grenv: I disagree... the fact he has 200 games now is vitally important. In this case he will be twice as slow because he has twice as many games... This is meant to be something indicating *current* ability to play quickly.

Previously you proposed the indicator, let's say it S = (moves per day) / (number of games currently running) for measuring players speed ability to play games.
So what you say now about "the fact he has 200 games now is vitally important. In this case he will be twice as slow because he has twice as many games" does not make sense at all.

Reality is that if S is a good indicator, then it has to be kept almost steady, so if the player has double games to play than before, then the indicator S, estimates that the player will probably double his "moves per day" rate.

If you disagree that the S indicator can estimate that, then you don't actually need such an indicator as it's useless to predict anything, like the tendency of "moves per day"(and yes this is an average obviously :-) ) when the number of games change, which is what you want.

+++++++++++++++++++++++++++++++++

Also about your:
"What happened to the idea of rating people... moves per day / # of games currently running.

If this number was >1 you would expect a move per day, but a number of 0.1 would be a move every 10 days and you could avoid those players."

This is not true at all. That would be true if the indicator was "moves per day" and not "moves per day per number of running games".

+++++++++++++++++++++++++++++++++

The best indicator of someone's pace to play moves per day, is obviously the move per day indicator so what is best to do is to see::
•A Y-X graph of Y = moves/day and X = days (for example for the last 365 days).
•The standard deviation of the number of moves/ day, over the average moves/day for the last 365 days.
•The arithmetic mean(average) value of moves per day, over the last 365 days for example.

The number 365 of days can be changed to the most desired, for example to be done for the last 6 months.

If the graph is "rough" with Everest types of maximums and many zero valleys and the standard deviation large, despite the value of the average moves/day, then the player is not very trustworthy to count that he will play quick.

The above 3 indicators, make the number of games a player has irrelevant, since a small standard deviation with a large average (that means a not "rough" graph), means the player's history says that this player plays fast, always fast, no matter the number of games.

Just my 0.02c.