If "leaving a game without resigning can result in a temporary ban"...

@IbnHussein said in #8:

Thanks for the data points. Around how many times does it take for the first ban to kick in do you estimate?
Idk, i did not do it in a row, but really a lot. Like, i aborted ~15 games of my last 30 and then got ban. Though, next bans were more easy to get.

@IbnHussein said in #1:

...then why the heck do people do it all the damn time? It seems like every day I play, at least a few players leave when their position is lost rather than resign. Is the punishment too lenient? Maybe make the bans longer / kick in sooner?

Here is a view from the other side as someone who has left games without resigning.

4 times now when my internet connection has gone down and I get the warning about leaving the game. So not everyone is trying to offend you, there are cases where the other user has little to no control over the disconnection.

I can't say how long the punishment is supposed to be. By the time I reconnect, nornally 3-4 minutes later, resetting routers etc, the game is over and the opponent has claimed the win. I curse a bit and start a new game. I would hate to think I would receive a long ban because of my internet provider.

@MaxTheLad said in #12:

Here is a view from the other side as someone who has left games without resigning.

4 times now when my internet connection has gone down and I get the warning about leaving the game. So not everyone is trying to offend you, there are cases where the other user has little to no control over the disconnection.

I can't say how long the punishment is supposed to be. By the time I reconnect, nornally 3-4 minutes later, resetting routers etc, the game is over and the opponent has claimed the win. I curse a bit and start a new game. I would hate to think I would receive a long ban because of my internet provider.

It happens to me too from time to time, but from what I've heard now from others the ban-threshold is still way too high. There's a middle ground I think, to greater discourage leavers while also not having sporadic disconnects get punished.

One way is for the punishing algorithm to incorporate whether the leaver was winning/losing. If it's just due to the player's bad connection, half of their disconnects should be while they're winning (or at least the ratio would be consistent with the games they don't leave). But if every time you disconnect you were losing, then that person is almost certainly doing it on purpose.

@IbnHussein said in #13:

It happens to me too from time to time, but from what I've heard now from others the ban-threshold is still way too high. There's a middle ground I think, to greater discourage leavers while also not having sporadic disconnects get punished.

One way is for the punishing algorithm to incorporate whether the leaver was winning/losing. If it's just due to the player's bad connection, half of their disconnects should be while they're winning (or at least the ratio would be consistent with the games they don't leave). But if every time you disconnect you were losing, then that person is almost certainly doing it on purpose.
The problem is determining who was winning is harder

@Autofill said in #14:

The problem is determining who was winning is harder

Uh, StockFish? This is just a heuristic anyway, doesn't need to be exact.

@IbnHussein said in #15:

Uh, StockFish? This is just a heuristic anyway, doesn't need to be exact.

Ha ha ha. Impose Stockfish's computational burden on the disconnector, when they reconnect.

@IbnHussein said in #15:

Uh, StockFish? This is just a heuristic anyway, doesn't need to be exact.
Ok lets see
In a 1200 rated game white is down 10 points in material but has a forced mate in 12.
Stockfish would judge this inaccurately

@Autofill said in #17:

Ok lets see
In a 1200 rated game white is down 10 points in material but has a forced mate in 12.
Stockfish would judge this inaccurately

This is irrelevant. We just need some metric that's uncorrelated with games that are disconnected accidentally, and correlated with ones disconnected on purpose. The former kind will be indifferent to the state of the game, since internet loss is random/uncontrolled by the player, so any metric will satisfy this. For the second kind, people leave when they're losing, so we just need a metric that correlates to loss, it doesn't need to be precise or worry about edge cases, since we're just dealing with the statistical aggregate.

If you did this, you'd have a very simple way of judging, from a set of games that a player has left, whether they did so on purpose or not. I can't even be sure lichess isn't already doing this, but I doubt it.

@IbnHussein said in #18:

This is irrelevant. We just need some metric that's uncorrelated with games that are disconnected accidentally, and correlated with ones disconnected on purpose. The former kind will be indifferent to the state of the game, since internet loss is random/uncontrolled by the player, so any metric will satisfy this. For the second kind, people leave when they're losing, so we just need a metric that correlates to loss, it doesn't need to be precise or worry about edge cases, since we're just dealing with the statistical aggregate.

If you did this, you'd have a very simple way of judging, from a set of games that a player has left, whether they did so on purpose or not. I can't even be sure lichess isn't already doing this, but I doubt it.
Ok so how would you define losing?

@Autofill said in #19:

Ok so how would you define losing?

My point is that you don't need to, you just need to measure something highly correlated with winning/losing. Like StockFish's centipawn advantage.