It was created on the old OGS (when I was a moderator and involved in its introduction) after luke, a 5d player, timed out down to around 5k, and subsequently entered and won a handicap division of one of the site title tournaments (maybe the Kuksu). With hindsight, I would have argued against rather than for its introduction.
EDIT: Actually might not have been handicap, but with divisions split by rank ranges (so he was playing even games against other real ~5ks, rather than even games with other mid/high dan players in the top group).
Without this rule, your ranking could fall very deep, we can imagine you played as a 4k and you are now a 15k… Imagine you gonna be a famous sandbagger for a while.
Without this rule the chips would actually fall where they may.
Folks who have the skills to play far above their apparent OGS ranking will beat other players one at a time (with limited effect on these outgunned opponents) until their rating stabilizes.
We apply this same logic for posting new player rankings as “?”. Things will right themselves if we trust the system.
Letting folks get well underway in a correspondence game and then annulling the game when they never return is effectively saying we don’t really trust our ranking system to correct itself over time.
It’s not we don’t trust. It’s we want to avoid too many boring or painful games to happen the time the ranking get fixed. That’s why we ask now to declare (vaguely) a level of play when you register (for example).
(Sorry… The above was quoted prior to your further editing)
I guess my own experience has been that when games I’m actively invested in playing are annulled because the other player abandons our game which was well underway, it feels both boring and painful.
That’s why I don’t like at all when I have doubts about the reasons of a timeout, especially when it’s well made, with just a few of them. I said I’d like to have a broader window but I guess it’s too much to run smoothly.
So I dunno, maybe best is to remove it? How would you manage yourself if this happens to you (losing 100 games at once)? Sandbag and say sorry each time? Start a new account?
How should we proceed for tournaments? (See @Uberdude historical explanation,)
If I were king for a day, yes, I would remove this convoluted game annulment algorithm (of course, as royalty, I’d first solve other world problems. )
How would I manage myself if I’d lost 100 correspondence games at once? Well, first I’d manage myself by not making 100 long term commitments I couldn’t honor.
But having forfeited 100 games, I would not feel entitled to be let off the hook through annulments while I leave my opponents in the lurch.
It should be far more appropriate to post wins for each of those 100 players who were playing in good faith before I had abandoned them.
(As for starting a new account? I could do that at any time for any reason. The annulment policy has no effect here.)
But that won’t solve the fact that we are not alone so that would still generate boring or painful games in the case of others who don’t give up their account.
Another way (especially considering tournaments) would be to get directly a more reasonable rating based on a previous one (like you were 5k so you get a 5+2= 7k without playing?
Yet this happens every time a strong player creates a new account, does it not?
I think the consensus on the preceding thread was that auto-vacation would be a good replacement for the serial timeout rule. (And, additionally, if possible, the option to disable vacation time for some challenges.) I believe we are patiently waiting for our technologically competent heroes to implement these changes. And meanwhile life goes on.
No doubt our techo-gurus have plenty enough to do. If a fix like this were on the “to do” list I’d be heartened (and patient).
As it stands, I never know when my active games will get annulled - or when they won’t - when an opponent doesn’t bother showing up after a few days of waiting.
Is there some way we can reasonably know whether it’s still worth wasting time thinking about a game before the next move?
Anyhow, indeed as you said, life (and death) definitely goes on!
I think the purpose of this rule becomes clearer if you stop thinking of rank as a reward or punishment. On ogs, the rating system is a tool to help players find fun, even (or appropriately handicapped) games. If a player times out of tens of consecutive correspondence games, those losses are probably not representative of their true skill in go but rather of some emergency in their life. I agree that signing up for a bunch of correspondence games you can’t complete is bad behavior, but the rating system is not a good way to hand out punishments.
Imagine a 2d player who times out of many correspondence games and ranks down to 10k. While ranking up, he will play many 5-10k opponents who expect an even game, then get crushed. These are the players who are being protected.
Regarding “?” players: yes, skilled players new to ogs may have a few uneven games, but they can select an initial rating and (I think) start as a dan. Their volatility is also much higher so their rating adjusts much more quickly.
Perhaps a better solution to serial timeouts is to increase volatility of the player who timed out and give him a “?”.
I share your vision that OGS should remain a place for enjoyable Go play… and that our rating system should be a tool aimed toward that larger mission - not as a stand-alone reward or punishment.
For me, as a newbie to the game, I regard any change in my (still paltry) rating as a reflection of whether I’ve any gained skill and insight through my hours of diligent study of the game. I come to OGS for the joy of playing with other like-minded folks around the world.
And so if I mess up my end game (something I’m prone to do) I forge ahead to maximize my chance for turning things around through learning. But at some point - unless it’s very close - out of respect for my opponent, I resign. I don’t simply disappear.
And if I were to simply disappear from a losing correspondence game (via multi-day timeout) and have the game annulled, I’d deprive myself of the healthy disappointment and stimulus to improve that a loss provides. (Whether that’s a reward or a punishment is in the eyes of the beholder.)
