|
The problem is you only get what you measure. Then you rank only/disproportionately by those narrow measurements. It leads to "gaming" of a system. To min-maxing. (To use an old RPG term.) See Congress and Wall Street for powerful examples of how "gamed" systems end up hurting the rest of us in tangible ways. I've seen it lots on GitHub, and the problem seems to be growing. The guy that parachutes into a project -- one with money-impacting code already deployed on 1000's of real world systems used by ordinary people -- submits a pull request where he's modified every single file in the project, making only superficial changes to formatting and whitespace, and probably the result of an automated tool, and most importantly DOES NOT THOROUGHLY TEST IT beforehand... yet, from GitHub's perspective he's now (a) an accepted code contributor, (b) modified lots of files -- prolific! -- and (c) has a high count of lines added/removed -- IMPACTFUL! -- and (d) very very busy beaver "working" on lots of projects. "Gosh, we want him!" But does he/she understand threading, leaks, races, parallelization, good architecture, good documentation, automation, reproducability, security, scalability, strategy, pacing, production support, risk mitigation, algorithmic complexity, tradeoffs, market priorities, DRY, YAGNI, edge cases, BATNAs, etc etc? Well... er... maybe not so much. But his/her eyes just light right up when you mention you have a foosball table and a keg in your cubicle-or-overturned-door "office"! |