When you add up all the kills and divide by all the deaths, the thing that you are averaging is "kills" and you are calculating it on a per-death basis. (Per the analogy dividing total miles driven by total gallons consumed is averaging "miles" on a per-gallon basis). When you add up KD ratio for each player and divide by all players, the thing you are averaging is "KD ratio" and you are averaging on a per-player basis. (Like averaging MPG on a per-car basis.) When you compare an individual player's KD to the first example, there are 2 problems. (1) you are comparing a "K/D ratio" to an average of "kills" (hopefully you can understand that a kills and KD ratios are not the same metrics) and (2) you are comparing a player to per-death statistics (surely you can understand that a player and a death are not the same thing and there is no value in comparing the statistics of the two). When you compare a player's KD ratio to the second example, you are comparing a player's KD against an average KD (comparing like for like) and the comparison you are making is of one player against per-player statistics. If you calculate KD ratio for individual weapons and average those you'd get something else. You'd get per-weapon statistics. It would make sense to compare individual weapons to this average. It would *not* make sense to compare players (KD per player) to this average (KD per weapon). It would also *not* make sense to compare one player's KD ratio with one weapon (KD per weapon per player) to this value (KD per weapon).