Has this finding been published in a decent journal? I couldn't find it anywhere. The best fit lines on those graphs look a little dodgy. The error at the each end of the line is huge (Just look at the Iraq one, there isn't a single result on the lower side of the line, all occur above).
The data sources used are unreliable. The media for being incomplete and the NGOs for over reporting.
The alpha value for the various countries differs a lot. Combine this with the fact that I think he uses a non-linear scale to attain alpha and I think we have another very important error going on here.
Finally, this trend is pretty obvious. Let's reword it to, "there will, in a war, be more attacks where few people are killed than attacks where many people are killed".
It's an interesting video, but I don't think we can read anything into it at this point.