Life In 19x19

topazg wrote:

ChradH wrote:What happens if the escaper comes back while you're away and makes another move? Who does the system assume now having escaped the game?

Perhaps the idea of being able to claim a forfeit after a reasonable amount of time (a day or two?), thus clearing up one's game list, could be really helpful. I think it wouldn't hurt the ranking system and most of all it would give people back some control over the situation the escaper brought them into.

Yeah, this would work for me too, I'm more than happy waiting a couple of days to clear each game.

...But if your opponent accidentally disconnected and you aren't on at the same time for a couple of days...

There isn't an easy solution.

daal wrote:...But if your opponent accidentally disconnected and you aren't on at the same time for a couple of days...

There isn't an easy solution.

I agree, I'm also happy carrying on doing what I do - I'm not an unreasonable chap really

If my opponent accidentally disconnected, he can send me a message if he wants to carry on the game - that's what I'd do anyway. Even with the current system, I have a friend who went through 3 accounts in 2009 because of bad internet connections and all 3 got marked as escaper accounts. I'm not trying to propose a magic pill or silver bullet here, nor am I trying to say I have a better system than the one present, but there are tweaks that I think would allow more flexibility for individual players to manage escapers in different ways without fundamentally damaging ranks.

Out of interest, we talk about the impact this will have on ratings as a whole - wms, is there any chance you could do your KGS database-fu and produce the number of completed ranked games in May and the number of players marked as escapers in May, so that we can see the sort of magnitude of effect my behaviour could have "if everyone did it"? I would be genuinely surprised if it makes up even 1% of games, but it would be good to know.

The main problem with "if everybody resigned when the opponent escapes" is that it encourages escaping even more.

topazg wrote:

hyperpape wrote:Invalidating games enables escapers.

What I dislike is that a third party sees no difference between games I escaped and games that my opponent escaped. I used to dislike playing people when I saw two or three unfinished games--I thought they might be an escaper. Now I think that was a mistake, but I bet some people still act that way.

I don't see why it should. If I mark the game as invalid, it becomes "free" and ended by forfeit or no result or whatever, but still exists as a game and my opponent should still be marked as an escaper for that game. I don't see how that enables them at all?

Escapers want to avoid losses--this lets them do that for at least one game.

Maybe they'll have enough other games marked as losses that it doesn't matter. I'm having a little trouble wrapping my head around the proposal.

hyperpape wrote:

topazg wrote:

hyperpape wrote:Invalidating games enables escapers.

What I dislike is that a third party sees no difference between games I escaped and games that my opponent escaped. I used to dislike playing people when I saw two or three unfinished games--I thought they might be an escaper. Now I think that was a mistake, but I bet some people still act that way.

I don't see why it should. If I mark the game as invalid, it becomes "free" and ended by forfeit or no result or whatever, but still exists as a game and my opponent should still be marked as an escaper for that game. I don't see how that enables them at all?

Escapers want to avoid losses--this lets them do that for at least one game.

Maybe they'll have enough other games marked as losses that it doesn't matter. I'm having a little trouble wrapping my head around the proposal.

Yeah, I think in effect, it will allow this to happen up to 10 times. After which, every game will be an auto-loss as per the current system. I don't see this as much different to how it currently works, with the exception that currently once the auto-escaper has triggered, it forfeits the 10 "in holding" games too. Frankly, it's so easy to just go and create a new account I don't think it's going to make much difference to those who escape - those who want to will find a way under either system. Those who are habitual escapers due to temper or whatever will continue to do it anyway, and those who only have one or two due to genuine problems don't get caught by the system.

As mentioned, it does at least allow me not to look like an escaper in my own unfinished games.

Just to toss my thoughts into the mix...

Regarding resigning games where you were escaped on. As I see it, there are two major differences to consider here (and perhaps at least one minor one):

1: Escaping potentially excludes a valid data point from the ranking data set, whilst resigning an escaped game assuredly includes a false data point. Further as others have stated, resigning the escaped game not only includes a false data point, but it increases the likelihood that more valid data points will not be included (because it extends the duration prior to have escaped games count as forfeits).

2: The negative effects of escaping have an upper bound that is automatically enforced, while the effects of resigning games you were escaped upon are unbounded and impossible to monitor automatically. I've done the back-of-the-envelope calculations in other threads and don't feel like fleshing it out again here, but the long and short of it is, if you play a significant number of games you will move your rank by less than 1/3 of a stone via abusing the escaper system. I would imagine you would see similar gains if you only played on KGS after having a good night's rest. On the other hand, by resigning games you should have won, there is no limit to how far you could artificially depress your own rank (and affect all those who play you).

