@Mephostophilis: in the beginning points were given back when someone was adjusted...and actually it changed because of me...
Imagine you have to adjust controlaltdelete (7000 games here), it will cause so huge changes to the ratings that elo won't mean anything.
Actually, after ~20-30 games, the points you loss during the game are compensated by the fact that you win more points when you win, and you lose less points when you lose.
Well that's the thing. I didn't discuss this explicitly when I mentioned the idea of giving points back to those who lost against cheaters, because I didn't want to lengthen my post drastically.
But still, I didn't think about it in the way papa explained. I don't entirely understand all of this, but I do know that point distribution is dynamic and based on current win/lost/draw and rating.
So the only realistic way to implement what I had in mind, was to make all point changes after every game dynamic, detecting the validity (cheating or not by opponent) of the previous games. This
way, the way points would be "given back" is a matter of recalculating all rating changes to be re-executed as if the cheater games were never played. The database is much too large to apply that
sort of measure to do on a per-game basis, but that goes without saying.
Therefore the only easy way to solve the unfairness of the points people do lose by playing cheaters is the natural way -- to prevent cheating. I do realize anything can be hacked :P actually I wrote
some cheat codes for virtual machines (emulated systems) a few months/years back, too. I still don't see the harm in making sure people know that switching windows during every single turn may signal
cheating. I never said anything about taking action, because if they know that the site checks for it, then they know not to give themselves any reason to switch windows during every single turn of a
rating game, thus already making completely cheated games (use computer for every move) impossible for them. I don't consider counting the game as "Casual" instead of "Rated" a matter of taking
action against people who knowingly switch windows every turn when they don't need to; I consider it a matter of not taking action--i.e. the action of giving points, because in rated chess people pay
more attention to the board than that.
It won't change anything: if you lose a game, then play 30 games, you will obtain the same elo than those you would have obtained by winning the first game and doing the same next 30 games
That's not what I meant. Rather than counting the first/cheated game as a win, don't count it at all. Recalculate all scoring for all games after the cheater game(s) as though they had never been
played, since the algorithm reads in previous successes and failures.
I know for the sake of the person wanting the points that it does not make a difference, whether the end result gives them more or less, but for the accuracy of the rating then it would make a
difference.
Besides, as I addressed it would be too complex on the database to go through those games just to re-execute finding the rating value. More simple would be to prevent cheating through stable means.
What papa is saying is that the ratings will balance out in time eventually anyway. Say I'm rated 1500 and I lose 4 games to a cheater, then I'm 1400. At that point I'm under-rated, everything else
being equal, and the games I win in future I will gain more points than I would have at 1500, and I will lose less when I lose. Therefore there is a sort of balance that maintains itself, despite
temporary local distortions.
The same is true in real life. A healthy society will do its best to deter/prevent crime, but it's still going to happen, and the effects cannot be entirely erased. A healthy society accounts for
it, tends towards the good even so, and moves on.
I do know the balance is there.
That's why every 2- / 3-day cycle my rating cycles between 1250 and 1500 XD. It's easy to gradually get up, then get down due to the higher vulnerability of losing more points if a game is lost.
This compromise that was described.
As for it being more healthy to accept fundamental issues that persist just about anywhere else--including real life--can't say I disagree. It's natural; I mean people learn more when they encounter
cheaters. There is the benefit of experience from even sorrow. Conversely, successfully preventing such issues would maintain a more stable community, except that they'd have to remember not to
switch windows all the time or whatever is in use to prevent the cheating.
I'm really not worried about the points at all. I'm confident that the lower my rating gets the more I'll just climax it later anyway. It's a matter of perspective, I guess, but I would experiment
with normalizing ratings whilst excluding the cheating games. It would be more consistent to decisively erase the history of cheating if it is disagreed with and disallowed, or to allow engine use
and not erase rated game history. But, inconsistency is healthy, too.
I agree our focus should be on prevention, when possible. I'd like to see IP ban if we can determine other innocent players won't be affected, but that may be too hard to do.
Refunding points the way you suggest would be more reasonable, it just seems way too complicated to actually do. Say I play a cheater then I play 10 other people; then say they play 10 other people,
and they play 10 other people; then let's say I play the first 10 people again. Would have to be an incredibly flexible algorithm, and we're talking about potentially amending thousands of ratings
each time we reset one cheater. And the net effect will be negligible in the long run.
As for window switching, there are times when it will pull up false-positives; I think leaving it as an advisory tool is best. If we see somebody window-switching, we can check if it is consistent
thru all or most of their games, look at move times, quality of play, and other factors; that will provide the necessary conviction to reset.
>Because in order to cheat in rated games, you would use a computer chess engine to calculate the moves for you (unless you see some other way?), which would require switching to the window for that
corresponding application.
never heard of browser add-ons?
Never heard of proper site design that can detect browsers and browser add-ons?
Is that seriously the reason to think that my claim of "almost completely preventing cheating" (note where I said "almost", because, again, anything can still be hacked) is so senseless? Because of
that one minor possibility?
Those engines suck by comparison, anyway, to the ones people install separately from attaching as a browser add-on. They're hardly even worth cheating with; I came across an Anonymous person who
tried to use one on me. :P
They can also use another computer side by side or something; in which case a window switching alg wouldn't pick it up.