Pitching is a ridiculous thing when you stop and think about it. It quite literally taxes the human arm to its biological limits, yet demands incredible precision and relentless repetition. As a result, pitchers can be quite fickle commodities and prone to injuries and irregular career paths. Some pitchers don’t figure everything out until they’re past 30 (e.g. Randy Johnson) while others achieve dominance at an early age but are unable to maintain that level (think Scott Kazmir). All these factors makes evaluating pitchers tricky, but there is plenty of exciting work being done these days in this area, especially with Pitch f/x data.
When evaluating a pitcher, rely mostly on the golden trifecta of FIP/xFIP/tERA. Batted ball data, BABIP, homerun rates, LOB%, and rate statistics are all very important too, though, and can shed light on individual players. Scouting data is invaluable as well, although unfortunately we have very little of this publicly available for use.
Links for Further Reading: