You raise an excellent point. It’s always wise to strike a balance between competing objectives.
Good spirited well-established players with high game volumes are an asset to the community. High active game count players who discourteously disappear on their opponents en masse are a liability.
So, how about a game limit cap that increases based on an absence of multiple timeouts?
Still not perfect, but perhaps a better balance than either extreme?
Putting a limit on the rating drop from a mass timeout could be a decent middleground.
Ultimately - the problem is that well-meaning players are currently being punished when other people just disappear and timeout. They are sinking time and energy into a game that will have no effect on their rating.
The annulment rule was a heavy-handed “solution” to a problem that happened once. And it goes against the spirit of these kinds of ideas → helping bad actors while disenfranchising good ones.
The wins should be counted as wins - and the winners should get points from these games. They sunk time and energy into them - they should get something for that.
@SomeGoGuy
That still sanctioning something which may be an accident.
That will induce to determine a boundary between an abuse and an accident which will be a source of disagreement and conflicts.
No I am not for limit when things can be repaired.
The rating deviation already increases in such a case:
Do you want more uncertainty? You can look at the code here. What exactly do you want to change?
Rank should not be a punishment. The entire point of the rating is it’s meant to help people find others of a similar strength to play against, and to help determine handicaps.
AFAIK there is currently no way to do this. The way rank is calculated is there is basically a formula that you feed someone’s game history into and it spits out their rating. I think that trying to implement your proposal would be quite messy.
I think if you tie the serial timeout feature into the rating system then you’d have the unfortunate consequence that changes to the serial timeout feature would then result in rating changes. Currently you only see rating changes when the rating system is modified.
Nothing messy if it’s not linked to your old rating but considered like a new registration (you give up the actual calculation for a fresh new start) with keeping the account name and history.
We could even process by declaring again your level (like the first time you registered) but I feel better to calculate your entrance right from your old rating before the timeout.
I broadly agree with your suggestions and have had similar ideas myself, but I would just add a precision to this barring from tournaments: only ones where a low rank is an advantage. So no handicap tournaments, and no tournaments where you are grouped or seeded by ranks. But an open even Swiss, or a ladder is fine (in a ladder a low rank is a disadvantage because you can’t challenge people so high via the sort r ladder by rank and your position plus x rule). This should be based on rank uncertainty being over some threshold. I would hope glicko2 would do this naturally after mass timeouts (if the average rank of your last 50 games is more than 2 standard deviations from the average of your prior 50 games then uncertainty should obviously be high), but if not some custom uncertainty boost would do the trick.
P. S. Above is in the context of ranks and avoiding sandbagging, a broader bar on the basis of them being unreliable in completing games could make sense still.
Clearly, you’ll all given this considerable thought and have commensurately nuanced and multi-faced alternative approaches to the current protocol.
And yet, perhaps this may be the problem?
Because I’m new to OGS I have a simple view. But oftentimes, simplicity wins the day.
The more complicated and multi-layered the OGS annulment procedure is for correspondence games - where timing out should not be excused - the less folks will grasp it and the more likely it can be “gamed.”
Alternatively, the more straightforward it is, the more likely it will be met with acceptance and regarded as fair.
The rule should be easy enough for a child to understand.
Simply put: Just as in OTB play, if you run out of time before a game is over, you lost. The other player wins.
Since there’s great concern that, unless such lost games are annulled, one player can wreak havoc on the rating system, then yes, limit the # of active games a single player can play until that player becomes a trusted member of the site.
Such a simple rule has the benefit of being easy to explain, easy to confirm, and easy to implement.
I realize this simple approach is anathema to some. Again, since I’m new here I can bring a “who-the-heck-is-that-new-idiot?” type of vibe to the discussion. Please don’t flame me. 
Perhaps we should just hide the annullment? Then people can enjoy their win, and the rank system stays intact.
Clever idea!
If all my annulled games were counted as wins, I might be 2 dan by now. Of course, actually playing at that level I’d have been repeatedly trashed until I wound up back where I am now. 
Yes! Just like how e.g. ranked 7x7 games are processed. It can just be a computation done on the fly, we don’t need to rely on a user-visible annulment.
@PRHG Yes - I would absolutely propose that the deviation should increase more than that. If I have some spare time I’ll look at the code and make a PR.
Rank should not be a punishment. The entire point of the rating is it’s meant to help people find others of a similar strength to play against, and to help determine handicaps.
While I ultimately agree that rank should not be a punishment → it is far more than just a way to help people match more fairly and determine handicaps. If that was all it was - no-one would care about AI cheating or the annul rules. Frankly - if that was all it was we should be able to just arbitrarily set our ratings to match what we think they are, and there wouldn’t be a rating formula from your wins/losses.
But I do think “punishment” is the wrong word here. The correct word is “consequence”
If you disappear from your games and time out - the consequence for that is that your rating falls.
