So everybody spoke, but what is next? Will something finally change or the community doesn’t have any way to suggest changes?
Personally it’s frustrating every time game is annulled because someone just decided to stop playing. How is that even something we should talk about? Correspondence game is a GAME and it has rules. It’s a deal between two players to follow those rules. If I can’t comply with my part of the deal - I lose, end of story. Yes, I can lose 20 games in a row if something happend in my life, but probably this was more important than a Go rating on some platform, if I have decided to prioritise that over my part of the deal. Also, if you are, say, 5k, than you will be 5k again very fast - just start playing and you’ll be back in no time.
I love OGS, I support it from day 1 I started playing again and today, after yet another annulled game by someone, who would lose the game, I’m seriously considering to just switch to some other platfrom despite my history here, because it makes 0 sense to be a victim of that kind of rules everytime you play a long game and someone just doesn’t care enough to respect the game, rules, and process.
Changes can also be suggested on github for instance
I guess the more the discussion and support for it one can point to, like in a thread, the more likely things get implemented. Sometimes github issues get marked as stale and close similarly if there’s not enough interest.
IIRC the main has been what happens when someone gets back after serie of corr timeouts. If their rank has dropped like a rock then their future opponents would also be affected by the unintentional “sandbagging”.
I have no idea whether the system can be tweaked so that winner would still get his/her points, but the user who times out wont get their rank decimated. Maybe the coders know better? @GreenAsJade@benjito@dexonsmith
^^ that’s the obvious suggestion, but unfortunately it messes with the ranking system if we do that … it causes everyone’s rank to drift up each time it happens.
This is “The Reason” why this remains in the “too hard” basket.
If a person wants this fixed, they’ll need to bring the maths that shows their solution will not mess with the rank pool. So far no-one has been able to suggest a way to achieve this.
The resources are available for this analysis, in the “go-ratings” GitHub repo.
Personally, I am not so concerned about rank drift until I see it. However, philosophically, I do not see why the “winner” should have their rank changed. If I get hit by a car and timeout of all my games, about half my opponents will be behind and win due to reasons outside of skill.
The whole point of rankings is to estimate skill so that we can have good matchups. It’s not a reward/punishment for good/bad behavior. In other words, “1d” should mean “1d”, not “2k, but a really good person”
I agree that a timeout in correspondence is fundamentally different.
In live/blitz, timeout usually means “this game was hard” (a reflection of skill).
In correspondence, timeout usually means “busy with other stuff” (a reflection of something other than skill).
There are exceptions to both. Some live/blitz games are abandoned (for other reasons). Likely some players (hopefully few) abuse the ratings system by abandoning a bunch of lost correspondence games en masse (perhaps some continue playing lost games hoping their opponent will have a mass annulment event). But I agree the current behaviour seems like the right general rule.
Maybe there is a way to fine-tune the general rule that would be cheap enough to implement. For example, if a player is systematically procrastinating in some games (not necessarily to abuse ratings, but just because they don’t know what to do next) while making moves in other games, then those games should (ideally) affect ratings if they eventually time out. Maybe it would be possible and cheap enough to detect this, and avoid triggering mass annulment in those cases.
But I think there’s a good chance it’d be a whole bunch of work with no real improvement in practice, since the “busy with other stuff” reason dominates.
Also, if you are, say, 5k, than you will be 5k again
Other than this, I’d agree with you completely. But it’s not just a matter of lost rating points, it’s also (and I think more importantly) lost reputation. If I get really sick and have to be hospitalized, and I have many games going, all of which I lose on time, not only will these opponents likely block me, but discerning future opponents will look at my history and decline challenges or even block.
I’m getting old and creaky enough that this may in fact happen to me at some point.
So I feel this is the main question, and not how to end the abandoned games. The games should be lost, I agree. But there should be some way to get those games marked later as “not walked away from”.
The reason this rule was introduced on old OGS (which was correspondence only, so no possibility to play live games in between correspondence timeouts), back around 2010 I guess, was there was a player luke a 5d who left the site and mass timed out down to 5k or whatever, who then returned and re-entered tournaments which seeded based on rating with his 5k rating and upset people who got crushed.
So one obvious alternate solution to this problem rather than an automatic serial correspondence timeout rule would be to let the timeouts happen and be rated, and then when he returns reset his rank to 5d. The downside is this does introduce extra rating points into the rating system so is inflationary, but if it is manual and rare might be ok. Rating systems do generally need points injected over time anyway to cope with people getting stronger (the delta value in the EGF rating system, does OGS glicko2 have similar?).
