inter-rater reliability question

#1
I'm trying to help a colleague and haven't done this kind of analyses in ages. Can't remember what might be best to use for inter-rater reliability. We have 3 raters (more later) examining videotaped session - each session will be rated now by all 3 raters, later by any set of two. I know regular correlation is the wrong way to go here - need either Kappa or Inter-class correlation (ICC). I seem to remember that one is more stringent than the other, but can't remember details or which might be better to use in this case.

Thanks for the help.