KGSRating Discussion

Frustrated by my lack of progress in the wake of the ratings system changes in the fall of '04, I had a look around on KGS to see if the effect was general, or whether I just stink. The latter may, of course, be true, but a brief exam seems to indicate that few mid-kyu players are going up and many are going down.

I went through the list of folks online about 11 p.m. West Coast time picked out any who were from 4k to 8k, no ? and at least one year on KGS, and looked at the change in their rank on the rank graph from Jan 05 to late May 05, and also from Jun 04 to Jan 05.

Of the 44 people I found, here were the changes from Jan to June

change  num
+3       1 (drimgere)
+1       3
+0      14
-1      20
-2       6

For June 04 to Jan 05, the changes were

change  num
+2       1
+1       2
+0       7
-1       5
-2      22
-3       5
-6       1, poor chap

The effect of the two rating system changes seems nearly universally to have been a two-rank decline, so it may make sense to add two to each level in the second list.

Three things jump out at one. First, in the earlier period, the dispersion of results was greater, with more people changing ranks quickly. Second, the earlier list (adding two to compensate for the big shift) tipped toward people being promoted. Third, the current chart tips hugely toward people being slowly demoted. Almost no one is going up at all and those who are are mostly going up slowly. Almost everyone is declining gradually.

This seems odd. If, as I assume, the math is working the way it is supposed to, about the only explanations I can come up with are a net influx up higher grade players or a big improvement in the playing skills of lower grade players.

What does anyone think?

ilan: My observation from the ranks of people who started playing at the same time as I did, and who were in the range 5K to 1D before September 2004 is that ranking immediately went down about 2 or 3 ranks, and then were frozen. Since people were still improving, but their ranks weren't, this meant that people who weren't improving were losing ranks. This caused another rank or two of inflation (ranks are stronger). I think this has changed recently and ranks are going up again.

RayTomes: New players generally seem to have to play at below their correct handicap to start with to get a rating and then work their way up. The result of this is to suck points off other players. This will be one factor in causing the shift that you notice.

AndyPierce: Everyone who's rating graph for KGS that I've looked at has had a huge jump upwards at about what looks like the third week in September '06. Was there a KGS ratings adjustment?

MrTenuki: The KGS rating system was tweaked when CGoban3 came out. Players weaker than 4d were shifted upward, while high dans were shifted downwards.

Lynx: My question is, what are the values of the new ranks? A KGS 1 dan =? And so forth. I will say that I noticed a HUGE play difference from 3k-2k. Some of those 2ks are now 2ds, along with some players who were 1ds from before. Is this because the 1ds from before weren't that much better? How does this whole thing work?

zinger: The table at RankWorldwideComparison has been updated to reflect the new KGS ranks. The purpose was to get the handicaps more accurate - previously the ranks were too "narrow". For example a 4k could not really give six stones to a 10k and have an equal chance to win. Supposedly the new system works better at this.

ww: Few questions: 1. Assume that I play a) 10 games today and get 8 wins 2 loss b) 5 games today 5 wins 0 loss c) 9 games 7 wins 2 losses d) 15 games 10 wins 5 losses. Which one of them shows best development, 2nd best,3rd, 4th on my rank graph on the next day when it is updated ? Let's also assume, that all games were played against the same player(s). If losses has same weight as wins, then a) is superior right? But when I have asked around, some have preferred 5-0, as you have suffered no losses. 2. Can the boost be different if my rank is x.1 or if it is x.9 ? Point is that is the system expecting better ratio from a player with x.9 rank for it to advance to a new rank ?