Proposed solution to the escapers problem...

[oren]

The reason that ratings is important with these definitions is because the LOSS is what an escaper is escaping from. To be an escaper, you must be escaping from something.

I view an escaper as someone who chooses not to hit a resign button whether they get a loss or not in order to avoid finishing a game.

You view an escaper as someone who leaves a game no matter for what reason and is not assigned a loss.

I'm curious what do you do in the situation that two players can not agree on the scoring at the end? Death match who stays online longer?

[judicata]

Just food for thought: I wonder if a timeout system (or similar) would lead to more people escaping games (or "abandoning" if you prefer). I don't know the answer, but it seems plausible. At least some would-be escapers aren't completely morally bankrupt, so they may hesistate to escape because (a) the negative effect it would have on their opponent and (b) they don't want the opponent to bug them afterwards. If escapers lose automatically, the know they will get the loss and their opponents are very unlikely to bother them later. Others just "rage quit" (borrowing a term from online shooters) so it probably wouldn't make a difference to them.

As a counter-point, I'm sure some (maybe even most?) escapers really want to avoid a loss or hurt their opponents, in which case the timeout system would be a deterrent (i.e. there would be fewer escapers). The result depends on the makeup of the player population (both current escapers and any new escapers the system would bring in), which is unknown. Unfortunately, we cannot simply compare the number or percentage of escapes under a timeout system (like Tygem) and the KGS system, because the player populations are too different.

Again, I don't know the answer. I'm also not saying whether a timeout system would be better or worse than the status quo. I just think it is an interesting human behavior/psychology study.

I've already stated my position, so I won't repeat; I'll just say that I wouldn't really complain about either system (and I still haven't had the time to respond to my colleague's counter-contract arguments, and since the point has sort of died anyway I might not).

[Bantari]

...

2) do not set the flag and only play within their group of high percent escapers (also not fair to them, I would say)
...

Why is this not fair ?

I am not sure how to explain it to you.
Are you Jasiek's relative by any chance? Just asking...

By the way, I forgot to directly address the previous comment:

Neither the KGS definition you give nor the Free Dictionary mentions ratings at all.
So - ratings is not important when defining escapers.

The reason that ratings is important with these definitions is because the LOSS is what an escaper is escaping from. To be an escaper, you must be escaping from something.

Yes, you are escaping from the loss of the game.
In a rated game, you ALSO escape from the loss of rating points.
Which does not change the fact that you escape from something in BOTH cases.
Ditto - you are called escaper when you escape.

Not sure why we have to go into these elementary definitions.

[Bantari]

...

And personally, I find it almost comical that it is against the rules to resign games before they have been decided, but it is not against the rules to simply leave a game when it has been decided. ...

You guys really just see the issue from your own personal perspective, yes?

Consider this:
A = all games properly finished
B = all games escaped - i.e. finished by fake disconnection in lost positions - or real disconnections in lost positions
C = all games finished by real disconnections in won positions

Th KGS system basically states that:
If the input is (A) the system is the most stable, exact, and reliable.
If the input is (A+B+C) the system is less stable, exact, and reliable.

Why?
Because A and B are the only 'correct' inputs - i.e. games in these categories have the same outcome regardless of connection/escaping status. Games in category C introduce bad data into the system - i.e. games which, if not for bad connections, would have a different outcome.

Example:
A 1d player with really crappy connection loses a lot of games and has actual rank of 2k if all the disconnected games are counted as losses. The domino effect then dictates that every player losing to him will have to take a rating hit equal to losing to 2k rather than losing to 1d. Same for beating this player - you only get benefit for beating 2k player. In addition, this nominal 2k player has to take 2 stones if he wants to play other 1d players - which also messes up the system, regardless of the results. Not to mention being really hard on the said player.

As a result - there are many players whose ranks are slightly off because of such games.
This is why the admins prefer you to escape (i.e. void the game) than to resign (i.e. introduce bad data into the system.)

And this is why the main problem in dealing with escapers is to differentiate between categories B and C.
Bunching them together and making them all relevant is NOT a good system.

Yes yes, I know that for YOU personally, getting a win is more satisfactory.
But this is a very narrow approach, as I already said.

Still - you LIKE your rank to be exact and reliable, right?
Well, I am not sure you can have it both ways.

[Kirby]

...
Yes, you are escaping from the loss of the game.
In a rated game, you ALSO escape from the loss of rating points.
Which does not change the fact that you escape from something in BOTH cases.
Ditto - you are called escaper when you escape.

Not sure why we have to go into these elementary definitions.

I do not think that you understand.

When you play on Tygem, when you leave the game, a timer starts. When that timer runs out of time, you lose the game.

