Originally Posted by maradine
Not quite - KPU is performance divided by popularity.
Consider a 24 hour global pull across all servers from two fictional MAX arms.
The ZN-99 BoatMurderer scores 10,000 kills from 5,000 individual people using it.
The AX-Y-U-MAD also scores 10,000 kills, but from only 2,500 individual people using it.
The ZN has a KPU of 2.0. The AX has a KPU of 4.0. Clearly, something is very different about these two weapons, or their wielders. The wielders are averaged out by the Law of Large Numbers - while it is technically possible, it is very unlikely that the ZN wielders are, on average, two times as skillful and productive as the AX wielders. Therefore, there is a likely incongruity between the effectiveness of the ZN and the AX.
People shouldn't get upset over small differences in KPU, especially this early in the Oracle's data set. Probably worthwhile pointing out here that I consider <10% "small". Large differences, however, point to potential imbalances. This isn't necessarily a bad thing - I like factional asymmetric balance, personally. The point of this project is to identify what those are, and have a discussion around how they effect the game.
|
The problem with your analysis is that there's nothing that proves that the 2 hypothetical weapons were used just as long. For example, the ZN could be a very good weapon stat-wise but the devs botched the model and textures. Players got 2 quick kills on average with it and switched to something else. The AX on the other hand could be an UP weapon but with fantastic effects, textures and model. Players prefer to us it even if its stats suck. It could have an individual player 2, 3, 4 times as much time to equalize the amount of kills he could have got with the ZN.
An example you provided in your stats page to prove my point is the low KPU for lock-ons on high BR players compared to low BRs. I suspect that high BR players prefer to use a RL with a higher skill ceiling. Lock-on weapons are popular among low BRs because they're easy to use, so low BRs have a higher KPU with them. If you'd search the data for the average kills per hour used for each BR-tier, the data would probably prove my point by showing more constant data across skill tiers.