On Mon, 21 May 2007 16:41:27 GMT, Ron
wrote:

In article ,
Mike Murray wrote:

There's no fool-proof way to identify someone cheating, but when the
lag appears toward the middle or end of a close game where your
opponent is behind on time, one suspects.....strongly.

It might be reason for suspicion, but perhaps not strong suspicion.

I mean, in half the games where lag appears, through nobody's fault,
you'll be ahead on time - completely randomly. In half the games where
lag appears, through nobody's fault, randomly, you'll be winning on the
board.

My impression is that lag doesn't tend to appear as often when one
side or the other is clearly winning on the board. There's little
point in cheating when one is winning without it, or when it's too
late to help.

Put those two together and 75% of the time, when lag appears midway
through a game, you'll have one of the two conditions which most people
would use as evidence of cheating.

I mentioned "confirmation bias" in another thread, and this is another
instance where you need to be careful of it. When you're ahead on the
clock, or have a winning position, you tend to REALLY notice when all of
a sudden your opponent's connection gets laggy.

If the position is even, or your opponent is better, or your opponent
has plenty of time ... you tend not to notice at all.

Your point about confirmation bias is well taken, but since I've
started to complain about lag cheating, I've tried to be alert for
this, and to notice lag *every* time it appears. As far as I can
tell, it tends *not* to appear as often when one side or the other is
overwhelmingly winning on the board. It tends *not* to appear as
often when my opponent is significantly ahead on time. It tends *not*
to appear as often at time controls of 3-2 or 5-2, or other TCs with a
reasonably generous blitz increment..

Of course, one tends to get distracted when playing blitz and bullet
and I'm not trying to sell this as a scientific experiment. It
wouldn't be that hard to play a couple hundred games with an observer
and try to gather some betters stats, however.