feature request:
1. a "suspicious game flag" feature for games of suspicious activity. flags (suspicious play tags) created by lichess and not elected by players will be created whereby games with players who disconnect at least once per game and go on to win game (particularly from a losing position) will be flagged. These flags will be private and not public stats.
2. players who get flagged by lichess and are in the top percentage of flags received will be investigated for cheating.
3. more than one flag per game should receive more attention particularly if this player has been flagged before in other games.
Background.
some players are playing a game and then leaving for 60 seconds (sometimes more than once) and suddenly playing remarkably well to turn around the game entirely. I have noticed this trend again and again where the players appear to go off line (black icon) and then reappear and instantly play a tremendous move on their return. sometimes it is an innocent disconnect, and sometimes it is honest play but i believe the system will help prioritize analyzing at least one type of cheating exploit and make lichess a better game.
@baabaa8 said in #1:
> 2. players who get flagged by lichess and are in the top percentage of flags received will be investigated for cheating.
Lol
@baabaa8 said in #1:
> players who disconnect at least once per game and go on to win game
Nobody needs to disconnect to cheat, that's just factually nonsense.
>players appear to go off line (black icon) and then reappear and instantly play a tremendous move on their return.
I've experienced this too, it's called taking you seriously now. Especially if you get rattled it's easy to go on to lose and I consider this behavior annoying, but nothing more.
@baabaa8 said in #1:
> some players are playing a game and then leaving for 60 seconds (sometimes more than once) and suddenly playing remarkably well to turn around the game entirely.
I have heard this often. And then I looked at the game, and more often than not, this was a misconception. Once you have that mindset, you really believe in that, but it's not happening.
Disconnects happen for many reasons all the time (bad connection, navigating to players' profiles, changing devices, or simply refreshing the pager). And while you technically could disconnect and cheat, as @lonelypeanut has pointed out, there is absolutely no need to do so.
Your "algorithm" will not help in fighting cheating. It will only fuel paranoia.
Simply put this is data collection and could contribute to improving cheating algorithms. Disconnects per game is no different than time per move.
i have noticed that if you click outside the browser the light turns off and thus some people may be tempted to get external analysis in a position where they have the time to do so. Streamers have also noticed this type of cheating.
The suspicious flag just makes collective observations. labeling suspicious activity is nothing more than looking for long term verifiable patterns which could be analyzed with statistics at a later date.
There is nothing to prove bias (misconception or paranoia) and data collection is independent of my mindset but a process. Try not to make this personal but instead analyze the system. An attempt to identify known exploits is logic.
Data is unbiased and a suspicious flag would not hurt the game but make lichess algorithms smarter.
I think you did not read or understand the answers given.
@baabaa8 said in #5:
> Streamers have also noticed this type of cheating.
Just because someone says it on stream does not make it true.
> An attempt to identify known exploits is logic.
It is not a known exploit. It is rather a known source of "false-positive" and mis-accusations.
It's like labeling someone a potential burglar because they bought a screwdriver. Yes, some burglars bought a screwdriver, yet it is neither necessary nor indicative at all.
says you.
data is data.....
log repeated disconnects.
@baabaa8 said in #7:
> says you.
And many others, both in this thread and in the past, including people who seem to be very versed in the topic of detecting cheaters.
Your "idea" is not new at all. It frequently comes up quite frequently.
> data is data.....
As told many times now, it still does not indicate cheating, no matter how much of it you collect. It does not indicate that someone cheats, nor does it make him suspicious. Nor does a lack of disconnection indicate the absence of cheating.
It is simply a worthless metric in that regard. Just reiterating "log disconnects" will not change that. There are better ways to kill time than waste it by implementing stuff based on someones' gut feeling. Things that have been discussed many times before.
Not sure why you think this trivial (albeit non-working) approach would be something that has not been considered in the past.
OMG , looks like the whole chess world is goiing paranoia over cheating . come on guys, just because you 'feel' someone is cheating, dosen't means they are cheating
OMG = feeling
data = data
This topic is now closed.