The problem here is looking at small samples and going with the mindset (especially with relievers) that results = performance. There's a ton of noise in small samples, and (again, especially with pitchers, especially with relievers), you can perform well, and get poor results. You can throw the ball well, hit your spots, and sometimes guys just hit the ball. In small samples, things like ERA, WHIP, etc, are poor indicators of "performance". You need to look at far more advanced metrics (which good GM's do) to indicate how well a guy is performing, and get an indicator of future performance, and hopefully future results.