Dan Drezner takes on Exum's anti-quant and essentially anti-social science rant on counter-insurgency knowledge. Indeed, as there may be problems of selection bias in cases used to understand insurgency and develop COIN strategies, doing quantitative work is precisely the kind of effort required to determine whether what we know from a few key stories applies more generally.
The whole idea that a guy in an infantry platoon has far more to say than trained and educated social scientists smacks of the old claims that those who study baseball (sabremetricians) with statistics have far less to contribute than someone who played first base for a while. They each bring different things to the table, but one is far more likely to see the forest while the other is far more likely to see the bark on one tree.
We certainly know more about civil war and insurgency than we did twenty years ago, both through the reports back from the field and via the systematic study by political scientists (and economists and sociologists and other social scientists). Kalvyas, Weinstein, Humphreys, Salehyan, Gates, Davenport, Sambanis, Ross, DeSoysa, Toft, Fortna, Fearon and Laitin, Mason, Moore, Gleditsch, Regan, and others have added much to our thinking over the past several years, even if sometimes they lead us in the wrong direction (like F&L causing everyone to use ethnic fractionalization indexes as their measure of ethnic something or other).
If I didn't have to prep for some long flights, I would enumerate some of things we know now. Perhaps my readers can help out.