3: (This is the minor one) Escaping a game is a passive behavior (refusal to finish), whereas resigning a game is an active behavior (making an effort to resign). It is far easier to make a valid judgment on something based on what is being done rather than what is not being done.

At the end of the day, people who are actively engaging in behavior that causes bad data points to be added to the ranking system may be subject to deranking. The nature of this behavior is irrelevant, as the whole point is to simply avoid the bad data from being included in the system.


Regarding "invalidating" games:

I don't understand what you hope to achieve with this. Is it to have the game not included for ranking consideration? Because an unfinished game isn't included, so there would really be no difference. Is your concern that people will see unfinished games in your profile and think you are an escaper? If so, you could include a note in your profile explaining them (or better yet, you can use the "Tag Game" feature to put an explanation on each individual game!). If the issue is really just that you don't like seeing italicized games in your games list then (and I don't mean this sound rude, but I know tone and voice inflection carry poorly over the internet) that sounds like a personal thing you might need to get over.


Once again just my 2 cents, cum grano salis and all that...

Mef wrote:...At the end of the day, people who are actively engaging in behavior that causes bad data points to be added to the ranking system may be subject to deranking. The nature of this behavior is irrelevant, as the whole point is to simply avoid the bad data from being included in the system. ...

every resignation is potentially bad data. deranking of players that have resigned games definitely is. ..

lefuet wrote:

Mef wrote:...At the end of the day, people who are actively engaging in behavior that causes bad data points to be added to the ranking system may be subject to deranking. The nature of this behavior is irrelevant, as the whole point is to simply avoid the bad data from being included in the system. ...

every resignation is potentially bad data. deranking of players that have resigned games definitely is. ..

I agree with this completely. If I play 20 games and resign one game where my opponent escaped, I have contributed 19 games of good data and 1 game of bad data. Deranking me removes them all, so is the action a) designed to safeguard KGS' ranking system, or b) to make an example to make it not happen again? It seems to be achieving the opposite of a), and if someone is willing to resign an unpleasant but won game, they probably aren't fussed about rank enough for b) to be effective - what is gained by the deranking process exactly?

wms wrote:The problem is, topaz, it doesn't just screw up your rating. It screws up the rating of everyone you or the jerk have played.

No, the KGS system is what's screwing up everyone's rating, not the individual users. I know this has been said thousands of times before, but I'll just say it again. Being able to escape rated games is crazy. Of course there might be a five-minute or so grace period for disconnects, and free games could be excepted. But this entire problem would simply not exist, if KGS automatically assigns a forfeit when somebody escapes a rated game.

I think there at least 90 % of KGS users agree on this. Then again, KGS is an amateur hobby project so I cannot blame its author for not caring too much about the actual users

I hope I don't get de-ranked if I misread a problem "below my level"... :-p

Seriously though, IMO, resignation is always a valid option in the game of go, and it's kind of funny that people are punished for exercising it.

Ideally, IMO, the system would be like this:
1.) "Escaping" is not an option on ranked games. If someone leaves the game, and does not return within 5 to 10 minutes, they get a loss.

2.) An option is added to the user interface to "request resume". If the user selects this option, their opponent must agree to let them have it - kind of like "undo".

This way, games can be resumed if people have to go, people have a time period to return to the game in case of connection problems, and escaping is not an issue anymore.

topazg wrote:

lefuet wrote:

Mef wrote:...At the end of the day, people who are actively engaging in behavior that causes bad data points to be added to the ranking system may be subject to deranking. The nature of this behavior is irrelevant, as the whole point is to simply avoid the bad data from being included in the system. ...

every resignation is potentially bad data. deranking of players that have resigned games definitely is. ..

I agree with this completely. If I play 20 games and resign one game where my opponent escaped, I have contributed 19 games of good data and 1 game of bad data. Deranking me removes them all, so is the action a) designed to safeguard KGS' ranking system, or b) to make an example to make it not happen again? It seems to be achieving the opposite of a), and if someone is willing to resign an unpleasant but won game, they probably aren't fussed about rank enough for b) to be effective - what is gained by the deranking process exactly?

Let's compare the effect of excluding good data with including bad data -- If you play 10 games and an escaper plays 10 games, one of those is against each other.....Perfectly handicapped you both should go 5-5 (50%). The escaper escapes his game against you and we ignore it, he is now 5-4 (55% wins) and you are 4-5 (45% wins). If you resign your game, he will now be 6-4 (60%) and you will be 4-10 (40%). Also, you have now enabled him to escape one of his next two games meaning he will effectively be 7-4 (64%). It has turned a +-5% difference into a +-10-12% difference. While it may not seem like much, The effect of resigning the escaped game was in fact as large or larger than the effect of escaping the game in the first place. So it would seem strange to think that action should be taken against someone who routinely engages in one of these behaviors, but not someone who routinely engages in the other. (As stated, the aim is to ensure the integrity of the ranking system).

