com/images?q=tbn:ANd9GcQ-Kyspb5cwFk4Dc7XySbiEHR84GhRc6JdLgihaoAxBdu3vdOAA&t=1[/img]

      Check your stats after 1 million hands ;)
    • kiromanAAKK
      kiromanAAKK
      Bronze
      Joined: 08.10.2009 Posts: 4,022
      Do we have a standard value for the set at a certain data (whitout count the variable)?
    • Alficor1
      Alficor1
      Bronze
      Joined: 16.06.2010 Posts: 7,291
      I flop a set 16% of the time (bout tree fiddy hands sample), FT gives your sets to me. There's noting you can do about it brohem, i have a special contract with ftp.
    • evertonroar
      evertonroar
      Bronze
      Joined: 26.06.2009 Posts: 737
      Originally posted by Hackett77
      Over 900,000 hands - flop set = 11ish %

      over that same sample I can have a flop set % of between about 5% & 25% over 2k hands
      well thats two. wonder where the other two of the OPs only four in 10 million are....