BGonline.org Forums

Rollouts - Long

Posted By: Matt Cohn-Geier
Date: Friday, 2 January 2009, at 4:50 a.m.

In Response To: Rollouts - Long (Rich Munitz)

GNU's "standard error" is standard deviation. The JSD (joint standard deviation) between 2 plays is determined by adding the individual variances: sqrt ((SD of play 1)^2 + (SD of play 2)^2)).

So Stick's first rollout is TCTC (0.5 JSDs separation), but his second rollout has 2.43 JSDs separation. I don't remember how to calculate the standard deviation <-> confidence interval stuff anymore but keep in mind that 1 JSD corresponds to about 84.1% confidence interval, 2 JSDs are about 97.7%, and 3 JSDs are about 99.9% confidence, so it is somewhere around 99% that 64R is better, according to the second rollout. This is different from how Snowie reports its confidence intervals as 95% around one play; in order to get the JSD there, you have to first determine the SD of each play and then compare them.

So, for the JSDs between the two ROLLOUTS:

There's about 10.9 JSDs separating 64S in the two rollouts, and about 8.5 JSDs separating 64R in the two rollouts. That should never happen and clearly isn't a matter of chance.

I am somewhat less optimistic than Stick; I am not sure the seed # is a factor here.

Messages In This Thread

• 64 @ GS
  Stick -- Wednesday, 31 December 2008, at 8:19 a.m.
  • 64 @ GS
    John O'Hagan -- Wednesday, 31 December 2008, at 11:54 a.m.
    • 64 @ GS
      Stick -- Wednesday, 31 December 2008, at 2:56 p.m.
  • 64 @ GS
    pontiac -- Wednesday, 31 December 2008, at 2:57 p.m.
  • 64 @ GS
    Matt Cohn-Geier -- Wednesday, 31 December 2008, at 4:27 p.m.
  • 64 @ GS
    Caryl Stephanson -- Wednesday, 31 December 2008, at 9:48 p.m.
  • Rollouts - Long
    Stick -- Thursday, 1 January 2009, at 9:01 p.m.
    • Rollouts - Long
      David Rockwell -- Thursday, 1 January 2009, at 9:11 p.m.
      • Rollouts - Long
        Rich Munitz -- Thursday, 1 January 2009, at 10:49 p.m.
        • Rollouts - Long
          Stick -- Thursday, 1 January 2009, at 11:36 p.m.
          • Rollouts - Long
            Rich Munitz -- Friday, 2 January 2009, at 12:03 a.m.
            • Rollouts - Long
              Matt Cohn-Geier -- Friday, 2 January 2009, at 4:50 a.m.
              • Rollouts - Long
                Rich Munitz -- Friday, 2 January 2009, at 6:07 a.m.
              • Rollouts - Long
                misja -- Friday, 2 January 2009, at 8:48 a.m.
                • Rollouts - Long
                  Matt Cohn-Geier -- Friday, 2 January 2009, at 9:11 p.m.
                  • Rollouts - Long
                    Matt Cohn-Geier -- Friday, 2 January 2009, at 9:17 p.m.
                    • Rollouts - Long
                      David Rockwell -- Saturday, 3 January 2009, at 12:46 a.m.
                      • Rollouts - Long
                        Stick -- Saturday, 3 January 2009, at 1:25 a.m.
          • Rollouts - Long
            Matt Cohn-Geier -- Friday, 2 January 2009, at 1:08 a.m.