[computer-go] Re: ZG1-trunMCx-50k

[Christoph Birk]

> ÿÿukasz did the experiment with 5 players, which is misleading.  If you
> put several identical players on CGOS like he did, you can pick and
> choose the very worst case and present it as a horrible anomaly.   Even
> so, he didn't find much difference,  his spread is about 54 rating
> points which indicates that we may be off about 27 rating points or
> so.

That's what I did. I used 5 identical players.
Maybe you miss-understood my result. I think a '12%-event' is
perfectly reasonable. I do not think the rating algorithm has
a problem. If at all this would point to a difference in the programs.

> By the way, I don't understand your data because I don't know how you
> calculated the ratings.   Did you do performance ratings as a whole, or
> did you run the simulation incrementally, rating them 1 game at a time?
> And what K-factor did you use if you did this incrementally?

I used CGOS settings (K=3) and run the simulations incrementally, starting
with 5 identical players and look at the max. rating difference
after 600 games.

This is the results histogram:
Col1: Max. rating difference after 600 games
Col2: fraction of events (10000 simulations) with 2 players
Col3: frations of events (10000 simulations) with 5 players

diff(rating)    2-players    5-players
10              0.67         0.90
20              0.38         0.72
30              0.19         0.48
40              0.08         0.26
50              0.03         0.12
60              0.01         0.04

Christoph