Timeout in correspondence games

I have read the forums on correspondence timeouts and find no clear answer.
What is the rule for timeout in correspondence games?
Can a player escape by timeout?

Can I pause games that are about to time out and ask Moderator to judge me as the winner?
If I have a big lead (30, 50, 100, etc.), and the opponent has not been online for a long time (months) and has a large number of timeouts.

This is quite tricky and there is a lot of discussion on the forums. This post is good context

Why pause a game that is about to time out?

I would equate it to kicking the ball away as far as you can after a lost football game and then not shaking hands. Or hiding buildings to stall a lost game in Starcraft or Warcraft, and then not typing GG.

Itā€™s basically rage quitting for passive aggressive people.

Because if it times out as part of a mass timeout it will be annulled, whereas if itā€™s paused then a moderator can end it as a win for whichever player is appropriate and it would count in ratings.

I think the issue is that the purpose of mass timeout is to avoid distortion of the ranking system but if selected games are removed from this (all the ones where the non-timing out player would win) then this removes or limits that protection.

Thing is, itā€™s applied chronologically.

So, if my first X timeout games are against 99k they will count as losses (and distort the system) but the rest of my games, that might all be against 9d, wonā€™t count as losses, although realisticallyā€¦

But it usually balances after a few games anyway, so I donā€™t think itā€™s that much of a distortion in the long run. But I donā€™t really care about my accurate rank. Some people do.

On the other hand, Iā€™m against the sentiment ā€œAI says A player would win, so they should get the winā€. I believe this mentality is the worst thing to happen to Go in recent years. Even worse than cheating. Because it removes the reality that the game is played. Otherwise, we wouldnā€™t have to play, just submit our ranks and the system automatically grant wins against 50% of equal ranks, none against stronger, all against weaker.

If timing out is a legitimate way to lose the game, it should lose the game.

Iā€™m just stating my opinion for the void, because I know this isnā€™t going to happen. :slightly_smiling_face::woman_shrugging:t2:

1 Like

When you engage in a game of go, you agree to follow the rules that are set for that game.
At the end of the game one of the players apparently wants to change the rules.

This doenā€™t seem to be correct to me.
I hope that the mods are not honouring this request.

Just to clarify my understanding, which I think is not at all your one. The ask here is not to change the rules in any way but rather to have a win on time recognised as such with ranking points rather than not due to annulment.

Ofc I might have the wrong end of the stick. Or a totally different stick!

So you think itā€™s correct that timeout games donā€™t count towards the ranking?
The current situation is that when a player has a large number of timeouts, his games are not counted in the ranking.
And that was the case with my opponent.
I won, but it didnā€™t count because the opponent timed out, that was the problem.
I just want Moderator to give me the victory I deserve.

1 Like

You have won, the victory is yours already.

2 Likes

And if you lost, would you ask a moderator to give you the defeat you deserve?

1 Like

Well, I somehow have doubts that @LetsFightLOL will do that, but I could be wrong on this.

In this case, there is no need for Moderator to intervene, because I can resign.
Most games where Iā€™m far behind, Iā€™ll resign, so thereā€™s very little opportunity for the opponent to time out in this situation.

If it really happens, according to the rules of Go, the opponent should be sentenced to lose if timeout. (Alright, Alright, some people say that the corresponding game is not a serious game. I can only say that this is your point of view, you are not serious and some people are serious.)

OGS designed this mechanism mainly because it does not want someone to drop a large number of ranks in a short period of time, because it is obviously unreasonable for games with a large lead overtime (sandbagging, so they are cancelled).
However, a game that falls behind by a large margin should have been lost.

Yeah, the ā€˜mass timeoutā€™ thing was implemented few years before alphago, so there wasnt yet any good enough aiā€™s around to make distinction between games where one player is losing a lot and winning a lot. Most AIā€™s at that time used simple monte carlo search tree, which was infamously bad with any complicated capturing races or even long ladders and such. The strongest bots at that time were only around 1k level at best, so it was useless to make any automated systems around their positional judgement

Tho now with the super strong aiā€™s the opposite we have the opposite problem. AIā€™s are able to see potential invasions and aji that human players are often totally unable to see, so making decicions of ā€œAIā€™s perfect playā€ may lead to results that both players would disagree with.

Bit crude example of what i mean, butā€¦


Who has more points on this board? Since the both territories are sealed, white should have lot more territory since their side is lot larger than blacks, right? And indeed if both players pass now, white wins by huge margin.

But according to katago (the bot ogs currently uses for analysis), black is ahead by roughly 25-40 points depending on who plays the next stone, just because theres so much room on whites side that black can can basically throw a random stone on there and create life without any trouble.

It is not difficult to solve this problem.
You only need to set the logic (if you plan to use AI to solve the problem): if the time-out party, whether black or white moves first, is greatly behind, then it can be judged that the time-out party must lose the game.
This can also solve the problem that the winning rate will fluctuate greatly when fierce battles occur. (Because the leading party wins by a large lead under the premise of passing once. Fierce battles obviously do not allow you to do so.)
If this mechanism can be implemented, then there is no need for Moderator manual intervention.

However, what games can be regarded as rankings and what games cannot be regarded as rankings, I think this is the first problem to be solved.
Itā€™s not that many people have no problem with a large number of timeouts not counting in the ranking, but they donā€™t know it at all.
Iā€™m sure more people would take this issue seriously if timeouts always showed as cancellations instead of sneakily not counting for rankings.

Yeah indeed, but who would you say is behind on that board? Should it be judged based on aiā€™s analysis, or should based on the actual situation on the board? White has about 70 points more territory (+komi) and its whites move, if players were to pass now its whites win. But to convince AI that white is winning, white needs to add multiple moves inside their own territory to prevent potential invasion by black from working.

But oc thats just a super crude example, most timeouts happen long before the territories have finished borders like that, so judging them is lot harder than the example i made.

It depends on how the OGS solves the problem of dropping games.
(Yes, I think the corresponding game timed out while falling behind by a large margin, itā€™s a discard game, it canā€™t be a death saver.)

If OGS decides to let the Moderator judge, it is obviously based on the Moderatorā€™s judgment.
(I believe they have the right to see the AI ā€‹ā€‹analysis, do they?)

If OGS decides to let the system judge automatically, it would be a good insurance judgment to roughly assume that Pass once.
But if it really doesnā€™t solve the problemā€¦
Well, how about setting the judging ranking to be effective when the playerā€™s winning points are higher than a certain number in a row?
(Such as ten moves in a row, leading by 20 points, then the opponentā€™s timeout is obviously discarding the game.)

  1. Suppose player A times out a series of rated correspondence games. If opponents of A who had a big lead ask a moderator to get their ā€œdeservedā€ win and if some of those who were far behind donā€™t want to resign, this will distort Aā€™s rank.
  2. Itā€™s not always obvious whether a game is ā€œresignableā€. In the following games, my opponent timed out and I didnā€™t earn rating points: game 1 (20 points lead according to AI at move 150, is that big enough?), game 2 (41 points lead according to AI at move 104, thatā€™s big but Blackā€™s stone L10 has aji and White still has plenty of opportunities to blunder).
  3. Why waste energy? I guess that your rank wonā€™t improve by more than 0.1 in the long run.