Auto-annulments ... a poll & a proposal ☑

Feedback
1

OGS’s current auto-annulment practice (prompted by serial timeouts) has been much debated.

One possible alternative would be to limit the number of simultaneous active correspondence games for (1) a relatively new OGS member without an established track record and/or (2) an OGS member with a poor track record for multiple time-out escapes.

I appreciate that any actual policy changes require support from both the community and, ultimately, OGS leadership.

But in an effort to assess community views, I respectfully pose a simple poll.

Please select the choice that best reflects your opinion …

  • YES, we should limit the number of simultaneous active correspondence games for new or poorly behaving OGS members.
  • NO, we should not limit the number of simultaneous active correspondence games regardless of an OGS member’s longevity or record for poor time-out behavior.
0 voters

For those who answer NO, thank you kindly for your responses, but please don’t waste your valuable time reading any further as you won’t likely be interested.

If you answered YES, you may find the following detailed proposal (subject to further refinement and consideration) to be of some interest.

Regardless, I appreciate your candid thoughts on the topic. :vulcan_salute:

= = = = = = = = = = = = = = = = =

TWO POSSIBLE POLICY REVISIONS TO CONSIDER:

#1 Correspondence games that progress beyond the current minimum move thresholds should not be annulled automatically in response to a player’s serial timeout. Instead, as with a live game, a correspondence game should be deemed a loss against the timed-out player unless OGS’s moderation process identifies a separate justification for annulment. Note: No changes to OGS’s overall moderation framework are being proposed.

#2 To limit the potential adverse effects of mass serial timeouts on the OGS rating system, a cap on the number of simultaneous active correspondence games should be established for newer members and/or those with a record of poor timeout behavior.

= = = = = = = = = = = = = = =

#1 RATIONALE FOR AUTO-ANNULMENT CHANGES

Reasons for this recommendation are based on considerations of fairness, respecting time controls, the irreversibility of annulments, transparency, and the need for symmetry in player treatment.

FAIRNESS: Once the current minimum move thresholds are reached, it is unfair for a player to have a correspondence game auto-annulled if the opponent fails to fulfill the commitment to play to completion by timing out. Whether the timed-out player has also failed to respect time commitments made with other players in other games is irrelevant to the game in question. The rules of each game, including time controls, are known to both players from the outset. There is no rule in Go that specifies that running out of time in another game with a different player will affect whether the current game either stands or is annulled.

HONORING THE TIME CONTROLS: By its nature, a correspondence game provides ample time for deep thought and reflection. In contrast, the accelerated pace and constraints of a live game force quick play and effective time management. When a player times-out in a correspondence game, it reflects a lack of commitment to honor the game’s agreed-upon terms; it does not reflect having lacked sufficient time to think nor having suffered an intermittent connection problem. While timing out in a live game is appropriately treated as a loss by default, the current auto-annulment algorithm in correspondence games can shield the offending player from this same outcome. If time controls are to be meaningfully respected, they should be treated the same in both cases.

INABILITY TO UNDO AN ANNULMENT: There are any number of circumstances under which an annulment, in retrospect, may be deemed unnecessary or inappropriate. Inasmuch as annulments are not reversible, the implementation of an auto-annulment practice produces results that cannot be reconsidered upon closer evaluation. Although moderated annulments are also irreversible, deliberate human consideration mitigates the risk that an instant (auto-)annulment will be regretted.

PLAYER TRANSPARENCY: There is no reliable way for a player to anticipate whether an active correspondence game will be annulled by an opponent’s time-out while it is still underway. Once a game has started, the current protocol for auto-annulment hinges on both past and future occurrences in one or more other games which may or may not have yet timed out. As a result, there is limited transparency concerning the risk of having one’s game annulled, leaving players unable to discern the consequences of an opponent’s behavior.

PLAYER ASYMMETRY: The current auto-annulment protocol often treats each player in the same game differently in the face of a time out. If one player has a previous timeout, timing out in the current game can produce an annulment. However, if the other player (with an otherwise clean record) times out in the same game, it is treated as a loss that is not annulled. It is not appropriate that the same timeout behavior by either player participating in the same game under the same time controls produces a different outcome.

#2. RATIONALE FOR LIMITS ON ACTIVE GAMES

Reasons for this recommendation are based on protecting OGS’s rating system, aligning trust where it is earned, promoting a culture of responsibility for commitments made, and mitigating circumvention.

PROTECT THE RATING SYSTEM: To mitigate the risk that an excessive number of games may be timed out by a single player’s failure to make timely progress across a large number of active correspondence games, players with less than [6] months activity on OGS and/or players with a record of multiple timeouts are limited to no more than [25] active games at one time [These figures are illustrative and subject to refinement]. The serial timeout auto-annulment policy of correspondence games is intended to protect the rating system from the effect that mass timeouts might otherwise inflict. Although it is unclear how many such mass timeouts would otherwise occur, avoiding system-wide rating inflation through game limits could help further protect the rating system.

