[Computer-go] scores histogram

[Vlad Dumitrescu]

Hi,

I'm sure that such a simple histogram won't add much information, but
it might hint at interesting things that can be investigated further.
I'm still surprised that the peak of the curve seems to be at the komi
value - my intuition regarding statistical results is usually wrong...

The peak for 19x19 is at around 11 for this experiment and this might
mean that correct komi is depending on the level of the players. For
the dumb playouts I use, it might be 11 and for top-level play it
could go down to around 7.

Or maybe komi is the same at all levels and then the professional
players aren't as close to perfect play as they think they are :-)

Or maybe many more samples are needed to get a significant result
(currently running 100 million playouts).

regards,
Vlad

On Tue, Jan 10, 2012 at 02:12, Hideki Kato <hideki_katoh at ybb.ne.jp> wrote:
> Hi,
>
> I've observed some and found there are several peaks when the positions
> are complicated, i.e., there are multiple semeais simultaneously, for
> examples, on 19x19.  Simple statistics may not work on such positions,
> though I've not tried any.
>
> Hideki
>
> Vlad Dumitrescu: <CAA-EFXtnEUUZf63K7Nx=4ThZnr71iiE5zt00Steyx__aEga_KA at mail.gmail.com>:
>>Hi,
>>
>>This is probably very close to useless, but it might be fun in a geeky
>>kind of way :-) It might not be novel, either.
>>
>>I ran some playout games without komi (with simple policies: capture
>>atari, escape atari, simple 3x3 patterns) and plotted the distribution
>>of scores. I thought that the top of the bell curve would be placed at
>>the ideal komi - well, close to it, as the playouts aren't very smart.
>>The engine I used is Orego 7.10 and its scoring functionality.
>>
>>For 9x9 and ten million games, the top seems to be around 7.
>>
>>For 19x19 I only ran one million games and the top is difficult to
>>locate precisely, it's quite flat between 3 and 21. The middle of this
>>interval is around 11, which feels a bit high. I plan to let it run a
>>while longer.
>>
>>regards,
>>Vlad
>>_______________________________________________
>>Computer-go mailing list
>>Computer-go at dvandva.org
>>http://dvandva.org/cgi-bin/mailman/listinfo/computer-go
> --
> Hideki Kato <mailto:hideki_katoh at ybb.ne.jp>
> _______________________________________________
> Computer-go mailing list
> Computer-go at dvandva.org
> http://dvandva.org/cgi-bin/mailman/listinfo/computer-go