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

Feedback
22

I think it is only the matter of fairness. It almost doesn’t affect the rank of the opponent. It is just a single annulled game for the opponent among lots of games.

I’m aware that there are players who don’t play multiple correspondence game at once. For example How to deal with a conspicuous player Annulment will affect the time to update their rank a lot. But considering timeout of an opponent as win for estimating the rank is not good too. I think in this case their rank will be incorrect. So the best way is to annul.

23

Thanks for your reply. I certainly respect your view but beg to differ for two reasons…

  • For new lower ranked players each game weighs far more heavily on their ranking. And newbies (i.e., their likely opponents) seem more prone to timing out - so it creates a poor introduction to OGS that’s hard to understand. After all, if the well-behaved player abandons, they suffer a loss. Doesn’t seem quite fair.

  • And if fairness dictates that time-outs should be annulled for correspondence games, then why are they not also annulled for live games? It’s unclear how the fairness principle differs based on time controls.

Anyhow, these are some of the considerations I struggle with under the current auto-annulment practice. :thinking:

24

I think annulments of such games don’t matter for newbies.

Who they? Newbies? Well-behaved player abandons? I don’t get it. Sorry but I’m not native English speaker. Please explain what you mean.

It is a good point I think. I think it is not about fairness. It is about rank estimation.

2 Likes
25

Yes, a fix is in development whereby vacation is automatically triggered when a serial timeout occurs. I should explain that I voted in the poll for capping games, because I think anything is better than the status quo. Until something is actually done, I favor abolishing the serial-timeout rule, I favor the vacation trigger, and I favor the game cap.

Tiers based on behavior already exist. Here in the Forums, we have Trust levels. On the main site, OGS rule violations are handled through tiers: warnings, final warnings, and suspensions.

Correspondence timeouts have not been reportable violations for the past 5 years. One consequence of this is that a culture of escaping has arisen in correspondence games, which has in turn led to many more serial timeouts than in the past. Many of these serial timeouts are doubtless “incidental,” in the sense that the player is not intentionally abusing the serial-timeout rule.

A further complication is that the serial-timeout system seems to be buggy. Years ago, a completed live game broke the seriality (because there is obviously no emergency if someone is playing live), but that is no longer true. I raised this issue in another thread a year or two ago, but I don’t think anything has been done to fix this.

2 Likes
26

With all due respect, they matter to the newbies :wink:

Sorry if I cut the thought short for brevity… Yes, if a well-behaved player abandons a correspondence game, just like with a live game, they lose the game. But if their serial time-out opponent abandons the same game, it gets annulled. Same game, same time control, same behavior, different outcome.

Good point. I understand that concerns about the system-wide ranking effects of serial time-outs is what originally prompted the current auto-annulment policy.

Although this issue doesn’t worry me too much, I do respect the concern as expressed by others. So this was why I included the game limit concept into the mix of the proposal.

Thanks for weighing in. My point in posting this thread was for us to gain a better sense for the community’s current thoughts on these topics. Seems we’re getting a healthy cross-section of views.

27

Edit: Meant to post a reply to @SomeGoGuy

I agree that there are issues with the current serial-timeout annulment system, but I strongly disagree with this proposal. Obviously I can only speak for myself, but I suspect that many others (including much of OGS moderation and leadership) have similar views. I think there are some pretty fundamental problems with the way you are thinking about annulments and I don’t think this or any similar proposal will be accepted until they are resolved.

Of course I can’t guarantee that everything in this post perfectly represents the positions of Anoek. The following is my impression of some of the design philosophy of OGS.

The purpose of OGS is to provide a great go-playing experience for its users. OGS is not in the business of judging people of punishing wrongdoers beyond what is necessary to ensure other users can continue to enjoy the site. OGS does not declare what is the “right” way of playing go or interacting with the site, except in ways that are necessary to support this goal. Thus users’ options are not limited unless there is a good reason.

Examples

Most go-playing sites only allow the standard board sizes. OGS allows anything up to 25x25. I have had some very enjoyable 2x25, 3x25, and 5x25 games. OGS allows extremely fast time controls. There is a risk of this surprising players or being used to exploit unsuspecting players, so there is a warning sign on these challenges - but they are not disallowed. Many would be tempted to say “3-second go is not real go” or “that is a clicking game, not a thinking game”. Personally I struggle with any byo-yomi shorter than 30s - but I am glad that players get to use these time controls if they wish.