DEMONSTRATE INCREASED TRUST FOR ESTABLISHED OGS MEMBERS: For players with longer track records who lack evidence of multiple timeouts there should be no limit placed on the number of simultaneous active games. By lifting the limits (imposed on newer and/or poorly behaving members), OGS sends the signal that longevity and adherence to proper game completion are valued traits that earn community trust. (Note: Players both with and without longevity would still be subject to warnings and escalated sanctions for timing out of one or more games, with the possibility that a game limit might be reimposed.)

EXPECTATION FOR SELF-MANAGEMENT: It should be understood that by accepting more games than one can reliably manage, a player invites the risk of multiple timeouts which, instead of auto-annulment, will be treated initially as losses by default. Game limits on new and poorly behaving players highlight the community’s expectations for responsible time management.

CIRCUMVENTION THROUGH MULTIPLE ACCOUNTS: For those players who seek to bypass the consequences of poor game behavior by establishing multiple accounts, imposing a game limit on new members would help mitigate the extent to which such players can immediately return to disrupting the OGS community. Protecting the community from large scale poor behavior by one or more (ostensibly) new members limits the potential for community disruption.

= = = = = = = = = = = = =

If by some chance you are supportive of one of these two proposals, but not both, please feel free to leave any comments or concerns in response to this thread. Of course, any and all comments would be welcomed. Thanks again!

2 Likes
2

Should say “Game limits…highlight….”

1 Like
3

Good catch. Fixed :+1:

4

I like the idea of limiting the number of games people can play, even if the limit was quite high, say one or two hundred games, and affected all members. I’ve played against people who struggle to keep on top of their games, timing out of a bunch and using their vacation as soon as one day becomes available just to get some breathing room. For players that play hundreds of games, reducing that just by a factor of two or three might help. The issue with these players isn’t just timing out, they also tend to drag tournaments on, consistently being the last players with ongoing games. You could say “just don’t play in tournaments they are a part of”, but the issue is, with hundreds of games ongoing, they’re in a lot of tournaments.

Some potential issues with signing up to tournaments & cyclical behaviour
I think you need to elaborate more on how you think this would work in practice as I see some issues with tournaments that would need to be addressed. Let’s say someone has 25 correspondence games ongoing, which is their limit, would they be able to sign up to a tournament that does not start for a long time? Even if it would be expected that most of the games they are currently playing would finish by then?

What if they have, say, 20 games ongoing and the sign up to a simultaneous McMahon tournament? Would you prevent them from joining based on the theoretical maximum number of games they might have to play at the same time, if that were more than 5? Even though realistically they may only have to play fewer than 5 games? Would you allow them to sign up in case it would be okay but once the tournament starts, if that would push the player over the limit, then boot them? What if the start of the tournament was triggered once the requisite number of players joined? The match would start, kick one or more players, and then there wouldn’t be the numbers that the tournament creator wanted.

Would there ever be a point that someone would be permanently limited? If so, then I think people would be encouraged to abandon their accounts and start a new one. If not, then you might see people vacillate between having a limited number of games, and then timing out a bunch once the restriction is lifted, only to then rise and repeat.

Another Suggested Change
I favour making it an offence to trigger the serial timeout feature too often. I’d favour making it automated, to reduce the workload placed on CMs, although the trade-off is that someone would have to develop the feature. I’d also favour increasing the threshold at which the serial timeout feature kicks in, I believe it currently only takes two games in a row. Maybe increasing it to five or ten. The feature is also referred to as the mass timeout feature, I believe this is because its purpose was to address mass timeouts. Timing out in three or four or even eight games will not affect the rating system too much, so I see no issue with increasing the threshold.

If we were to do this I think warnings, and eventually suspensions, would be a better deterrent than periodically restricting the number of games an abusive player can play at one time. Currently escapes in correspondence games are not an offence, so it could come as a shock to some players that they are suddenly getting warnings/suspensions. We could have a transitional period of a few months where people would receive a friendly notice/reminder that escaping is not allowed instead of warnings/suspensions.

Even with this system I’d still favour a high cap, say 200 games, on all players. But if some cap is implemented, whether a permanent high cap on all players or a low temporary cap on new users/abusers, you’d have to find a way to resolve the potential issues with tournaments that I mentioned earlier.

Sure, but there are also no rules in Go that say you can’t play hundreds of correspondence games at once. :stuck_out_tongue:

EDIT: Removed link as I realised a lot of people would not have access to it.

3 Likes
5

@PRHG Lots of great points raised.

I fully agree that effectively calibrating a simultaneous game limit (or set of graduated limits?) remains key. And avoiding unnecessarily bollixing up responsible tournament participation would also be an important consideration to work in.

The tactic of opening multiple accounts to dodge the consequences of prior poor behavior is, at least in part, the concern behind establishing the longevity requirement to access unlimited simultaneous game privileges. At least this puts a speed bump in the way.

Thanks for your thoughtful reply.

