Hi, I recently replaced a bunch of repetitive tables with new templates on enWikipedia, resulting in my avg edit size dropping from 560+bytes to 469bytes, which is kind of misleading because edits made in either positive or negative direction should still count as edit size. For example, a person making two edits, first adding 20 bytes then removing 18 bytes, will see their avg edit size =(20-18)/2=1byte, when in reality it should be =(20+18)/2=19 bytes, as edits in either direction counts as size (the magnitude, not direction). Thanks!
Topic on Talk:XTools
I think that could arguably be considered a different metric, such as "net average size". The average as it is calculated now gives insight into editing behaviour. For instance, recent changes patrollers will usually have a negative average size since they make lots of reverts. I can look into adding what you suggest as a separate metric, assuming it won't slow things down much (I don't think it will as it's limited to the past 5,000 edits anyway).