Thank you @anoek for trusting the users of your site. I appreciate my fancy ogs cockpit.

What does this have to do with your proposal? You propose placing a limit on the number of simultaneous correspondence games a player can have. I hope you can see now how this conflicts with the purpose of OGS. Many people would not support this proposal without a very strong justification. I have never had more than 25 simultaneous correspondence games, yet I would be sad if this proposal was implemented. To me this feels a lot like “there is something that annoys me (getting my games annulled), here are some people that are doing something I can blame (playing many simultaneous correspondence games), let’s stop them from doing that”. This type of reasoning feels very threatening, and tends to produce bad outcomes.

I think even conducting this poll at this stage is a bit misguided. OGS is not a democracy. If we had a poll where 90% of forum users said yes to “should we remove the option for <5s time limits to preserve the respectability of go”, it still wouldn’t be a good reason to disallow bullet games. We need to be reasoning about whether this policy would overall improve the user experience on ogs. One poll I think would be good is “Have you been bothered by excessive correspondence timeouts”. This would help judge the size of the problem.

To decide if the solution is worth its costs, we would need data. Do players with more correspondence games time out more often? You seem to think this is true, and that it is being overburdened with correspondence games that causes them to time out. It is also possible that people time out for real-life reasons unrelated to how many correspondence games they have, and there is no correlation. Even if you have seen that many of your opponents who time out have many games, it is still possible that they time out at the same rates as everyone else and they just play against you more often. If having more games does not lead to more timeouts, then a sitewide maximum for correspondence games would not reduce timeouts at all. What about only applying the limit to “untrusted” players? Then you have to determine if the players restricted by this policy would tend to time out more often. My guess would be that

  • players with 25+ games do not time out significantly more frequently than other players, and
  • mass timeouts happen at most twice per year for almost all players.

If these two statements are true (and players stay “untrusted” for at most 6 months) then the proposed limit would bring no value to OGS.

@za3k and @siimphh have produced a database of all ogs games ( 2025 OGS game dump - #24 by za3k ). It would probably be very useful for collecting data on timeouts.

Fairness

A large part of your justification is about fairness. This is not a fairness issue. Timed-out correspondence games still count as wins for tournaments and ladders. The only difference this kind of annulment makes is that it doesn’t update the ratings of the players. The rank you deserve is the rank that accurately predicts your skill in go. Like with all other features of the site, we have a rating system because it improves the experience of users. This is because it is useful for players to know each other’s approximate skill, mainly for getting appropriately-matched games. Any method which does a good job of predicting predicting player skill is valid. OGS used to use some sort of sliding-window system where sometimes you could win a game and your rating would go down. You could even have a game where the winners rank went down and the loser’s rank went up! This was eventually removed because it was confusing to some people, and because fixing a bug made the complicated sliding-window system unnecessary. It was not removed for fairness.

6 Likes
28

@calebjdastrup You raise a number of thoughtful points.

FWIW, we agree that OGS is not a democracy as such. Yet getting a sense for the pulse of the community is valuable and I think we’re gaining that through responses - including yours above.

I sense we also agree that rankings are valuable as a means to promote appropriate player matches. As a player who takes his game commitments seriously, I’d like to be matched not only according to raw Go skill level, but also by shared respect for the playing terms agreed upon. For correspondence games, time limits shouldn’t be hard to honor.

Regardless, I respect your point of view and the spirit behind what you’ve shared above.

1 Like
29

The issue isn’t just that they’re burdened with a lot of games, one issue is if you time out of a bunch of games at once your rank will drop and you could be unintentionally sandbagging. This isn’t just annoying for the person who timed out, but all the other opponents who have to face them. The serial timeout feature protects against this somewhat, but there’s a lot of unhappiness with it.

The other issue is that escaping is an offence in live games, but it’s not in correspondence games. At least with live games people can receive warnings/suspensions which can correct or deter behaviour. This doesn’t really happen with correspondence games. Currently, people in correspondence games are free to timeout as much as they want.

2 Likes
30

I can’t remember correctly but I think on KGS if your rank drops a lot you go back to having a question mark in your rank? And it remains until you go back to your stable rank again. Not sure whether that’s possible here.

