Id appreciate the score to get some kind of static absolute gravity

so make them fall towards 1500 each day per 3 points. sounds fair to me.

It simply is no fun at all if the topplayers stopped playing. This is a simple but effective way to encourage playes to keep on playing. And I like it.

I disagree Overlats. When someone is able to do well enough to climb the ranks and secure themself a top position in any competition, they are given the right to a high ranking based on their play. Oherwise record-holders in sports would be useless, wouldn't they?
If it were a percentage thing (it states your winning percentage) then for sure this method would not work because someone who played and won a single game would be on top of the list, but in kdice it takes many many games to climb to the top of the list.
I say once you're there, good for you, you keep your points. Other "good" players will catch up.
Perhaps if inactivity was in the magnitude of MONTHS, maybe, but daily? That's just silly.

"record-holders" in sports only really exists in sports where there is an "objective" way to record your doings. On the team vs. team sports, you have the rule: You stop playing, you fly out of the score boards! Simply as that!

If they must fall then they should fall to about 2000, any less would simply be unfair IMO.

Why is it unfair? The rating is a relative measure of skill, and is subject to inflation and deflation. If higher ranked players stop playing points are effectively removed from the system, making it harder to actually achieve a high rank.

With decreasing points you have to keep playing to some extent to maintain a rank, keeping your points in the game, therefore opening the chances for other (possibly better) players to climb to the highest ranks.
I do agree with willy that this should be a form of gravity, to keep the system from deflating.

Currently the top players list has very little significance as most accounts in there built up a score currently almost impossible to reach, due to the top players not actually playing.
It neither can be seen as an 'all-time' high, because the scores are not absolute.

3 points per day is imo very gentle for the top players. Right now the nr. 1 is still 85 points above the highest person who played today (jtav13), which would take roughly a month to solve itself by just the decreasing scores (while jtav13 should still maintain his score by playing against people (signficantly) lower ranked, risking losses).

Personally I think the decrement should even scale with inactivity (for example, active-5 days inactive = -3, 5-10 days inactive = -5, etc.), to really force people to play.
As a compromise records could be kept though ('blah's highest rank was Xst at day/month/year, with Y points), indicating some form of record.

I like the compromise idea...

We need some input from Ryan lol

At best, a score decay will circulate a few more players through the top ranks. It still gives players no reason to play when they are in the top ranks.

They are still likely to get screwed on points if they do play. So it is still just as sensible for them to take the daily drop instead, as long as they're near the top. That month that it would take #1 to drop 100 pts down the rankings could be 10 minutes if he gets a couple of bad rolls. And for what, the chance of extending his lead over #2 by a handful of pts? (People who play purely for fun without regard for score would play just as much with or without the point decay, so that's beside the point.)

If you want to consistently get people to play when they are actually near the top of the rankings, something else will need to be tweaked.

Ok, so #1 needs incentive to play (assuming he is not just playing for fun). Because he's already #1 there is little gain in playing to extend his lead on #2 (unless he's very close).
So the only option really is to make him lose the top rank if he doesn't play (enough). Two ways to achieve that: take away points from him, or give more points to the people trying to gain on him.

The first, like you said, has little effect if it is too insignificant to the risk. So either lower the risk (lower the K factor in the elo adjustments), or increase the decay. Still #1 would have to play just enough to battle the decay. Yet, if the risks are lowered this does bring his points back into the game somewhat.

Rewarding more active players would be risky. I can only see that working in an inflating system (i.e. the points needed to be #1 continually rises). While such a system is not necessarily a bad thing, it's hard to create a reasonable balance (at what point should activity weigh more than skill?).

A different thing that might work is to have a threshold on activity for appearing in the ranking (i.e. you would need to play X games per week to appear). So even if you would have enough points to appear as #1, but haven't played enough, you would not be ranked in the top players list. So the risks of actually playing would be worth it (if you intend to keep your top rank listed). At the same time leaving for a month will not completely ruin your rank (at the current rate of decay), but just make it unlisted.