So, back to my comment:

To be an escaper, you must be escaping from something.

To which you replied:

Yes, you are escaping from the loss of the game.

Your reply is incorrect. In the case of Tygem, you are not escaping from the loss of the game, because you lose the game when the time runs out from your absence. This is one of the "loss conditions" on the server. It is one way that the server defines the loss.

Since a loss is always awarded in the case of people that leave the game, nobody can escape from the result.

P.S. We need to go into these elementary definitions, because that is where the misunderstanding is.

[Kirby]

...
As a result - there are many players whose ranks are slightly off because of such games.
This is why the admins prefer you to escape (i.e. void the game) than to resign (i.e. introduce bad data into the system.)
....

Seems you did not listen to the proposed solution, or take Tygem's way into account. People with poor connections are provided a time limit for which they can return to the game.

[Kirby]

...

I view an escaper as someone who chooses not to hit a resign button whether they get a loss or not in order to avoid finishing a game.

Again, in the case of tygem, one of the loss conditions is not returning to a game within the allotted time provided for poor network connections. This is defined as a loss condition by the system. Therefore, if you do not return to a game within the allotted time, you lose the game.

This implies that people that leave the game without returning in time for the timer to run out, by definition, lose the game.

If the person leaving the game has lost the game, what, exactly, are they escaping from? By the definition of the system, leaving the game and letting the clock run out, hitting the resign button, and letting the normal clock run out, are ALL lose conditions.

...
I'm curious what do you do in the situation that two players can not agree on the scoring at the end? Death match who stays online longer?

This is determined by the rules of the game. Actually, in the case of tygem, admins have the ability to dictate the result of the game to prevent cheating in this phase.

[oren]

If the person leaving the game has lost the game, what, exactly, are they escaping from? By the definition of the system, leaving the game and letting the clock run out, hitting the resign button, and letting the normal clock run out, are ALL lose conditions.

I view this as escaping the game. They should have to resign. Just leaving means they have not finished the game. Therefore, it's escaping either way.

This is determined by the rules of the game. Actually, in the case of tygem, admins have the ability to dictate the result of the game to prevent cheating in this phase.

That would put quite a load on KGS admins.

[Kirby]

If the person leaving the game has lost the game, what, exactly, are they escaping from? By the definition of the system, leaving the game and letting the clock run out, hitting the resign button, and letting the normal clock run out, are ALL lose conditions.

I view this as escaping the game. They should have to resign. Just leaving means they have not finished the game. Therefore, it's escaping either way.

That's the thing, "just leaving" means that they DO finish the game. It's the way the server is designed. By server definition, "just leaving" and letting the time run out is the same as clicking the resign button. In both cases, it's a resign. The game ends - there is no game left to finish - 끝 - 終わり - bye bye - game over.

It might be that you make a distinction since you are used to the KGS system, but in systems that use this method, there is no difference.

This is determined by the rules of the game. Actually, in the case of tygem, admins have the ability to dictate the result of the game to prevent cheating in this phase.

That would put quite a load on KGS admins.

Sure. I'm not proposing such a system here (though it is nice) - it's kind of off topic.

[danielm]

Bantari, I don't think you were really responding to the quote of me which you quoted, as I don't see the direct relation, but I would like to respond anyway.

I do not disagree in general that an adjourned (or cancelled) game is better for the rating system and the players than one that is decided by forfeit due to external circumstances. If winning would be all that mattered, we should be playing tournaments.

It is not as clear-cut as you say though, unless you assume an unbiased sample. In case of escapers, the very definition of it is about bias (not resuming games which should have been a loss). This bias is artificially inflating ratings, because systematic abusers can increase their winning percentage far above what it should be. We know that this is happening because some people are fair enough to admit that they do it, so it's not just a theoretical problem. Since KGS ratings are indeed fairly inflated, especially on the lower ranks, where escaping is more rampant, it does not seem that far fetched to suspect a relation.

My point is, that systematic escaping can be as harmful to the rating system as systematic sandbagging, yet it is officially sanctioned. I have not heard a convincing argument why that is the case, other than that it would create too much strain on the admins if complaints about escapers would be allowed.

While that's fair enough, let's not pretend that the policy is flawless. And I do believe that it would be possible to find a compromise, e.g. a flagging system which would allow admins to focus on the most prominent cases of abuse as they see fit, without being swamped in complaints. Like sandbagging, it would be impossible to perfectly get rid of the problem via policing, but even small efforts could make a big difference.

