Brian,

Good to hear you chime in on this.  I don't think Quackle is seeing 
anything; when you sim the position individually, the plays are 
within 1 point of each other.  The plays are just DIChOTI(C) G1 72 
vs. DICrOTI(C) G1 72.  So it's basically the same play, just making 
the blank an H instead of R, and it's fairly early in the game.  

I think the championship player doesn't do very many iterations 
because it sims the whole game in about 10 minutes, so that's why the 
discrepancy is there. But I'm hoping JKB or John O can confirm that 
that's what's happening. 

Mike



--- In [email protected], "sapphirebrand2000" <[EMAIL PROTECTED]> 
wrote:
>
> --- In [email protected], Russell Honeybun <tcheq@> wrote:
> >
> > 
> > The 6 point difference isn't due to triple-triples,
> > it's just due to random bonuses played.  At most 1000
> > sims would have been processed, which is <1% of the
> > required sims to produce results with any certainty.
> 
> This doesn't square with my general expectations about simulation 
> accuracy. A 1000-trial simulation that runs from the start of a 
game to 
> the finish should have a standard error of maybe 3 points. This is 
a 
> much shorter simulation, so I am thinking the standard error is 
maybe 
> 1.5 points. A difference of 6 points is certainly significant.
> 
> It is possible that the simulation is using many fewer trials, or 
the 
> variances are much higher than typical. But I would bet that 
Quackle 
> has seen something.
> 
> Brian
>


Reply via email to