rating

User avatar
adum
Site Admin
Posts: 422
Joined: Tue Sep 30, 2008 5:09 pm

Re: rating

Post by adum »

:) okay. hard to tune these exactly right for me...
User avatar
BartTM
Posts: 119
Joined: Mon Sep 21, 2009 1:46 pm

Re: rating

Post by BartTM »

I find that new problems with little tries have a huge impact on my rating while older problems have almost no effect. I feel this should be the other way round.
User avatar
adum
Site Admin
Posts: 422
Joined: Tue Sep 30, 2008 5:09 pm

Re: rating

Post by adum »

the only thing that impacts your rating is how recently you tried the problem previously (max is if you never tried) and the problems' rating relative to your own.
wurfmaul
Posts: 6
Joined: Sat Sep 01, 2012 5:37 pm

Re: rating

Post by wurfmaul »

The rating system is really too generous. I'm 5k on KGS (and I'm probably lower according to other rating systems), but i can reach 1d here. Also I've submitted 2 problems, and they are rated as 1 dan and 3 kyu at the moment, though i think 10 kyu would be more appropriate. I've tracked the ratings of my problems a bit, and it seems like they are not converging very good.

When I am rated as 1d, and I fail a "1d" problem, it's rating is immediatly adjusted to 2d, so i feel my impact on the problems rating is too large.
I don't know how the current system works, but my suggestion would be to lower the current people and problem ratings (but i don't know by how much...), and to lower the impact of failing/solving a problem on the problems ratings, for example, the effect of a 1d failing a problem that is rated as 1d and has been tried 100 times should be lower than the effect of a 1d failing a problem that is rated as 1d and has been tried 10 times.
I like the rating system of KGS, but that's hard to implement for sure.

...Anyway, thanks for this site :)
User avatar
adum
Site Admin
Posts: 422
Joined: Tue Sep 30, 2008 5:09 pm

Re: rating

Post by adum »

the current rating system does in fact reduce the effect on ratings according to the number of previous attempts.

getting the levels right is hard...
wurfmaul
Posts: 6
Joined: Sat Sep 01, 2012 5:37 pm

Re: rating

Post by wurfmaul »

Ok, I didn't know that. Still I think that single attempts effect the problems ratings too much. For example, my two problems changed now from 3 kyu to 1 kyu and from 1 dan to 6 dan in only three days. It is true they are fairly young, 3 months (but without time trial) and 1 month respectively, with 300 and 200 tries, but I think the rating should be steadier already.
hippophile
Posts: 4
Joined: Thu May 10, 2012 9:28 am

Re: rating

Post by hippophile »

Me too. It seems some problems normally around the 1 dan level are frequently marked as 7-8-9 dan. Also some problems by 10 kyus or weaker are marked hard, perhaps because of inaccurate solutions...

I guess in the second case one just lives with (or avoids "hard" problems posted by weak kyu players). But I am a tad frustrated that I get to 9 dan so easily, and still get fairly easy problems at that level.
hippophile
Posts: 4
Joined: Thu May 10, 2012 9:28 am

Re: rating

Post by hippophile »

Of course, the other possibility is that doing the problems on this site has just made me incredibly strong! :-)

I am not really complaining, this site is a great repository for sometimes quite different problems. And I guess the rating system is hard to implement. Just expressing a point of view...
Jhyn
Posts: 1
Joined: Wed Dec 22, 2010 2:49 pm

Re: rating

Post by Jhyn »

If I may add my voice to the chorus.

I am 3k EGF, 3k KGS, and I am rated 6d after a few days. Maybe this is due to the fact that I like tsumego and I usually won't play before I read it carefully, but still the gap seems very wide. Also, I met some problems that are rated 6d here in a korean book which is supposedly aimed at 3k - 1d.
Aristotle
Posts: 5
Joined: Thu Nov 29, 2012 1:22 am

Re: rating

Post by Aristotle »

One of the ways to improve the quality of the solver rating, is to penalize "skipping" problems when solving. Since the quality of problems varies, some allowance must be made for the skipping of low quality problems. But when a solver skips more than 1 in 3 on average, a penalty is justified.

Another point is "solving in categorries" like Joseki, Fusaki, Life and death, Endgame and "Unusual". Someone may be great in solving Life and death and awful in Fuseki and skip Fuseki problems for that reason. By selecting a category at the start of a solving session, he need not skip too many offerings and be penalized for it.

Just a few ideas. I'm probably not the first one to bring them up.

Aristotle.
Post Reply