I'd like to have AniDB derive the episode lengths from the new "file length" entry you added today.
I'd love to see how long an episode is everywhere and even I was to lazy to add all those episode lengths back last year

If the file is more than 10% longer than a certain percentage of all files then the file most likely is a double episode and that could be another indicator for the future or whatever

Build yourself a nice algorithm... maybe if 80% of all the files are withing a 1 or 2 minute range caluclate the average of those 80% of files and use that as the episode length...
Hmm... what would be a nice algorithm? Take file length of ep 1 and substract the length of file 2. Put in array[0]. Substract length 3 from length 2 and put in array[1] ... upto n files. once done go through array and take every value from -1 ot 1 (minutes) and add them all up to $lengths while $i++ for every file within -1 to 1. $eplength = $lengths devided by $i
Or something like that.
1st run: do it for all episodes and put all EP IDs with lengths +/- more than 2 minutes into a file to check later (sorta creqs).
daily run: only for episodes that have a length of 0 or over 150 (which is close to impossible, no movie is that long.
Just a thought
