KGS Ranking adjustment?

[xela]

Did you actually observe people who play in a lot of tournaments and demote themselves? Personally I've never seen that.

Yes, I've seen that. If someone notices that they've been losing all their games in recent tournaments, they sometimes conclude that they should be playing at a different rank.

[dfan]

In the US, AGA tournament rules do not allow someone to enter a tournament at a lower rank than their AGA rank (see https://www.usgo.org/sites/default/file ... ns2014.pdf, rule VI.A.1.d).

[gennan]

Did you actually observe people who play in a lot of tournaments and demote themselves? Personally I've never seen that.

I've seen it a lot actually. Looking at the top 25 of the Dutch rating list(http://www.europeangodatabase.eu/EGD/cr ... dgob=false), I already spot 8 players who demoted themselves in recent years, even though their previous higher rank was officially awarded by the Dutch ranking commission. It's just that most have seen their rating go down slowly over the years and at some point they give in to the rating system (or maybe they stop playing in tournaments).

Some examples from active strong players in the Netherlands:
http://www.europeangodatabase.eu/EGD/Pl ... y=10298596
http://www.europeangodatabase.eu/EGD/Pl ... y=10233861
http://www.europeangodatabase.eu/EGD/Pl ... y=10501216
http://www.europeangodatabase.eu/EGD/Pl ... y=10986514
http://www.europeangodatabase.eu/EGD/Pl ... y=10486410

I'm sure there are many cases like this in other countries and at all levels.

[gennan]

In the US, AGA tournament rules do not allow someone to enter a tournament at a lower rank than their AGA rank (see https://www.usgo.org/sites/default/file ... ns2014.pdf, rule VI.A.1.d).

I skimmed those AGA Tournament regulations. I think this part is interesting for the discussion:

Appendix F: Re Self Promotion and Assignment of Provisional
Ratings
The TD has broad discretion to assign players to appropriate ratings for the event. Players who feel their
ratings are lower than their actual playing strength often request to be assigned a higher rating for the
purposes of pairing the tournament. The TD may and should take any evidence into consideration to
determine whether or not to grant the request. The TD's discretion in this matter is absolute. Any
adjustment (even less than a whole stone) upward is permitted. Downward adjustment will be made only
in a clearly demonstrable case of disability, such as a stroke. Lack of recent practice no matter how long,
is insufficient.
The TD has the same discretion regarding players who do not yet have an AGA rating. Ratings assigned
in other systems are the ideal gauge, but results from casual play reported either by the player or others
may also be used.
If the initial estimate turns out to be clearly wrong after a round or two, the TD may make an adjustment to
provide appropriate competition for the remainder of the tournament.

So an AGA Tournament Director (TD) has the power to bump a players' ratings upward by any (small) amount that he considers appropriate.

The EGF doesn't have such regulations about ranks and the rating system. Instead, each national federation has its own regulations (some even have their own rating systems).
But regardless of national regulations, an EGF TD doesn't have many options to change a player's EGF rating.
1. He can set the intitial rating of an unrated player (in many cases this will just be the declared rank of the player, but there is a lower limit of 20k.
2. He can adjust a rated player's rating by bumping a player's rank upward by 2 ranks from his previous highest rank in the tournament file upload. This will bump the player's rating to the higher rating.

The reason for triggering this bumping mechanism (called a rating reset) will differ in different countries. I think self-promotion is usually accepted up to a certain rank. In the Netherlands, self-promotion is allowed up to 1k (higher promotions are awarded by a committee, but those are almost always by 1 rank only, not triggering a rating reset). I think in Belgium and France the self-promotion limit is lower, but I don't know the exact limit (10k or so?). In Germany there is no limit AFAIK (it's always self-promotion).

So the EGF system has no mechanism to bump a player's rating by a small amount. So after their first tournament, players generally have to "earn" all their rating points by taking them from their peers, unless a player's strength increases so quickly that a 2 rank promotion is warranted (but still, the TD may not allow the triggering of a rating reset depending on national federation regulations).

I suppose this difference in rating bumping policies could make higher EGF ratings tougher than their equivalent AGA ratings, even though the AGA rating system has higher winrate demands. But it still depends on the frequency by which AGA TDs use their power to bump players' ratings in practice. Is it common or exceptional?

[jlt]

Looking at the top 25 of the Dutch rating list(http://www.europeangodatabase.eu/EGD/cr ... dgob=false), I already spot 8 players who demoted themselves in recent years, even though their previous higher rank was officially awarded by the Dutch ranking commission. It's just that most have seen their rating go down slowly over the years and at some point they give in to the rating system (or maybe they stop playing in tournaments).

