Re: Luck and

Gary Carson wrote:

The statistical terms that have some relevance are fat-tails, skewed,
non-stationary, large deviations, and similar concepts.


I'm glad you mentioned this.

I keep rather crude records (balance at the end of every day). In a typical day I play 4 hours/2 tables/online micro-stakes NLHE.

I know what it means that I'm averaging 15 big blinds per day profit; I'm not at all sure what it means that my standard deviation is about 30 BB. I even think I know what it means that my profit has been rising slightly at the same time that my stddev is declining slowly.

But...I have no clue as to what skew and kurtosis have to say about my results. Right now, skew is nearly -0.0 and kurtosis is just barely (but solid and predictable) positive.

And...I have serious doubts that the distribution of day-to-day changes is a unimodal distribution. I haven't bothered to histogram it, yet - but I'm moderately sure that it's got at least three (3) peaks - one big one in the middle, one cluster of "big wins" and another cluster of "big losses". My efforts lately have been on eliminating the "big losses" - which shows up in a lower stddev and a higher mean. I *think* that this (reducing the big negatives) should generate positive skew (which is what I see right now in my short-term results - say, over the past month), but I have no intuition on how it "should" change the kurtosis.

All clues gratefully rented.

If you could pick (reasonable) values for stddev, skew, and kurtosis - values that would imply you are playing well - what values would you aim for?


--
Kenneth Sloan KennethRSloan@xxxxxxxxx
Computer and Information Sciences +1-205-932-2213
University of Alabama at Birmingham FAX +1-205-934-5473
Birmingham, AL 35294-1170 http://KennethRSloan.com/
.

Relevant Pages

  • Re: Adjust for Skewness and Kurtosis in Excel
    ... If I have a distribution which is approximately normal, with just a slight skew and kurtosis, can I somehow adjust for that? ... Lacking access to raw data, ...
    (sci.stat.math)
  • Re: OT: Even discussing IQs in the range of 200 is absurd
    ... skewed and/or flat/peaked distribution curve. ... You can throw around words like skew and kurtosis all you like. ... See if you think there is no face validity for extremely high scores. ...
    (rec.gambling.poker)
  • Re: Contradictory tests of normality
    ... A histogram shows virtually no left tail, ... positive skew and negative kurtosis. ... rejects normality, ...
    (sci.stat.consult)
  • Adjust for Skewness and Kurtosis in Excel
    ... If I have a distribution which is approximately normal, with just a slight skew and kurtosis, can I somehow adjust for that? ... Again, I'm a novice at this, so please correct me if I've done anything fundamentally wrong. ...
    (sci.stat.math)