Life In 19x19

What's up with this anchor business? Before Christmas, I was quite happy starting the holidays having managed to get all of my playing accounts up to 3k and then *poof* Christmas is over and they are all at 4k. I know that the purpose of ratings is to get even games and not to boost my ego, but but but... How does this anchor thingy work? Are our ranks based on the distance to some steady player? Why do they get periodically adjusted, and how come they never seem to get adjusted up?

daal wrote:What's up with this anchor business? Before Christmas, I was quite happy starting the holidays having managed to get all of my playing accounts up to 3k and then *poof* Christmas is over and they are all at 4k. I know that the purpose of ratings is to get even games and not to boost my ego, but but but... How does this anchor thingy work? Are our ranks based on the distance to some steady player? Why do they get periodically adjusted, and how come they never seem to get adjusted up?

Ratings systems are good at reflecting the relative differences in skill between players. However, all ratings systems drift, because there is no absolute measure of skill. The use of anchors is a way of reducing drift.

Back when I ran the New Mexico ratings system, after a couple of years it became obvious, from players who visited from elsewhere and from visits by our players to other clubs, that our rankings were not aligned with those elsewhere in the U. S. So I gave everybody a 1/2 rank promotion and altered the system. A couple of years later I did not find a reason to make any adjustments. We anchored our ratings to clubs on the coasts.

As for why, in your experience, adjustments are one way, is perhaps that the ratings systems are based upon theories that do not include anchors. The adjustments that I made to our ratings system to take ratings deflation into account were purely empirical, not based on any theory. It may be that the characteristics of the combination of the go servers and their ratings systems produces a bit of ratings inflation. We got ratings deflation in NM, I think, because our dan players were young and increasing in strength as much as most of the kyu players. (That is not typical, I believe.)

daal wrote:What's up with this anchor business? Before Christmas, I was quite happy starting the holidays having managed to get all of my playing accounts up to 3k and then *poof* Christmas is over and they are all at 4k. I know that the purpose of ratings is to get even games and not to boost my ego, but but but... How does this anchor thingy work? Are our ranks based on the distance to some steady player? Why do they get periodically adjusted, and how come they never seem to get adjusted up?

They are probably adjusted because the anchor's strength changed or they became inactive. As for why it always seems to be adjusted down, that is probably just selective memory. We want to associate any rise on our rating graph to our improvement.

emerus wrote:We want to associate any rise on our rating graph to our improvement.

That's a nice way to put it indeed.

Optimistically, you could think of this as a good chance to trick yourself: Since you know you are clearly at least 3k in strength and all your accounts are 4k, then you must be able to easily beat your opponents!

Because you can easily beat the now 4k opponents, if the ranking system drifts back to where it was before - poof! You're 2k!

Disclaimer: This only works if you don't think about it too much. Think too much, and then you're back to 4k!

I think thread drift should naturally be in the upward direction. Sure, it's hard for high dan players to improve, but the average skill of other players should improve with time. But they are playing each other, so that improvement won't show up, so their rating will need adjusting. Except I just gave a hypothesis that suggests the opposite of what daal is asking. Never mind.

Happy New Year.

I've been playing long enough to remember the great rank inflation of 2006? 2007? I went from 14kyu to 7 or 8 kyu overnight.

Of course that wasn't from changing anchors.

Thanks for posting this - I'd been working really hard for my most recent rank increase (which I finally earned a couple weeks ago). Then, I happened to lose several games coinciding with this huge rank shift, and I was just stunned that I not only had lost my rank increase, but I was a considerable distance away from returning to my previous level. I was actually really bummed out until I came across this topic, checked several accounts and realized that most everyone's rank has the same sharp downward spike.

Was this KGS rank adjustment publicized anywhere?

hyperpape wrote:I've been playing long enough to remember the great rank inflation of 2006? 2007? I went from 14kyu to 7 or 8 kyu overnight.

Of course that wasn't from changing anchors.

same for me, and about the same time period -- made my transition from DDK to SDK rather anti-climatic

I pretty much quit tracking my rank progress after that

xed_over wrote:

hyperpape wrote:I've been playing long enough to remember the great rank inflation of 2006? 2007? I went from 14kyu to 7 or 8 kyu overnight.

Of course that wasn't from changing anchors.

same for me, and about the same time period -- made my transition from DDK to SDK rather anti-climatic

I pretty much quit tracking my rank progress after that

I think you're both talking about the server upgrade from cgoban 2 to 3, where, indeed, rank jumps of 6 or more weren't uncommon.

(Apparently dan-level players got a rank drop, though.)

It is my firm believe that anyone who is too preoccupied with rank will always have problem improving quickly. There is an overall tendency for the general go population to improve anyway an so it the top of the scale is fixed it becomes harder for those lower down the scale it increase their rank, but they are still increasing their ability.

Do you understand more than you did three months ago? Do you feel stronger? Then you probably are.

Maybe I am at fault here . I played my last game on KGS on 11/30/2013 and since then I moved completely to Tygem. Before that I played steadily at KGS and had a stable rank at 1d (before I was a 1k for like 3 years). But it's not the first time stuff like that happened, anyway.

As for the rank, I may share a personal approach. I think if you win like 5 games against rank x, you can call yourself: x. Because at the end you showed that you ARE ABLE to beat those x-players and in a tournament no one of these players would give you handi, because they'd knew that you may not beat them regularly, but that you CAN beat them on an even game. In that sense I am a 2d-KGS, because I win like 3 out of 10 games against those fellas, but I never won against a 3d (even one or two wins wouldn't count because of the coincidence factor).

daal wrote:... and how come they never seem to get adjusted up?

Looking at the KGS archive graphs for some of the most active players seems to show that the similar overnight adjustment at the end of last December *was* up.

Counter-example: I did not play any KGS games in November, but my rank magically jumped from 13k to 8k! Of course, then I played a few games and quickly dropped back to 10k But there was definitely some sort of odd drift there, maybe that was anchoring? Our just the players I beat getting better?

http://www.gokgs.com/graphPage.jsp?user=sligocki
http://www.gokgs.com/gameArchives.jsp?u ... 3&month=10
http://www.gokgs.com/gameArchives.jsp?u ... 3&month=12

The optimistic viewpoint is that we have more and more professional players joining and playing on KGS as 9d and consequently pushing everyone else downwards in rank.