Another would be not trying to let that human use the same account, so instead he needs to make a new account (which will have lower volatilty) and play some ranked games to get back to a realistic 5d rank before being allowed to enter a tournament which seeds based on ranks. This is extra work for tournament directors though to screen out the inaccurate ranked people. Also it’s not only relevant to tournaments (though is more annoying when a 5d as a 5k wins some kyu division) but also affects the regular ranked games played by a 5k vs this 5d as a 5k who wants to play a real 5k.
It feels to me like this suggestion to change time outs is more about the sense of moral justice than a specific systemic effect on rating. Let me know if I’m wrong in this!
But if that is the case, then it might be worth considering changes that more directly addresses the sense of justice about the games, rather than specifics of ratings that would more difficult to change.
It might be a bit condescending to suggest, but maybe it would already be helpful if we changed how these games are presented, maybe not striking them through and maintaining the colour coding (unless they time out right at the beginning?). So the site would acknowledge the win but still make it evident they weren’t included in rating calculation.
Maybe even include katago evaluation in there to acknowledge games that the player was winning?
W+Timeout (W ahead, annulled) ← green
But ultimately, I’m just wondering if we could improve the situation at all by not taking away all the sense of accomplishment, especially where the player was already doing well.
I think a small improvement, which shouldn’t be controversial (and we’ve already said it before) is only to annul games of inactive players. I’ve seen a couple of cases of a player slowly timing out of correspondence games but playing live games in between but the correspondence games get annulled as part of a bulk timeout, since they look consecutive when only considering correspondence.
It is the definition of bltiz (that this is allowed to be true) but timeout in liveusually means “crap, I’m losing, I’ve had enough, I’m not considerate enough to even close the tab (and loose by disconnection), I’m just going to look at cats on this other tab instead”.
Or at least, from a moderator perspective that is what we see - a steady stream of “escape” reports that report exactly this.
It may be that there is also a steady stream of live games lost in the endgame due to “wow, this is hard”, I don’t have data on that.
You could only assert this if you have data showing that timeouts in correspondence don’t correlate with a losing position.
I think that actually we get “busy with other stuff” when we are losing
Personally I think there is absolutely no excuse for timeout in correspondence while other Go is being played.
(dayum, there’s a graph somewhere that shows opponent response time plotted against score rate … this is the left hand end of that line … but I can’t dig that graph up now rats Any meme hounds or post archeologists able to help?)
In live/blitz games, I think it matters if the winner by time-out was leading, and if they were leading, how long they had to wait to get the win.
If they had to wait for many minutes, I’d usually think that the loser abandoned the game while they should have resigned. That would be disrespectful and poor sportsmanship.
But if the loser by time-out was already short on time and timed-out while frantically looking for a move to play, the winner probably didn’t have to wait very long, especially when it was a blitz game, or when their opponent was already in byoyomi. In such cases, timing out is just part of the game.
Another would be not trying to let that human use the same account, so instead he needs to make a new account (which will have lower volatilty) and play some ranked games to get back to a realistic 5d rank before being allowed to enter a tournament which seeds based on ranks. This is extra work for tournament directors though to screen out the inaccurate ranked people.
It could be automated, relatively easily I think. I.e. software can recognize the event of a large drop in rating due to many timeouts.
I think we agree in principle. That’s still a reflection of skill, right?
I don’t have data, but anecdotally, I’ve seen opponents time out when I was winning and when I was losing. When I look, the timeouts are usually part of a large cluster on their profiles.
But if you have data/experience that contradicts this, maybe it’s just luck of my games.
My lens is probably distorted by spending more time looking at the report queue than playing live games.
However, there is one important difference between:
In this live game, I timed out while trying to decide what next move to play
and
In this live game, I time out because it was too hard and I decided not to even try to play anymore, I just left.
The difference is that the latter case is not allowed. It’s called “escaping”.
You are required to “end live game properly” - which means resigning if it is too hard and you are not going to try.
The reason why this is relevant is because it seems as if we don’t have the same expectation for correspondence games.
Personally, I think people who stop playing correspondence games when the game is too hard are rude, and if there was a vote I’d vote for deeming that to be “disallowed escaping” just like the live case.
Any meme hounds or post archeologists able to help?)