Best Way To Evaluate A Pitcher

Interesting question and good discussion. There is no single silver bullet statistic but there are several useful ones. ERA+ is a great measure because it normalizes park and league effects. In Pedro's case it clearly shows how amazing he was at a time when pitching was at an all-time high. His 1.74 in 2000 is much more impressive than Gibson's 1.12 in 1968 (not to take anything away with Gibson, certainly - it seems like every time I say that I get accused of being disrespectful to Gibson, as if being worse than Pedro in 2000 is worthy of scorn.)

For looking at an entire career, I do prefer WAR measures to ERA+ generally. WAR is cumulative, so players who decline slowly. continue providing post-peak value will often have better rate stats than those who burn out quickly. So an extreme example: a player who plays for 10 years and posts an ERA+ of 130 will have a better ERA+ than one who pitched for 20 years and had an ERA+ of 135 for the first decade and 110 for the second, even though the second player pretty obviously was both superior at his peak and provided significantly more career value.

For a retrospective look at career value, bWAR is superior to fWAR. FIP has more predictive value than runs allowed for smallish samples like a single season, because in most cases runs allowed will regress pretty close to FIP. For players whose RA does not regress accordingly over very large samples like a 4000-inning career, that usually indicates some skill that FIP isn't able to pick up - an aptitude for situational pitching, exceptional pitcher defense, ability to hold runners, etc. So players like Tom Glavine or Mark Buehrle will get underrated in fWAR because they outperformed their FIP consistently over a career.

So that we get a feeling for what we're talking about, the difference in the two flavors for Wins Above Replacement (click to expand):