And no, I'm not personally bothered by it. I haven't encountered an escaper since I started playing again recently (well one, but it was forfeited so it doesn't count), and can barely remember what it feels like. I just like to discuss matters of user interfaces and preventing abuse in multiplayer games, and I feel that neither the system itself nor the policies are perfect yet. It almost never is, so as long as at least some people perceive this as a problem, it is worth looking for solutions and improvements.

In summary, my proposals so far are:

A) Don't allow escapers to start new games when their opponent is still waiting on the board (though provisions for simultaneous games could be made). If you are connected to the server and you have an open game, you should be in it.

B) Update the dialogs and wordings to make the situation more clear. E.g. don't scare users if they want to leave a game after their opponent escaped, and at least provide a warning when the KGS app is quit. I wouldn't be surprised if the latter accounts for a fair amount of "escapes".

C) Change the policy as described above, so abusers know they may be sanctioned and players know that their concerns are being taken seriously.

I do not see how either of these changes would pose a problem for anybody currently affected by the system, and combined I believe that they would improve the situation enough for an already small(ish) issue to become a non-issue.

[Mef]

I'm still not interested in arguing any points in these threads, but just to clear up some inaccuracies...

This bias is artificially inflating ratings, because systematic abusers can increase their winning percentage far above what it should be.

This is false, as has been stated several times in other threads -- An optimum escaper who plays a significant number of games could raise a 50% "properly handicapped" winning percentage to a 55%, this difference is still far less than the winning percentage needed for a promotion (approx 66%). This assumes of course people aren't resigning unfinished games to them.

My point is, that systematic escaping can be as harmful to the rating system as systematic sandbagging, yet it is officially sanctioned.

This is false. There is a very strict upper bound on how much you can improve your winning percentage via escaping, and it is less than a stone. A game will not be mismatched due to this. There is no lower bound to how far you can reduce your rank by resigning won games, there is potential for great skew.

[danielm]

I'm still not interested in arguing any points in these threads, but just to clear up some inaccuracies...

This bias is artificially inflating ratings, because systematic abusers can increase their winning percentage far above what it should be.

This is false, as has been stated several times in other threads -- An optimum escaper who plays a significant number of games could raise a 50% "properly handicapped" winning percentage to a 55%, this difference is still far less than the winning percentage needed for a promotion (approx 66%). This assumes of course people aren't resigning unfinished games to them.

My point is, that systematic escaping can be as harmful to the rating system as systematic sandbagging, yet it is officially sanctioned.

This is false. There is a very strict upper bound on how much you can improve your winning percentage via escaping, and it is less than a stone. A game will not be mismatched due to this. There is no lower bound to how far you can reduce your rank by resigning won games, there is potential for great skew.

This is all well and good, but how can you start with the assumption that the abuser will play "a significant number of games"? As rating changes on KGS are inverse-proportional to the number of games played, and one can always create a new account and even play on many accounts alternating, I would not expect a systematic abuser to play many games on a single account.

[Kirby]

The problem with escapers is created from the desire to win. Eradicate this desire and the escapers will disappear...

Sure. But if you do that, let's also get rid of the entire rating system.

Otherwise, if you want to have a rating system, make it count losses to people that leave the game with no intention of returning.

The fact that a rating system exist, to me, gives reason to have that rating system give losses to those that leave with no intention of coming back. Make some allowances for poor connections, sure. Make some allowances for people that want to postpone the game, where both players agree, sure.

But don't let someone just leave a rated game they've agreed to without giving them a loss for it.

Removing the desire to win is fine and good, but if we are going to have a rating system, let's make it as good as we can.

[judicata]

The problem with escapers is created from the desire to win. Eradicate this desire and the escapers will disappear.

This would make go, at best, the equivalent of working on collaborative math problems (which may be fun for some people in some circumstances--but leave our ancient competitive games alone). Watching pros sit around and not try to win would be equally enthralling. It wouldn't even make much sense to have two opposing players--just have everyone play some version of Zen Go. Sure, this is fun in study groups, but "eradicating" any desire to win would rip the soul out of the game.

You can backpedal now if you want, and say you really meant "don't get too obsessed with winning" or "don't focus only on winning." I'll forget the hyperbole. (Disclaimer: my argument is intended neither to defend nor oppose any escaper system.)

[Kirby]

The problem with escapers is created from the desire to win. Eradicate this desire and the escapers will disappear...

Sure. But if you do that, let's also get rid of the entire rating system.

To me the supreme purpose of the rating system is to create games that will provide joyful challenges for both participants. You're throwing out the baby with the bathwater...

I guess I just meant to say, if you're going to have a rating system, you might as well make it as good as you can. The rating system takes wins/losses into account. An escape with no intention of returning is a loss that should be counted.

Talking about the "desire to win", etc. is beside the point.