I buy that. Not enough people report bad behavior, and they should. I believe that’s a separate issue, but I’ll make a suggestion to address this anyway, let me know what you think.
We don’t need to make “correspondence timeouts” a reportable offense in order to make this issue reportable. IMO the offense is “a sustained pattern of serial correspondence timeout”.
Making (single instances of) correspondence timeouts reportable would be a mistake in my opinion. It would put an unnecessary burden on mods, and unlikely to do much to reduce the issue. Life is going to cause people to time out occasionally. Also, people make mistakes.
To address your concern about people not submitting reports… I think it should be relatively straightforward to automate detection of serial-serial-timeout. The criteria would look something like this:
User has triggered serial timeout X times in Y months
I’m not familiar with serial timeout code, but I imagine it’s not terribly complex to detect and store these numbers. Parameters can be tuned, but I think it would be something like X=3, Y=4.
In combination with your suggestion to use logins or live games to suppress serial timeout annulment, I think we could reach a better outcome in a lot of these cases.
Congratulations. You are the first person to answer a question I posed a while back: does it make sense to prohibit reporting of correspondence timeouts while allowing reporting of abuse of the serial timeout rule? In other words, it’s okay to timeout once or every other game, but don’t timeout twice in a row. I think there is a bit of philosophical inconsistency in that; however, it is probably the best solution to the reporting problem.
I think the idea of automating detection of potential serial-timeout abuse is excellent if it could be implemented. However, even with that, it is unclear whether anything would be done. As I noted, my report of the worst case I ever saw—five times in four months with long strings—was dismissed. Recently I saw a case almost as bad, but didn’t bother to report it.
FTR: I did not suggest logins to automatically break serial timeout strings, although I love the idea. I was told it was @Gia who suggested that.
To get this thread back on track, here is what I think needs to be implemented:
Logins break serial timeout strings, as suggested by Gia.
Probable wins with XX% count as wins by timeout, as suggested by @GreenAsJade.
If possible, automate detection of excessive timeout strings (X strings in Y months), as suggested by @benjito.
If possible, reinstate a flag without bugs for people who timeout from correspondence games, as suggested by me.
So there could be a reputation system to deal with the moral/etiquette side of the equation.
And the rating system could be simplified to focus only on computing the best estimate of the rating.
i know that’s not what you are suggesting, as your #1 would not improve rating estimates. I’d just really like to separate the rating system from other reward/punishment issues.
does it make sense to prohibit reporting of correspondence timeouts
I think it’s worth a clarification.
It’s not “prohibited to report correspondence timeouts”. (*)
What is the case is that the moderation team doesn’t view “a regular uneventful timeout from a correspondence game as something that needs to be corrected”.
So for example, it makes sense to report a mass timeout that appears to be cheating…
… if you get a moderator that sees it the same way, they will take action.
( * ) the quoted sentence didn’t say that it is, but its been said or implied before
This is close to the status quo. There is no rule that says “don’t timeout twice in a row”, but moderators will take action on repeat mass timeouts.
I re-read the article and am glad this mechanic was implemented when I joined OGS.
I think it’s different.
Generally speaking, you will lose to players who are stronger than you and beat players who are weaker than you.
If that’s not the case I suspect you’re sandbagging
That said, this mechanic will only cause you to lose a lot to players who are better than you, assuming you have a lot of correspondence games.
If my understanding is correct, your rank will only drop a few kyu/dan at most and then it will not drop again.
This is completely different from the fact that everything will be considered a loss after timeout, which will cause you to drop directly to 25k.
I just ran into quite a few cases of this myself (timeouts when as White in the beginning of handicap games & being White in all of them due to rank being higher than average players in the handicap tournament) when a serious life event kept me from keeping up with games for a while, and also noticed this in another player’s profile – should something be done to adjust for handicap games ?
(especially as this would affect the stronger players in handicap tournaments disproportionately, basically making it likely to lose a large number of them in the event of a timeout, such as in such circumstances :
I propose that for handicap games, White’s timeout is always canceled if the number of moves is less than 20*n, where n is the handicap number.
For 13x13 and 9x9, it is 9*n and 4*n.
This thread is a year old and I may have missed further developments - if so, my apologies in advance. But it all seems quite straightforward.
“Time-out => loss” seems both fair and straightforward enough to both players. Isn’t that how most major competitive sports work? A no-show is a forfeit.
When a game has time limits, one’s window for making a move is limited by time. Yes, things come up in life, internet connections fail, etc. Unless your opponent is willing to make an exception, it seems the timed-out player is simply out of luck.
