I was literally spending my last month on a ranked game. Looks like my opponent hasn’t logged in for the last few days, and the server’s timeout protection kicked in and annulled the game. Link: dlhanson1015 vs. jane.eiffel
I understand the need for the rule, but this result still makes me really frustrating. My opponent had a solid record, didn’t seem like someone trying to mess with the system. Does anyone have idea on how we can address this?
Same thing just happened to me today. I don’t see how timing out is any different from resignation. It obviously can be abused to protect one’s rank in a losing game. Even if that is not the primary reason timing out happens, it is surely a large downside to annulment.
It’s different in a lot of ways. When you’re winning the game and your opponent times out sure, not so different. When you’re losing the game or the game is undecided or just still very early in the game, it’s quite different when someone just forgets to log into OGS for a few days or a week, or clicks the resign button.
I agree it is a downside, and if someone is regularly trying to timeout only lost games it’s something we can probably contact the user about.
On the other hand, when someone is playing a lot of correspondence games (which some users do) and then something comes up, there can be drastic drops in rank for that player like (extreme example say)
should the games not be annulled. That’s probably not going to be very enjoyable for that player should they return to play again, or for their opponents that they have to play to regain their rank.
Having to manually annul games is also not great especially if it’s like 10s or 100s.
I’m not sure what the best solution is?
One might think to annul only games where the player timed out and they were winning or something like this, so that when the opponent wins by timeout but was also winning the game, they get the rating they “deserve” say.
That can still lead to some strange looking situations where someone might have played and lost like 20 games in a row (by timeout and maybe losing slightly or by a lot) and statistically it probably suggests they should be much weaker than their current level, as how likely is it to lose so many games in a row? Then their rating is still likely to drop by quite an amount.
So I don’t really know what’s best.
Does it make sense to count the game as a win for the player who did not time out (and affect their rank accordingly), but count it as annulled for the timing out player (and have no affect on that player’s rank)?
This sounds like a recipe for rank drift since on average everyone is getting free points and no one is “paying” for those points
Unpopular opinion but what’s best is to CTFD about rank and play another game! 
Maybe
helps?
I’m on team timeouts should not be annulled. The rating system can handle the noise, and if not, don’t punish the players for it: two wrongs don’t make a right
Maybe the rating system will be fine, it probably will, but it doesn’t stop the problem that somebody who plays people that regularly time out will find themselves over ranked and someone that forgets to log in will find themself under ranked when they drop a number of games.
At the end of the day you want the rating update to mean something, you played better than this person in the game, you won, and so maybe you’re higher ranked given this new information. If you lost a game, maybe you’re not as high ranked given this new information.
When someone times out, is it really reliable information in which to base anything on? Should there be some points margin where you think “yes this timeout probably accurately reflects the result the game would’ve had anyway” ?
Why intentionally add noise is the point?
Managing your time schedule is part of the game. Losing by timeout is same as by points or by resign. In itself there are no better way to lose a game as another one.
Now there are etiquette and politeness, we should not encourage players to let the game timeout as a variation of a resign. But even when it seems relatively obvious we have to stay cautious that it can result from something else.
That annulation when losing by timeout is not fair. Especially now when players are somehow manipulating their ranks with it. Mass timeout shouldn’t be automated but managed ponctually by a mod on request only.
Exactly
I would be fine with abolishing the entire rule and timeout=loss. But from past threads on this topic, there is no consensus on this and good reasons for the current rule.
My compromise solution is to increase the number of timeout games before the protection/annulment kicks in. Currently only the first game in a series of timeouts is ranked, and all the rest are annulled. I would raise this to the first 3 games. In a mass timeout, the player’s rank is still relatively protected with 3 games counted.
In a situation where a player timed out in two or three games in a row but wasn’t otherwise completely offsite (as happened to me recently) the rule doesn’t kick in, as it isn’t a mass timeout.
Also with three games counting as lost, this is a disincentive to anyone who would try to game the system, and would hopefully reduce such incidents.
