So FineArt will be removed after next daily update right? I can still see it now

Update: yes, it finally gone for good

Sorry to bother you again but FineArt is not gone yet.

I'm sorry to bother this thread again but goratings just let DolBaram (which just had a big upgrade) in to the calculation which I'm afraid will affect human players, currently it has two wins against Xia Chenkun and Liao Xingwen
https://www.goratings.org/en/players/1928.html
at least its competition 3-2 with DeepZenGo are not counted
http://www.go4go.net/go/games/byplayer/1928

Thanks for your message. I removed it. It should disappear at the next update.

I should have let Rémi know when I include more AI games into the underlying database. I think these games are of really good quality now and provide very good studying materials. So it is right to have them included in a Go database.

Rémi, would it be helpful if I modify the data feed I supply and give AI players a new value in the 'SEX' field? Currently we have only 'M' for male and 'F' for female players. I can add an 'A' value for AI players. It should be really easy to filter them out from your side.

Yes, that would help. My code already uses 'C' internally for computers, if you don't mind using it instead of 'A'.

Done! Those AI players are now marked by 'C' in the data feed.

My data also contains games from Taiwanese AI CGI. And I will add more games by Chinese AI abacus. Both of these are playing at strong professional plus level now.

This new ranking method claimed to be better than yours. Do you have any comment on this?

https://www.sciencedirect.com/science/a ... 1717306574

It doesn't seem to me that they claim it is "better than WHR". They do note that WHR has high computational cost and only works for players with both wins and losses. In general one can handle undefeated players by incorporating a prior (otherwise there's no way to predict an undefeated player will ever lose); I haven't thought about whether this would be trivial or hard to fit into WHR if you wanted.

They do not seem to have any quantitative evaluation of their method other than computation time. In particular, I don't see any place where they attempt to characterize how accurate their system is at making predictions; as far as I can tell the paper is basically just "here's the motivation for our system; here's our system; here's how it rates some top players". Given that, I don't really see how you can compare its quality in any sense other than speed (and I find it hard to believe that Glicko, which they mention but don't compare their system to, would be any slower).

Thanks for letting me know about this paper. I was not aware of it.

The method looks to be of little practical value. The paper does not claim that the method is better than WHR. It does not measure the quality of predictions made by the method at all. And it seems it does not consider the variation of player strength in time. I don't have time to read the paper in details. It ranks Lee Sedol first, ahead of Gu Li (6th), and Ke Jie (9th). That's enough to make me not want to invest more time reading it.

I use a prior in WHR, and compute the ratings of players with 100% wins and 100% losses. I just don't display them on the web site, because they are likely to be very inaccurate.