If I understand correctly, the self-demotion refers to the blue line ("declared rank" = Dutch ranking system), while the red line (GoR) continues to move smoothly? And you suppose that self-demotions w.r.t. the blue line occur at all levels and in several countries?

Assuming that's the case, then I agree that these self-demotions deflate national rankings (blue line), but I don't think this deflates the GoR for players weaker than 5k.

[jlt]

In the Netherlands, self-promotion is allowed up to 1k (higher promotions are awarded by a committee, but those are almost always by 1 rank only, not triggering a rating reset). I think in Belgium and France the self-promotion limit is lower, but I don't know the exact limit (10k or so?).

I don't know about Belgium, but in France you cannot decide by yourself to self-promote. On the other hand, a "reevaluation commission" can promote players, usually quickly improving young players or players who haven't participated in a tournament for several years and who have improved a lot since then.

http://ffg.jeudego.org/echelle/liste_jo ... values.php

[Javaness2]

Oh, I thought that in France you can self promote up to the rank of 4 kyu. When did that change?

[Javaness2]

The EGF doesn't have such regulations about ranks and the rating system. Instead, each national federation has its own regulations (some even have their own rating systems).

I think the EGF should really make the effort to change this situation. What is wrong with asking all its countries to agree to a standard system. If all the results are going to the same database, wouldn't it make sense to have results regulated in the same fashion? It's like every country has their own Brexit. Some countries don't even seem to know how to submit results to the database.

[jlt]

Oh, I thought that in France you can self promote up to the rank of 4 kyu. When did that change?

Apparently in 2010: http://ffg.jeudego.org/informations/off ... 100508.pdf

see page 8 "Création d'une commission permanente de réévaluation" but I wasn't there so can't confirm.

[gennan]

Looking at the top 25 of the Dutch rating list(http://www.europeangodatabase.eu/EGD/cr ... dgob=false), I already spot 8 players who demoted themselves in recent years, even though their previous higher rank was officially awarded by the Dutch ranking commission. It's just that most have seen their rating go down slowly over the years and at some point they give in to the rating system (or maybe they stop playing in tournaments).

If I understand correctly, the self-demotion refers to the blue line ("declared rank" = Dutch ranking system), while the red line (GoR) continues to move smoothly? And you suppose that self-demotions w.r.t. the blue line occur at all levels and in several countries?

Assuming that's the case, then I agree that these self-demotions deflate national rankings (blue line), but I don't think this deflates the GoR for players weaker than 5k.

The blue line is the self-declared rank (but for 1d+ it is in fact determined by the Dutch ranking commission). The red line is the EGF rating computed by tournament results. What you see is that first the red line drops below the lower bound of their official rank. When it stays lower for a long time, the player eventually gives in and requests a demotion to a lower rank by the ranking commission. When the rating is indeed lower than the lower bound of their rank for some years, the ranking commission usually grants the demotion.

[gennan]

The EGF doesn't have such regulations about ranks and the rating system. Instead, each national federation has its own regulations (some even have their own rating systems).

I think the EGF should really make the effort to change this situation. What is wrong with asking all its countries to agree to a standard system. If all the results are going to the same database, wouldn't it make sense to have results regulated in the same fashion? It's like every country has their own Brexit. Some countries don't even seem to know how to submit results to the database.

I think it will be hard to find a uniform system that all national federations will accept. There is a lot of variation and I expect that every national federation thinks their system is better than all of the others.

[RobertJasiek]

Today (2020-05-05), the effect of my opponent (with equal rank) losing his (even) game was that his rating-rank increased. Any rating system allowing such anomalies is fundamentally flawed. (Not to mention days with 80 or 90% wins not being enough to increase one's rank.)

[jlt]

How do you know this is the "effect" of losing his game?

Ratings continuously fluctuate, whether you play or not. Maybe he was n+0.99 dan, the game made him lose 0.02 dan but fluctuations made him gain 0.04 dan?

Three days ago I played a non-rated game on KGS. At the beginning of the game I was 3k, in the middle game 2k, and after the game back to 3k.

[RobertJasiek]

I know but it makes no sense. Ratings should be updated often enough so that anomalies do not occur.

[Tryss]

I don't see why you consider these things "anomalies". The goal of a rating system is to estimate your true rating. With KGS rating system, each time you play against an opponent or each time a previous opponent has his "true rating evaluation" reevalued, your "true rating evaluation" gets reevalued.

Meaning it's because the rating is reevaluated more often than just after a game and with more information than just the last game that you see these "anomalies"