Apart from this I also fully support the recent proposal to change the language around annulled games. Calling a clear winner but indicating the game as deranked will be less bile inducing than White wins Game Annulled.
I don’t agree with this solution, but would be happy with it as a step in the right direction
It’s a good compromise when everyone is equally unhappy.
Wow. It’s a long time since I’ve played correspondence on this site. I had no idea we’d implemented such a crazy rule. In every other game or sport I can think of, turning up and playing your matches is a core skill. If you don’t manage to do that part, then you lose. If there’s a rating system, your rating should drop.
Sure, we can forgive people if they literally have to go to hospital suddenly and can’t get online. But how many times per year does that honestly happen on OGS? Surely they can just contact a mod when they get back? Do we really need an automated system (with all the potential for abuse and confusion)?
Is there any way to know this rule exists other than stumbling across the appropriate forum threads? I couldn’t find anything on the main help page at Home · online-go/online-go.com Wiki · GitHub or in the FAQ at FAQ - Frequently Asked Questions
Yeah I do particularly like watching correspondence soccer where a single match can take six months.
I mean I agree with you if the game is being played in one session. Each person has a certain amount of time they agreed to and managing your time is a resource like everything else.
Except in correspondence it’s about fitting your life around a Go game. What’s that you didn’t plan for your car to break down and that messed your plans up for the day, that family emergency, that overtime you have to do as part of your job, that college assignment that’s taking far longer than you hoped etc etc, well tough you didn’t manage your life properly around your Go game XD
Is the points for that one game that didn’t get finished that important? How often do you run into it, and how often does anyone actually “abuse it”?
Like we can make the exact opposite arguments fairly easily. I don’t think they’re obviously weaker arguments?
Exactly. This is one of the skills of playing correspondence. If people are making a habit of signing up for games that they can’t finish, it’s pretty disappointing for their opponents.
Anyway, regardless of what the policies are, I think documenting them is a really good idea. I did try searching, and it’s not easy to find the previous discussions or the reasons for doing it this way. I’d really recommend putting something on the help pages and on the FAQ thread.
There are a lot of good thoughts in this thread. Let me propose another compromise:
We should definitely tune the text and show clearly it’s a win by timeout.
As many already suggested, there’s no significant difference between resignation and timeout, especially under the context of correspondence games. Even with very little or no ranking changes, people will feel a lot better to see a win than annullment.
We should discourage the timeout anyways and reward the winner.
Managing the time is very important in correspondence games, and people tends to abuse the rules if the punishment from timeout is almost zero. In the current rule it’s only 1 losed game, it looks way too small to me. I suggest we should at least increase the number to 3-5. Not sure if it’s applicable in current ranking system, I’m proposing this new rule:
First timeout: 1 losed game
Second timeout: 0.8 losed game (the elo change are scaled down by a factor of 0.8)
Third timeout: 0.64 losed game (further scale down by a factor of 0.8)
Fourth timeout: 0.512 losed game
…
In this new rule, every winner will be somehow happy because they still get some points from this game. If you sum up everything, it’s no more than totally 5 losed games for the timeout person so still keep their rating relatively stable.
I’m not sure about the. It used to be that these games were silently annulled and we often got people complaining about not getting points for a game they won.
I like this idea. I feel that 3 games timed out is a good balance and that game 2 and 3 could have a lower weighting.
I guess in terms of finding out information going backward it was mentioned
though the tag has since been removed to say a player has recently timed out of a game.
There was an old suggestion to annul only if the player hadn’t been logged in in a while (i.e. they’ll probably time out of all games vs they are just ignoring losing games)
I guess this was the mention of inflation problems of not taking points away from losers but awarding points to winners
But in terms of
I suppose I don’t see much difference between these except that a few people should be happier. It’ll still be a random selection of timed out games so not everyone will be happy.
So for instance if you had 7 games in a round robin or McMahon where the next round of the tournament started unexpectedly (which happens a lot - I mean I timed out of fast correspondence games like this recently) then 3-5 of your opponents will be happy but a random subset won’t be.
Probably could be an improvement, but probably not guaranteed to make everyone happy 