[Computer-go] cgos bayeselo

Robert Finking robertfinkng555 at o2.co.uk
Sat Nov 13 03:33:42 PST 2010

Thanks Don, makes sense. Very impressive that you recalc based on the 
results of ALL games! I think it is good to have two separate rating 
systems for another reason - some engines learn. Bayeselo will give the 
overall score across all games, but the rating system will potentially 
give a better indication of the current strength.

Thanks again.


On 13/11/2010 02:38, Don Dailey wrote:
> On Fri, Nov 12, 2010 at 8:54 PM, Jason House 
> <jason.james.house at gmail.com <mailto:jason.james.house at gmail.com>> wrote:
>     On Nov 12, 2010, at 7:40 PM, Don Dailey <dailey.don at gmail.com
>     <mailto:dailey.don at gmail.com>> wrote:
>>         I have a question here. What is the rating on the main list,
>>         and how does it relate to Bayeselo? E.g. if I look on
>>         http://cgos.boardspace.net/9x9/bayes.html, I see that
>>         _Gnu3.8L1_2 has an Elo of 1857 after playing 258 games. On
>>         http://cgos.boardspace.net/9x9/standings.html (the only page
>>         I normally look at for ratings) it has a rating of 1760
>>         (after playing 258 games). How do these two numbers relate,
>>         and also how will it work, copying the Elo across to the main
>>         table, which at the moment doesn't look like it contains Elo?
>>         Sorry to ask basic questions - I'm still new =)I see on the
>>         Elo page that the Lazy,* bots have between them accrued 144
>>         games. and have a combined Elo of 2111. But on the main
>>         ratings page I see two separate bots, Lazy,0 and Lazy,1.  Is
>>         the idea that when you do the copy-over, the main ratings
>>         page will also only show Lazy,* ?
>>     No, the main page will always track the players as separate
>>     individuals.   So you should see Lazy,0 and Lazy,1 on the main page.
>>     On the Bayeselo page, which has the most accurate ratings,  you
>>     will only see Lazy,*
>>     When bayeselo runs it produces as output a set of sql commands
>>     that (for now) I will manually run from time to time.    They
>>     just update the ratings field for the comma bots.
>>     So as soon as I update the ratings of the comma bots you will see
>>     them start drifting away until I do the update again.
>     I think any sync to baeselo should update all bots, not just the
>     comma bots. Without that, it may take a while for other stable
>     bots to resettle. It'll also make the comma bots drift faster if
>     you don't.
> It's pointless to do that,  we might as well just use bayeselo 
> exclusively.     I actually considered making bayeselo the main page 
> and removing the other stuff.
> But I think it would be annoying to get a certain rating and then 
> without playing any games come back a week later and the rating has 
> changed.
> I think incremental ratings are best and what we are doing here is 
> making it a lot more stable.
> It will take some time for the bots to re-sync,   but that's ok.   As 
> it turns out Bayeselo rating seem to be higher so no bot will be 
> rewarded for not playing.    I think the rating pool in general has 
> been deflating so this will keep it stable.
> Don
>     _______________________________________________
>     Computer-go mailing list
>     Computer-go at dvandva.org <mailto:Computer-go at dvandva.org>
>     http://dvandva.org/cgi-bin/mailman/listinfo/computer-go
> _______________________________________________
> Computer-go mailing list
> Computer-go at dvandva.org
> http://dvandva.org/cgi-bin/mailman/listinfo/computer-go

More information about the Computer-go mailing list