Question about variance.

    • 8979687
      8979687
      Bronze
      Joined: 11.11.2008 Posts: 2,225
      My finish distributions are always way in favor of first place and I am running miles above EV.

      I have a sample of 2.4k games of $6.50s and $12s mixed.

      The thing that gets me is being so far above EV. Like 100+ BIs and I have been running over EV for 1K games.
      with one drop back to the red when I started the $12s and then I fired right back off again to over EV.

      Last year at the end of my $6.50s the red and green line just dropped like an anvil and all profits were erased.
      I was never higher than 20-30 BIs above EV and not frequently either. I was also a much worse player and my
      finish distributions actually have 3rd as the highest. It is only a 796 game sample but I included it for a larger
      sample.

      So am I just extremely lucky and can I expect a crashing return to reality or is it safe to think I am doing
      something right in looking at the ITM and finish distributions over these 1K+ samples? Or do I need way
      more games to have any idea yet?

      I do review my games and I do study and I am constantly finding leaks to plug as I am sure I always will
      no matter how long I play for or how many games. It makes me worry being so high above EV that I am
      just gliding along on the wings of good luck and its going to end with a slap in the face sooner or later.

      Here are this years finish distributions...



      Uploaded with ImageShack.us

      And the whole thing over 2K sample including my heavier learning stages.




      Uploaded with ImageShack.us
      http://imageshack.us
  • 5 replies
    • lessthanthreee
      lessthanthreee
      Bronze
      Joined: 30.06.2009 Posts: 16,300
      come back with a 10k sample and finish distributions might have some relevance.

      sounds like a lot of variance, not saying ur playing bad or anything.. just enjoy it while it lasts. as a 'believer' in the red-line . all roads lead to red. so you must prepare for the bad times by not getting overly excited when u do run hot, and not getting too depressed when you do run bad. the long run is a very very long time and 99.99% of STT players will never get a statistically significant sample size in their lifetimes.

      even if you do manage to get a large sample (10k+) on the same limit, your game would have evolved so much overtime and the player pool might have gotten harder/easier so your underlying real roi during the first 2k games would be completely different to the underlying real roi over the last 2k games... which would skew the over-all result....

      if only we could create a robot free of tilt to mindlessly play 50k games exactly the way we would play. then we might actually get some reliable data on attainable/sustainable rois for any giving limit.
    • pzhon
      pzhon
      Bronze
      Joined: 17.06.2010 Posts: 1,151
      As you play more, your estimates for your ROI and finishing distribution (under past game conditions) gradually get more accurate. A rough 95% confidence interval for your ROI might be your observed ROI +- 310%/Sqrt(# tournaments). A rough 95% confidence interval for your first place finishes might be your observed 1sts +- 70%/Sqrt(# tournaments).

      So, after 1600 games, that would be ROI +- 310%/40, or ROI +- 8%, and 1sts+- 70%/40, or 1sts +- 1.75%. You can say that you have strong statistical evidence that you are winning and that you are placing first more than 13% of the time, but you don't yet have tight estimates on either your ROI or your first place finishes. The red line is a better indicator of your ROI than your results alone, and it might let you multiply your sample size by 2.2 or more.

      You bust out in 8th and 9th more than many serious players. This may be because you are playing looser in early levels, which is probably a good thing. It is usually better to bust out early than to bubble out, since you save time. Particularly if you have a cash game background, you should have a high win rate in early levels, both in chips and equity, and one of the results is a high number of first place finishes since you will more often have the chip lead on the bubble. The more common nitty style leaves chips on the table and may result in more finishes in 6th-4th place, with only 11-13% firsts.
    • 8979687
      8979687
      Bronze
      Joined: 11.11.2008 Posts: 2,225
      Thank you both very much.

      I also just saw the variance video and I am sold on the red line for sure now.

      That was a massive sample and the two lines are married over and over again.

      The good news is that the green line does not have to drop for them to meet.

      You can have a break even stretch while the red line rises for example.

      I was just so worried about a dip in the red bringing death to the green.

      Results orientated FTL.

      I will look into my 9 handed and 8 handed play as well.
    • lessthanthreee
      lessthanthreee
      Bronze
      Joined: 30.06.2009 Posts: 16,300
      Originally posted by 8979687
      I also just saw the variance video and I am sold on the red line for sure now.
      donde estas este?

      yeh... i cant wait for pzhons new vids. topics sound really good.. pzhon is like the bomb
    • santostr
      santostr
      Bronze
      Joined: 11.08.2009 Posts: 663
      Looks like its pretty standart... Both ways

      I have 3% ROI and 5.9% evROI over 1377 games....
      And I just had a 10 BI > ev session..

      A 3rd 50BI downswing before the 2k games would be very sick...