Derate games of short lived accounts?

OGS Development
1

Having checked my profile page I realized one of my very recent opponents had closed their account. Their entire games history was played on the same day (and they may have closed it having found themselves engaging in anti-social behaviour). It struck me, rather than them sucking rating out of the system, why not unrate all their games (or most recent games) where they were playing with their rating farthest from their level.

As I understand it the system holds a bunch of recent games not fully resolved for rating purposes, which makes this possible.

(I had won our game, because they had stopped playing and timed out, but as the AI review said they were better had felt they were potentially quite strong and underrated against the many players they beat.)

4 Likes
2

People who won against one of the short-lived accounts may not like their game to be derated.

1 Like
3

Yeah, i think that might make sense. My gut tells me that thereā€™s often something fishy going on with those deleted accountsā€¦

1 Like
4

Hard to tell if anything is suspicious as closed accounts donā€™t get a reason given (never get marked as AI cheating as far as I have seen). I do agree, accounts closed as AI cheats should have all the rated games possible derated.

5

I do think that happens already, but Iā€™m not sure.

But I can see the following problem that your approach would solve: Strong players register with a weak rating to have some easy wins, once they get their rank close to their real level they stop and start a new account. If this happens a lot it would contribute to an overall rank deflation.

6

Iā€™m not sure how it happens in practice that so may DDKs in correspondence games seem to play at 5-10 stones stronger. 14k will have flawless joseki, garbage fuseki, and flawless yose.

7

AFAIK joseki dictionaries are allowed in correspondence games, so that part is not that surprising.

1 Like
8

Thatā€™s pretty reasonable; you can take your sweet time reading joseki books, etc.

1 Like
9

I doubt so. Even strong dan players can lose 20 points compared to perfect play during yose.

2 Likes
10

Iā€™ve never heard that.

I think yose is the most calcuable part of the game.

No, Iā€™m not an expert. I just have feels about this.

11

Random example: in this game Championnat de France Open - Ronde 1 3d vs 6d, yose starts at move 131. At this point, White leads by 12.5 points according to AI. The final score is W+24.5.

2 Likes
12

This does happen, within reason and technical feasibility.

Some accounts clearly only started cheating recently - where this is the case, only games since that time are annulled.

There is a technical limit on how far back (time and number of games) we can anull.

Other than these constraints, cheated accounts get cheated, likely cheated, and rank-affected-by-cheating games annulled.

That said, I hadnā€™t thought you were suggesting annulment for the mere ā€˜crimeā€™ of being short lived. I imagined you were suggesting ā€œderatedā€ in the sense of ā€œlessenedā€?

Or maybe this was loss-in-translation and you really mean to say ā€œannulā€ when you said ā€œderateā€?

1 Like
13

I mean make the rated games unrated where possible. The games still exist normally that way.

The point being, the closed account sucks its rating points from the pool, never to be reclaimed back in further matches. My assumption is the maker of short lived accounts is probably experienced and underrated for much of their rating history.

14

I wonder why we started saying ā€œderatedā€, that was confusing :sweat_smile:

ā€œAnnulledā€ is the action that we have at OGS, which ā€œtakes away the rating of this gameā€.

I think that jlt is correct that folk would strongly dislike having their victories against these account annulled.

It would add to the already unfortunate feeling of ā€œI donā€™t want to play new accountsā€.

Now who would want to play a new account, when the game might rapidly become pointless (literally).

Note that legitimate new accounts have a reasonable chance of being correctly ranked (since people choose their starting rank range) and also a chance of being over-ranked.

Itā€™s only the nuisance new-account-sandbaggers who have consistently underranked accounts.

1 Like
15

It would be useful to introduce a delay of a few days before the account is actually deleted, so that we have time to check whether something was wrong, like sandbagging or AI cheating, and report the player. Once the account is deleted, itā€™s no longer possible to access playing history.

1 Like
16 
show_dialog("Your application for account deletion is currently being processed");
sleep_days(3);
player.delete_account();

:wink:

3 Likes
17

Is that already implemented?

I think that is not true as a generality. It is certainly true in some cases, but not even the majority, I think. A number of players leave because they were trying out go and decided they did not like the game. Other leave because they said they did not like OGS, or got into some controversy, and left in a huff. By far the largest category in my observation consists of habitually nasty characters: score cheats, stallers, and trash talkers of various kinds. They get some warnings, they know the jig is up, and they self-delete and reincarnate.

1 Like
18

No, that was a joke.

It says ā€œHow about I implement this suggestion by telling the person weā€™re processing their application, but in fact just doing nothing for a whileā€.

1 Like
19

My description of players as ā€œexperienced and underratedā€ doesnā€™t exclude them being ā€œhabitually nasty charactersā€. My proposal is to derate, ( e.g annull) the games to lower their impact on the rating pool.

20

I had the idea that annulment of games overshadowed the recorded result in some ways.