Comment by NightSquirrel

There are 4 methods of determining developer quality: Empirical metrics which can be misleading, or gammed. Team consensus which can devolve into a popularity contest. Manager "impression" which can be manipulated with strong people skills and "impression management", or letting an AI algorithm decide who is the most productive. No evaluation system is perfect and most are very flawed which is why management is both an art and a science and those that can do it well are very rare and in very high demand and therefore often deserve the compensation they receive.