So perhaps I’m simply confused about how things work. It appears that if one ignores their losing correspondence games, the games can simply be annulled.
If this is actually the case, then I do believe there is something wrong with the incentives inherent in our rating system - however we might describe their larger purpose.
(A touch of sarcasm: Why not ditch the rating system and just give out participation prizes to all, so no one gets rewarded or punished for how their games turn out? )
There is general agreement that the current system needs adjustment and from time to time there are discussions about what alterations should be made.
Previously serially timing out would just result in annulments. However, I believe a change was made so that if it looks like the player was going to lose the game it will not be annulled but instead count as a loss, but other games will be annulled.
I appreciate the context. Being new to OGS I lack a good sense for the history of prior exchanges on many topics. The forum search function helps but doesn’t always provide the full picture.
A couple of thoughts on this discussion from someone who HAS had a problem with mass annulments.
I hands-down agree that a player should not have their game annulled because their opponent disappeared. Especially if the game is in a tournament → a win is a win even if it is on time, even if the timeout happens at move 10.
That said - mass timeouts can disproportionately affect a users rank, and cause the user to look like they are sandbagging.
I personally just started a new account rather than sandbag for a while.
SO - my thoughts on a possible solution:
Don’t annul the game. The winner gets their rating boost from winning.
If a user times out of more than X games in a certain time period they are barred from entering tournaments (similar to how provisional players get barred from some tournaments). Perhaps they have to play a certain number of games to completion/resignation after a mass timeout before they can start joining tournaments again.
If a user has a mass timeout - their rating variability should increase. They are basically back to being a provisional player - treat them that way so their rating can more quickly move back towards actual as they play a few games.
We can have a display hook that displays the highest rating a player has non-provisionally achieved. This would frankly be nice regardless - I often find myself looking at a players profile to see their highest rating so I can better understand how they are going to play. I’m sure there is even a way to balance the auto-handicap formula to account for large disparities in highest rating vs current rating.
@theswarmking Thanks for your reply and suggestions.
The most serious objection I’ve heard against allowing mass timeouts by a single player is that it would adversely affect the entire rating system via mass losses - i.e., therefore these games must be annulled.
The solution to this seems simple enough … Set a cap limit on the number of active games an individual player can participate in at any point in time. Limit the mass.
I have no strong view on just what that active game limit constraint should be (it could increase with higher trust levels).
Whatever # limit is chosen will be arbitrary. Then again, the way the current annulment protocol works is also arbitrary.
At least an active game cap policy can be clearly communicated and easily understood by all upfront.
As a result, the need to annul correspondence games in connection with timeouts will come to an end. Expectations become simple: You fail to show up to play one or more of your (limited) games, you lose.
But why? Being a provisional player isn’t mean to be a punishment, the reason we have provisional players is because there isn’t enough data for the system to determine their rank. From my understanding the rating and volatility come from glicko2. I’m not sure how you’d make having a serial timeout influence volatility, and I’m not sure it would be a good idea. If you have a suggestion for how to do it the ratings code is on github.
One issue is it might not stop you from checking the profile page. There are some people who peaked years and years ago, and then there are some people who peaked a week or a month ago. Not only that, but matchmaking is not done based on your all time best rating. Even if it is useful information, it wouldn’t prevent the stronger player being matched against much weaker players, it could just result in more cancellations.
I disagree. It’s wrong to limit the number of games you can play because that’s the way that a lot of players use and in most cases without timing out. Even some went quite high (over a thousand simultaneously) without causing any troubles.
It’s going against what people like to do.
If a players rating drops several ranks overnight → can you really say that the rating system should be “certain” about their rating? That much movement should be a flag showing volatility and decreasing certainty.
I’m not necessarily saying that we should mark them as provisional → but that something has happened that decreased the certainty of their rating so we should be handling it appropriately.
The idea is intended to show a pathway for these players to quickly regain their actual rank, not as a punishment.
The punishment is their rank falling to begin with, and ostensibly them being barred from tournaments for a period.
The problems that people keep bringing up in these threads is that a person dropping in rank suddenly makes it less fair for others that then play against them at a lower ranking. Keeping them out of tournaments for a period would prevent that kind of problem altogether. People playing one-off games have more latitude to pick and choose their opponents - so can more easily keep their games fair.
I also disagree with a cap on the number of active games. It would have to be arbitrarily low to prevent mass timeout problems → and would just discourage people from playing correspondence games at all. It would frankly be worse than just keeping the annul rule.
If you reset the rating to something like 3 levels below what it was at the start of the mass timeout I think everyone will still feel happy (And we put away the actual system). We ll keep the results (no annulation) and trigger the “reprisal” rating as soon as it is falling under 3 levels for the coming-back game.
That s answering the problems:
you can keep your account
you don’t lose so many levels and start sandbagging
your opponents who dedicated efforts will be rewarded
you can’t abuse a system in which you could go away for only two games to get them annuled as it was till now.
