display | more...

A method used for the evaluation of pitchers at the major league level. An idea proposed by the improbably named Voros McCracken in January, 2001, this metric focuses on the individual ability of pitchers free from the help of their defense or other extenuating circumstances. DIPS, as it is known in the sabermetric community, was designed to dispel the myth that pitchers have a measure of control over where or how hard a ball is put into play by the batter. This is quite a statement, the premise of which flies into the face of pretty much every established theory on what 'pitching' is.

McCracken's postulation relies on four primary rate (counting) stats:

$BB; BB/((IP*3)+H+BB); measures how often the pitcher gives out walks against how many guys he faces.

$SO; SO/((IP*3)+H); measures how often the pitcher struck out a batter against balls he allowed to be put into play - essentially, how difficult it is to make contact against the pitcher.

$HR; HR/((IP*3)+H-SO); measures how often batted, fair balls left the park out of the pitcher's hand.

$H; (H-HR)/((IP*3)+H-SO-HR); measures how often a ball, when hit off the pitcher and not foul or a home run, drops in for a hit. This is the most important stat for proving DIPS.

DIPS has shown that, of the numbers listed above, only three of them have any correlation to one another from a given year to the next. Of those three, $SO and $BB are the two that are most likely for a pitcher to duplicate. The fourth, the outlier, is $H. Where &SO, the strongest of the correlations, is repeated at a .792 ratio, $H is repeated at .153. This means that a pitcher who allows 100 balls hit into play (BHiP) in 2003 is shown to repeat that skill in only about 15% of cases studied. That number, about 15%, carries over from year to year and can be applied to any pitcher, from Greg Maddux to Greg Myers.

We'll use Greg Maddux as an example. In 1998, Maddux posted the 5th best $H rate in the league and had an excellent year. The following year, 1999, Maddux had the 4th highest $H rate. Think about it. If $H reflected pitching abilities, would that make any sort of sense? Does Barry Bonds ever finish in the bottom of the league for HR% or Randy Johnson for $SO? Does Rey Ordonez ever finish at the top for HR% and Kirk Reuter at the top for $SO?

In every other category, the stats are elemental and easily tracked - predictable. It is only for $H that there is this wild variance from season to season. What many people call 'veteran pitching' is actually a combination of luck and good defense and a 'skill' which should never be expected to repeat, unless a given pitcher has more luck then is really healthy.

Correlation and repeatability are the building blocks of the DIPS theory. The contribution of DIPS comes in two parts. One is that it has helped to destroy the myth of what pitching was assumed to be and, two, helped fans realize what should be used when predicting whether a player's exceptional, or fluke, performance will be repeated the following year.