I have added a lot of words to my queue which use the same characters in different contexts, and I get the feeling that the way Skritter schedules them somehow keeps them together. This makes it a lot easier for me to guess in the cases where I don't REALLY know the character. For example, if I have recently reviewed

xia4 you2 下游

it's easy to guess that shang4 you2 should be 上游 because the recent review of 下游 gave me a hint. If I hadn't seen 下游, I would probably never have gotten 上游 right.

Maybe the prioritization/scheduling of a word could take into account the time passed since any one of its characters was last reviewed?

James

It does this on a small scale, within the words that are already loaded into the Flash. I just realized that it probably doesn't work there as well as it could if I did it another way.

But last time we thought about doing this on the server, we decided it's infeasible because we don't have an efficient way to find all the related items in order to space them out.

We recently added another mechanism that might make it slightly more possible, but I'm not sure. I'll ask Scott.