KissMyAxe 说:
Although, I think it's flawed because it basically assigns players their typical roles/classes with no regard to the fact that many players can and do play as two or three different classes depending on the situation and on the caller's demands (Marnid, please, forgive me if I'm wrong in that respect, and clarify if you have a different opinion). And, as they say in Balion, K

ratios are not a reliable indicator of player's value (although, Angus might disagree with me on that

).
When I made my statement that "archers have lower K/Ds, cav have highest," in whatever way I said it, I was generally referring to the high-profile players that are generally dedicated to a single class in scrims. So, I counted King John, Peasant, Peers, myself, and a couple others as archers. I imagine in season 1 that they spent the majority of their time as a ranged class. The same standards were applied to cavalry. Infantry is a bit different, because at the time it seemed to be a situational thing. If a team needed infantry, they would swap their general-purpose players to infantry. Sometimes a general-purpose player is actually dedicated to another class most of the time (like me), and because of the players involved or any other reason they might get switched back and forth. I even played cav in a few scrims for LES.

So, you're right that the stats aren't 100% accurate, but if the players I picked as archers only played as archers 50% of the time, the explanation for their lower K/D ratios & total kills is still the same (in part, I don't want to bore everyone to tears here): archers get less kills than other classes.
As for K/D being a less-than-ideal stat, that's also true. It's also true that K/D is about the only stat I could record with a high degree of accuracy from screenshots, other than attendance.

If the game tracked assists, then it would be much easier to compare the discrepancy between an archer's kills and their assists to determine whether or not they're truly a support class. As it is now, we have the NASTe 1 data and conclusions based on that, which are reliant on assumptions made based on a wealth of (sometimes conflicting?) anecdotal evidence.