I mentioned it in the other thread also but we should add a question mark (?) to the rank if a user’s rank changes drastically just like KGS. The (?) remains until the user’s rank is stabilized again.
This, this, this, this, this.
Why are so few people talking about the rating problem this way!
Players whose rank change drastically from mass-timeouts should immediately go back to ? and should be unable to enter tournaments until they play enough games to stabilize their rating again.
I think we need some data, but it’s possible this won’t work. If the winners gain rating and the losers don’t lose rating it could eventually spread rating inflation across the whole site.
Also, two players could schedule a bunch of correspondence games against each other, both time out of them, and gain a ton of rating.
Fair point. (But live games would be faster.)
It strikes me as a bit ironic that we worry about schemes to lower ratings and we worry about schemes to increase them.
Perhaps we should just annul all games. 
main problem is what beginners feel
it doesn’t matter for them if they play against sandbagger or against correspondence-player-who-mass-timeouted
As a newbie I can attest: It feels very crummy to have one’s game annulled through no fault of their own.
[Hence my whining… and then being tasked with putting other ideas forward.]
This isn’t really how Elo works! If when there are mass-timeouts the “winners” gain ranking points and the “loser” gets reset to their original rank, the result will be to inflate OGS ranks over time.
Would correspondence players be happier with separate rankings for correspondence and live games? Then you could do whatever crazy things the correspondence go community feels work best for correspondence tournaments, and OGS could still point to the live rankings as the “official” site rankings that reflect tight matchups, appropriate handicaps, no sandbagging, no rank inflation, and so on.
There’s been a lot of discussion on separate ratings for different board sizes. From my understanding, that used to be the case, but then it was found that overall rating was a better predictor of how people would do. If you do a search you’ll find all the various discussions that have taken place. I don’t believe this will change unless someone can show per board ratings work better than just using an overall rating.
There’s no way to get data on that question. Most OGS users don’t participate in the Forums. Even worse, many if not most users don’t even know about the serial timeout rule, which is why we constantly get queries in the Forums and in the chats about “why was my game annulled?” History, going back many years, suggests that it s a major issue for people who frequent the Forums.
Although I’ve pointed it out before, this discussion takes no notice of the fact that timeouts are not prohibited in correspondence games, and it is consequently superficial talk in a bubble. The abolition of the prohibition on such timeouts greatly increased the illegitimate occurrence of serial timeouts. Many of these new timeouts are inadvertent in the sense that the perpetrator is not malicious, only rude. They may not even know abut the serial timeout rule. They timeout because OGS now has a timeout culture for correspondence games.
In addition to all this, the elephant in the room is that the legitimate use of the serial timeout rule is extremely rare. I doubt that its occasional occurrence is significant to the ranking system at large. This makes the whole controversy tragically comic to me.
Sure there is - remove the mass timeout rule, and see how many people complain
First, the implication of your original statement is that you want the data a priori. Second, what you propose would just give us more anecdotal evidence, not data in the usual meaning of the word.
Not at all. Your reading comprehension needs some work.
You’re the one who suggested we need “data” not me. There are also ways to get non-anecdotal data, but I think it would be overkill. We know the existing rule brings in complaints. If we remove the rule and there are no more complaints, that seems like a pretty strong signal.
Right, but the overall combined rating is a better predictor with current rules governing annulment for serial timeout and other edge cases. I don’t think anyone has analyzed yet whether the overall combined ranking still works if ranked correspondence results include meaningfully more serial timeout losses. My prediction would that these outcomes are possible:
It’s not really possible for there to be a consistently measurable improvement in predictive accuracy for correspondence games from this change without there also being a consistently measurable difference in accuracy between correspondence and live games. (Of course, it could just be worse for both types.)
At the risk of further complicating things, the behavioral response to a change in the auto-annulment policy may also be material.
It’s been said that there’s been an OGS cultural effect of allowing serial escapes in correspondence games without consequence (i.e., they just get annulled).
Changing the consequences may lead to a change in culture. 
There’s a database of all ogs games. It would be possible to write a script to determine the frequency of serial timeouts.
Okay. That would be interesting. Maybe someone will do it.
That isn’t at all what I said.
I don’t think the mass-timeout player should have their rank reset to original.
I think they should be treated as provisional for a number of games - since their rating shifted by a large amount overnight.
As provisional - they would be restricted from most tournaments (preventing sandbagging) and other players would know at a glance that their rating might not be accurate.
Their rating would balance out over time - as all ratings do.
And the winners of these games would properly have their rating increase.
The current system punishes good actors to prevent the bad actors from losing a lot of rating quickly.
There is absolutely a cultural effect in place of allowing serial escapes.
Because there are no consequences for it.
Personally → If I’m ever planning on resigning from a tournament in the future I’m just going to let my games time out so I don’t lose rating.
The last time I resigned from a tournament I dropped 3 ranks.
But hey - that’s what I get for playing on a server that promotes bad behavior over good.
I have to admit that I didn’t really vote or have an opinion on this matter since I haven’t really encountered had that problem, but now that I have, I can understand the frustration:
I had noticed that the opponent tended to time out in other of the tournament games and I was more-or-less expecting that annulment, so I wasn’t very surprised and I had many days to prepare for this kind of thing happening and I just tried to enjoy the game as much as it lasted.
However I did work hard to kill that group and potentially win that game on merit and not just timeout, so I am a bit annoyed that I put in the work to get into an actually winning position, just to get a “not today bro” kind of result, without any fault of my own. 
IMO it should be displayed as a win, and count as such in all ways, except maybe the ratings. That may confuse people though, if their rating doesn’t change after a win.