The Margin of Error

Player A has hit for a .302 AVG and Player B has hit for a .300 AVG.  Player A has been "slightly better" than Player B in this category.  Given 500 AB, the difference between hitting .300 and .302 is one hit.  The difference of one scorching line drive hit.  Or the difference of a blown fair call down the line.  Or the difference of a blown safe call on an infield hit.  Or the difference of someone getting charged with a tough error.  Umpires and official scorers aren’t perfect and neither are the stats the players put up.

This same principal also applies to sabermetrics.  One common misconception is that sabermetrics attempt to be more accurate and precise than the traditional stats.  That isn’t true at all.  All sabermetrics does is take a deeper look at the stats; looking at things that the players can control themselves.  For example, traditional stats say that Austin Jackson has 117 hits in 373 AB this year.  Sabermetrics say Austin Jackson has 61 line drives, 97 fly balls and 118 ground balls (FanGraphs).  The batted ball data isn’t more accurate, it’s just looking at the AB in a different manner.  (Not to mention that two different sources can disagree on the batted ball data itself!).

For a pitcher, a timing issue can result in skewed stats.  Take Doug Fister’s game last night.  Fister gave up 4 runs, but due to a couple of errors behind him, none of them were earned runs despite giving up a HR.   Had there have been no errors in the game, the HR would’ve resulted in at least 1 ER.  Unearned runs are recorded to prevent pitchers from getting hurt because of poor defense behind them.  But when those runs are scored because of a HR, that kind of defeats the purpose since HR are out of the field of play and the defense has little to no control over them.

It’s getting to that time of year when people look at players stats and determine who has had the “better” year.  The next time someone says that Player A is better than Player B because of a couple of percentage points, just remember the margin of error.  It’s closer than you might think.