Two players commit to the terms of a game. If circumstances arise that are beyond one’s control, that player can always work their way back up in the rankings by future earned wins. It’s just a game. What’s the big deal?
The following argument seems thin: We need to protect others against a strong player who “sandbags” opponents through a depressed ranking arising from multiple time-outs. After all, that would-be sandbagger can merely play a bunch of ranked games poorly on purpose to achieve the same result.
Again, sorry if this topic has been resolved elsewhere. I’ll keep searching for relevant threads.
I agree. There are many ways for rank to fall. Timing out of games that should have been won is probably not the most common among them. A 4d loses to 3k who’s really a 5d? My condolences… but I would think this happens all the time, any time a strong player makes a new account. Theoretically even if the 4d’s rank falls, it should soon find equilibrium based on the results of future matches. These ups and downs are how the system works.
I would like to add that, having enjoyed now quite a few correspondence games on OGS, I believe I have timed out of only one of them. I have timed out of more than one live game. In both cases I timed out more or less because I couldn’t think of a good move in the allotted time. In the correspondence timeout, I felt the game was close, and I was not considering resigning… yet I didn’t play a move on time and therefore lost the game.
This is not a dishonorable way to lose, is it? In the live game is it so much worse to time out than to resign with one second on the clock? Say I have an six hours left in the correspondence game and I haven’t come up with a move yet, do I need to resign then and there if I’m not sure I’ll be able to visit the server again before the clock runs out?
In other words, timing out of a correspondence game should be an honorable way to lose, therefore it should count as a loss. I think this holds true even occurring for more than one game consecutively.
Activating vacation time seems like an OK solution. Unfortunately there too there appear to be some conflicting opinions about whether it is really honorable and fair to use vacation time. I say unfortunately because I don’t wish to disrespect any of my opponents or cause them to feel the game has became unfair or unfun or deviant from the agreed-upon settings—yet I do occasionally need to use that feature to avoid timing out of games that have been ongoing for weeks and months. I suppose I wish there was more clarity from the outset of a game that pausing for “vacation” is fair play. (The option to disable vacation in a challenge, suggested before, would help. I understand some tournaments already do this implicitly. Then one could avoid those challenges and be confident that one’s opponent accepts the validity of “vacation”.)
Anyway I commend the developers and the community for doing their best to navigate these difficult issues. There will inevitably be compromises and there’s no way to please everyone. Overall the whole project is very impressive and really enriches many lives, as well as the Go-o-sphere as a whole.
It hasn’t been dealt with yet, but it is ripening.
There is a proposal that I hope will get the nod. It’s in the queue for me (or someone) to get to fixing “the things that have to be fixed in order for this to be fixed”.
Serial timeouts in correspondence still occur, when players want to get rid of games where they think they are at a disadvantage. One guy had five series of serial timeouts in four months. I used to report those when I saw them, but don’t anymore, because all my reports expired without action.
BTW: “mass timeout” is a horrible term for this phenomenon, because of its lack of clarity. The original term for this phenomenon, serial timeouts, makes a lot more sense. The defining characteristic is seriality, not massiveness. Two consecutive timeouts are serial, but they are not massive.
The right thing to do would probably be to force-annul all the correspondence games of the player, whenever the mass/serial timeout behaviour triggers. That is, even if they still have some time left on some of them.
Then they don’t get to abuse the system, at least not at a per game level. In that case, at worst they could decide that their mix of games is disadvantageous overall and force annual all of them. But that’s much less useful than being able to pick the games.
Also, I just wanted to say that I’m very impressed how there’s always another abuse pattern that @Conrad_Melville has been investigating and tracking . It sucks that so often the resolution is that “and then the reports expired and I had to stop doing it because there was no follow-up”
This may have already happened years ago with OGS moving from a simple Elo rating system to the more advanced Glicko 2 with its uncertainty and volatility parameters.
I think the serial timeout rule likely causes more problem than it solves, and would be in favour of simply scrapping it, at least for a trial period of several months.
I haven’t missed your point. Simply considering that whatever you argue on the difference between mass and serial, triggering it just with two consecutive time out is too short and even more prone to abuses.
IMHO that may have been chosen because of the complexity for the rating system to manage a longer serial of timeouts afterwards, when a larger amount of time has passed. But the choice of two still remains doubtful.
. It sucks that so often the resolution is that “and then the reports expired and I had to stop doing it because there was no follow-up”