Yeah, IMO the rule just reduces the seriousness of timeouts in correspondence games, which shouldn’t be the case. People need to know that timing out has consequences, as John Wick said.

2 Likes
31

I suspect this is a big part of why people dislike the annulments. We only have “annuled” but we use that to both exclude a game from rank updates AND to signal the moral win/loss status of the game.

So in the UI annulment results in the game being grayed out and result stricken through, indicating there really wasn’t a winner. This is annoying if you really did win the game with a timeout, as in this case. On the other hand, we also annul games where the winner is incorrect because of the score cheating and people would be similarly upset if score cheating results still stood appearing as the cheater had won in the UI, even if the rank change had been undone.

So I think ideally we’d separate “moral winner” status of the game from “included in ranking” status of the game, so we could more clearly indicate who “really” won the game, because that’s what I think we as players are really emotionally invested in.

And someone might still care about the ranking effects of annulment too, like people care about how volatility affects rank updates. But I think people wouldn’t be nearly as invested.

5 Likes
32

If you lose 25 games in a row your rank will drop drastically. Unless the limit is much lower it won’t prevent accidental sandbagging. I agree that there are problems with the way we currently handle correspondence timeouts. I’m not convinced that a limit on correspondence games is an improvement.

1 Like
33

I initially voted “yes”, but changed my vote to “no” because making the player “?” with rating deviation 350 would be a better solution. The player would then be automatically banned from entering tournaments which don’t allow provisional players, but could still create/accept custom games. Their opponent could then check their history and decide whether to continue or to cancel the game.

2 Likes
34

A good point I was waiting for someone to make, and another reason I voted No.

1 Like
35

Downside of this is when someone has left OGS, you now need to wait for the game clock (typically 7 days) plus up to 30 (or 60 for subscriber) days for the game to timeout, holding up tournaments and ladders for a lot longer. This seems quite a big downside of this design to me, is this deemed acceptable?

3 Likes
36

Maybe you could handle the time-out differently for ranked and unranked games, and correspondence tournaments could use unranked games in order to get more expeditious handling of time-outs?

37

Perhaps the devs are looking for a way to except tournaments and ladders, which would explain why the fix hasn’t been rolled out yet.

As I said, I think anything is better than the status quo. However, my preference is simply to abolish the serial-timeout rule. The legitimate situations the rule was meant to address (emergencies and truly massive abandonment of games, rather than the paltry two, or three, or whatever) are quite rare. The damage done by such cases if the rule were abolished is no worse than the damage done every day by score cheats, stallers, sandbaggers, and assorted trolls. Moreover, a reasonable compromise would be to change the trigger on the rule: for example, the rule would not go into effect until, say, five serial timeouts had occurred.

Restoring the prohibition on correspondence timeouts is important to any fix, I believe. As I already noted, the rise of a culture of escaping from correspondence games has exacerbated the bad effects of the rule.

Also, as I already noted, the bug whereby a live game does not break seriality needs to be fixed.

38

Yes indeed. I refrained from voting but, yes. :+1:

39

Tournament games are exactly those ones where people take them more seriously and put in more effort, so have even more reason to be ranked than random non-tournament games.

3 Likes
40

One may well say “people take tournament games more seriously”, but you could also say “people don’t time-forfeit all their correspondence games at once if they take them seriously”. A real paradox!

1 Like
41

No paradox. You can take all your games with the most serious attitude but you can understand that’s may not be the case between a unique tournament game and a bunch of correspondence ones played at the same time. By design a tournament game is the one which should be ranked at first.

The mass timeout feature is not working as it should because it is triggered as soon as you timeout 1 game only. That’s incorrect
Another missing point is that someone coming back to play live will not get the trigger modified.

So the first thing to consider is if this can be modified and here I doubt for technical reasons, like how to modify rating calculations applied on past games.

If this is a reason why the system works from the first timeout, then it’s wrong. Like others I am to abolish the rule too.

We still have to manage the consequences of a real mass timeout. A suggestion would be to get a reset of your rating (to avoid the sandbagging, or the loss of your account for a new one). This would be a deterrent feature and still a pleasure to be able to keep your account.

If it’s something rarely happening, it could be a manual process (asking the moderation team). If it’s too common to become an annoyance then it can be automated (x consecutive timeout, then automatic reset, you get the start screen of a newcomer on OGS).