Further, when comparing the two behaviors, the escaper (should he or she continue this behavior) will eventually trigger the escaper system, the impact of their behavior will at worst level off (at 10 escaped games and 100 games total), or at best drop to 0 (if they forfeit more than 10 games every game is an immediate forfeit). The impact of the resigner will only grow until the games themselves expire in six months. There is no system to turn these resignations into wins (in fact the system that is in place to deal with this has effectively been disabled). The only way to reduce the extended impact of this behavior would be to derank the player.

In effect, yes, the person who is resigning escaped games will have a much larger and longer impact on the efficacy of the rating system than someone who is escaping games.

Mef wrote:

topazg wrote:I agree with this completely. If I play 20 games and resign one game where my opponent escaped, I have contributed 19 games of good data and 1 game of bad data. Deranking me removes them all, so is the action a) designed to safeguard KGS' ranking system, or b) to make an example to make it not happen again? It seems to be achieving the opposite of a), and if someone is willing to resign an unpleasant but won game, they probably aren't fussed about rank enough for b) to be effective - what is gained by the deranking process exactly?

Let's compare the effect of excluding good data with including bad data -- If you play 10 games and an escaper plays 10 games, one of those is against each other.....Perfectly handicapped you both should go 5-5 (50%). The escaper escapes his game against you and we ignore it, he is now 5-4 (55% wins) and you are 4-5 (45% wins). If you resign your game, he will now be 6-4 (60%) and you will be 4-10 (40%). Also, you have now enabled him to escape one of his next two games meaning he will effectively be 7-4 (64%). It has turned a +-5% difference into a +-10-12% difference. While it may not seem like much, The effect of resigning the escaped game was in fact as large or larger than the effect of escaping the game in the first place. So it would seem strange to think that action should be taken against someone who routinely engages in one of these behaviors, but not someone who routinely engages in the other. (As stated, the aim is to ensure the integrity of the ranking system).

Further, when comparing the two behaviors, the escaper (should he or she continue this behavior) will eventually trigger the escaper system, the impact of their behavior will at worst level off (at 10 escaped games and 100 games total), or at best drop to 0 (if they forfeit more than 10 games every game is an immediate forfeit). The impact of the resigner will only grow until the games themselves expire in six months. There is no system to turn these resignations into wins (in fact the system that is in place to deal with this has effectively been disabled). The only way to reduce the extended impact of this behavior would be to derank the player.

In effect, yes, the person who is resigning escaped games will have a much larger and longer impact on the efficacy of the rating system than someone who is escaping games.

That doesn't answer any part of my question. In reality, 50% of the games I, or anyone else plays, are not escaper games. By deranking me, you remove far more good data than bad, do you not? Doesn't removing the additional amount of good data damage the integrity of the rating system further than the 1 game in the other direction?

gaius wrote:I think there at least 90 % of KGS users agree on this. Then again, KGS is an amateur hobby project so I cannot blame its author for not caring too much about the actual users

Having a hard time figuring out what the smiley is doing there.

Mef wrote:...

In effect, yes, the person who is resigning escaped games will have a much larger and longer impact on the efficacy of the rating system than someone who is escaping games.

IMO, you are punishing the wrong person by de-ranking someone who has resigned a game to get some closure.

When two people agree to play a match, they have entered an agreement: We will play a game of go under the specified time settings. We alternate turns, and on each turn that you get, you can choose to play a move, or pass. At any time, you may choose to resign (though some may argue that pass and resign options are moves).

When an escaper leaves the game, this agreement is violated. The escaper is the one that reneged on the agreement - they left the game without fulfilling their agreement to play a game within those time settings. The person that did not escape did not violate anything at all. And, still, it is the option of either involved party to resign at any time.

They key point to take away is: Resigning is always a valid option for either player. The same is true in real life games.

Now it may be argued that resigning unnecessarily adds bad data to the system. But there is no way to fairly measure this. It's possible that a person's weakness is that they resign prematurely. I have sometimes resigned prematurely. But that is my weakness. My rank should reflect my weakness.

Actually, this is kind of cool if you want to sandbag... Call an admin, make sure he/she notices that you resign a game somebody escaped, and not only do you lose rating points for that game, you also get deranked. Double-sandbagging combo!