So, right now the leader board is just rank-uncertainty. It'd be a little more meaningful if you used a multiplier to the uncertainty, how so?

Rank-Uncertainty means that there's an 84.1% chance that the player has a true rating higher than that number (an 1 in 6.3 chance of being wrong).
Rank-1.28*Uncertainty would mean that there's a 90% chance of a higher rating (an 1 in 10 chance of being wrong).
Rank-1.64*Uncertainty would mean that there's a 95% chance of a higher rating (an 1 in 20 chance of being wrong).

Of course since it's pretty unlikely that I'd ever hit that leader board, it's a moot point for me. Still round numbers, like 1 in 10, are nice to use.

I think if you don't play for awhile, your uncertainty should increase again (thereby knocking you off the board in most cases). It could be a weekly or bi weekly server task during off hours... but I think people should get at least 3 points of uncertainty for each week that they don't play.

And it could just be a very simple check... i.e. "have they logged in in the last seven days" ... if no, add 3 uncertainty points... etc.

uncertainty does increase the longer you go w/o playing, the rate of decline is unknown.

Hammy's been on the board for as long as I can remember but I've never seen him play. I went weeks without playing and didn't see any noticeable change in uncertainty (wasn't quite down to 35 at that point). So if this is in place, I think it needs a tweak.

This would suggest that the rating period needs to be shortened or c (the constant associated with the uncertainty increase over time) needs to be adjusted.