6

@SomeGoGuy Are you aware of spicydragon, rumoured to be Shibano Toramaru 9p’s big brother Shibano Ryunosuke 3p, who had hundreds of simultaneous correspondence games, over 1000 at one point, and managed them by blitzing without timing out? Why should he, and indeed his many opponents who had the rare chance to play a pro, be prevented by your proposed restrictions?

2 Likes
7

That would probably more an issue for a blanket cap on all users, like I mentioned. @SomeGoGuy was advocating for a limit on users who are not established, or those who are timing out a lot. spicydragon is not a new user, so that part wouldn’t apply to them, and if they can manage games without timing out it wouldn’t affect them either. That being said, a number of their recent games have been annulled because they triggered the serial timeout feature.

8

As noted by @PRHG above, under a simultaneous game limitation as proposed, an established OGS player without a clear track record for serially timing out would not be subjected to any game limit. So sounds like it would work in the extreme? :+1:

1 Like
9
  1. How serious is this problem? I don’t see many people complaining about it other than OP
  2. Why aren’t we solving the annulment problem instead of trying to solve another problem? (which may not be a problem in the first place) I feel like this is a long way around the original problem. Are we assuming that nothing can be done about the annulment when timeout problem?

Edit: I just read your #1. I didn’t realise you were talking about the original problem since there was only one poll.

Basically I agree with your #1 but disagree with your #2, though I’m not sure whether #1 is the best solution to go about it.

5 Likes
10

Thanks for clarifying. I suspect we largely agree. Personally, I’m a bigger fan of #1 than I am of #2, but respecting the concerns I’ve heard about the effects of mass timeouts on ratings, I proposed a way to mitigate that potential issue. I hoped the combination might be more acceptable.

If you’ve got a better solution than #1 I’d be all ears.

11

Is it really a serious problem though? Are we actually using a sledgehammer to crack a nut? Are all the impacts just purely imagined? I’ve heard @Uberdude say that there was only 1 actual case.

3 Likes
12

You ask a fair question. Short answer: I don’t know.

If so, the proposal offers a solution.

If not, it’s unclear what harm it does to reinforce the idea that folks should not abandon games before they’ve been finalized. Seems reasonable enough.

13

To be clear, there was one case which led to the introduction of the serial timeout rule on old OGS long ago. The rule has since been applied many many times. I don’t know how often for its intended purpose where the player has left OGS permanently or for some real life emergency, versus undesired or abused cases. I suspect the latter may outnumber the former. My preferred solution to the problems the serial timeout rule causes and is intended to solve is to simply drop the rule, and instead use rating uncertainty, giving glicko2 some extra help if it needs it.

5 Likes
14

Instead of a hard limit of the number of games, maybe introduce a more soft rule:

You are only allowed to create / join new games / ladders / tournaments when you don’t have too many games waiting for you.

This way, players who do cope with a high load can have many ongoing games, but players who already have most of their games waiting for them and who are constantly behind are hindered from increasing their load even further.

3 Likes
15

@Azedo A very workable improvement.

I’m beginning to wonder if the game limit idea is causing more resistance than simply eliminating the auto-annulment policy would by itself?

The game limit was offered for those who fear “mass” time outs. But perhaps that’s less of a concern than assumed?

16

I am fully in support of removing the mass timeout rule and fully against this proposal. It seems an unnecessarily convoluted way of getting at your stated aim of reducing or eliminating mass timeout annulments.

2 Likes
17

@junubia I can’t disagree. In trying to ameliorate the mass timeout concern, I was drawn to a more complex combination of proposals than I might have otherwise put out here for consideration.

But having said that, there are other benefits to be derived from limiting active games for those who abuse timeouts. Reinforcing expectations for good community behavior and mutual respect for bringing one’s games to proper completion would certainly be positives.

Whether a game limit (for new unestablished players and/or poor timeout behavior) would be a net plus or negative would then be in the eyes of the beholder. On balance, I think it could be constructive. But it’s the auto-annulment policy that is higher on my list of issues to revisit.

18

I thought there were other good solutions to this well-discussed problem, which are on the development docket. (Could be wrong about that.)

It is something that could be improved. But regarding this proposal, I have an immediate bias against creating formal tiers of users based on behavior. It reminds me of the “was your opponent a jerk” feature.

Also, I timed out of a correspondence game today.

Please don’t drag me before the timeout tribunal, I promise I meant no harm.

4 Likes
20

My suggestions…

  1. Newbies by default should not be limited until they got timeouts.

  2. For a record of poor timeout behavior purposes multiple sequential timeouts without any account activity should be considered as single event. It is obvious that if player is completely offline for any reason they timeout all their games.

  3. Correspondence tournament games for the limiting purposes should be considered as single game.

  4. Ladder games should not be accounted.

21

@mikhail.trusfus These sound like constructive ideas to further refine the simultaneous game limit concept.

Any thoughts on the idea of lifting the current auto-annulment protocol?