For people that serially disappear - or that time out from a large number of games in a short period of time - we should be looking at ways to handle punishing the bad sportsmanship while also giving them a way to build back quickly.
We should absolutely not be punishing their opponents by giving them no credit for the time and energy they put into a game just because the person they were playing against timed out.
When it comes to tournaments, the game lost by timeout is counted as a loss, even if its result is annulled for rating purposes. AFAIK for ladders, the consequences are even bigger and you (might?) get kicked from the ladder altogether if you time out of a ladder game.
So the timeout losses are losses and carry competitive consequences even if annulled.
Sounds perfectly fair that negative consequences are faced by the one abandoning a game.
Would be equally fair and logically consistent to provide positive consequences for the one left at the alter.
The opponent moves further in a tournament or up a ladder, so the same amount of positive. I understand it doesn’t feel fair, though.
Agreed. It doesn’t.
I’ll suggest something entirely different (the whole mass annulment thing looks strange to me but I’m not the one to argue about it): If the player who times out is losing, the game is not annulled even if it fullfills mass timeout criteria.
IMO that prevents it being abused while still allowing it to be a feature?
Personally, I believe this would be an improvement - although it raises the need for a determination as to who is winning vs losing which isn’t always so cut and dried.
But especially for multi-day correspondence games, there’s ample time to reconnect if internet connectivity has become an issue. In such cases, a timeout ought to be a loss regardless of (a potentially fraught interpretation of) who is winning and who is losing.
You don’t show up, you lose without annulment.
This makes absolutely no sense. You understand that AI cheating makes the game uneven? You think that if rating was just about helping to find even games, then people wouldn’t care about AI cheats?
Again, this doesn’t make any sense whatsoever. If rating was just about matching people fairly then we should have no formula to determine rating? How does this follow at all?
Allowing people to chose their starting rank, or not, is a design decision. One reason you might not want people to chose is that beginners who are unfamiliar with the system might think that 1k is a low rating. There are arguments that go both ways, and some people will come down on one side or the other. I don’t understand how rating just being used for matchmaking and handicaps makes the choice clear.
I’m not entirely happy with the current state of affairs either. It’s not just the serial timeout feature though, there is no escaping offence for correspondence games currently.
I believe this is already the case. There are some thresholds that have to be met, but if someone looks like they will lose then it will count as a loss. If someone is behind but it’s still anyone’s game then it is annulled.
@PRHG To be clear → I don’t think there is a problem with the rating system. I believe it is more important to the players than just a basic level of information to help with handicaps and confer some feeling of fairness.
Frankly - your reaction to my statements tells me that you also believe this.
Because if it wasn’t more important then what I said is rather true → it wouldn’t matter if we set our ratings manually because the ratings wouldn’t matter!
Even more-so –> this annul rule is in place to ostensibly protect the ratings system by not having a single person drop in rank by a lot overnight.
My argument is that this doesn’t hurt the ratings system - that the ranking drop is a natural consequence of poor sportsmanship (allowing many games to time out) - and that the harm of annulling these games is actually to the players that should have won those games.
Even the weird idea that “if it looks like the player was going to lose, then it is a loss - otherwise it is annulled” doesn’t make any sense to me.
Do you know how many times I’ve seen games (my own and others) turn on a dime when the clearly winning player blunders. I witnessed a 150 point blunder in a game just the other day. I have personally won against a 3 dan player because they blundered at the end.
The game isn’t over until one person resigns or it is counted. If a person times out - that is a straight up a loss.
Annulling games should really only exist to retroactively correct harms (like a person was caught cheating).
How?
Again, this doesn’t make any sense. If ratings were just to help with fair match ups, then ratings wouldn’t matter? How does that entail people no longer caring about AI cheating? Why would it mean we should just let people chose their rank? Why would we get rid of the formula for calculating ratings based on wins/losses?
You argue that if ratings were just for fair match ups, then we wouldn’t care about all these things that make match ups unfair. This makes absolutely no sense. You keep restating it but seem unable to explain why all of these things would follow.
You understand the point was not to protect their achievement, right? The point was to stop someone having a significant drop in rating BECAUSE that would result in unfair match ups, not because it was some badge of honour that was being unfairly taken from them. The point is it could result in unintentional sandbagging, and people looking for a fair game would be matched up with someone much stronger than them. Such a person would be able to enter tournaments with rating restrictions that would normally prevent them from joining. The system might be a good idea, or a bad idea, but the reasoning behind it is for fair mach ups. Yet you keep saying that you’ve proven rating is about so much more than this.
Emphasis added by me. Yes, the game isn’t over unless someone resigns, the game goes to counting, or someone times out. You understand this is what we are talking about, timing out? How do you expect a game to turn on a dime… after a timeout? We are not discussing annulling ongoing games.
