Cross-posted from LL/DRB
The traditional way of evaluating pitcher effectiveness is ERA. There are a number of glaring flaws with this statistic, the primary one being that you can describe it as follows. 'How many earned runs (unearned vs. earned determined by the subjective home scorekeeper) a team concedes per nine innings when a certain pitcher is on the mound.' If you can't spot a couple of major issues here, stop reading now.
The primary problem, of course, that ERA is much closer to a team run prevention metric than a pitching one. Even at the major league level, defensive ability varies considerably. An adjustment is made for errors, but that's in the hands of a scorekeeper, and you can't make an error if you don't get anywhere near the baseball*. This seems a rather unfair thing to penalize a pitcher for. You see the defense issue popping up all the time - the early 00s Mariners, for example, had one of the best outfields in the history of the game, and their pitchers got all the credit for it. When Mike Cameron was removed from the equation, their ERAs rather predictably collapsed.
A couple of tertiary problems follow:
- Different stadiums have different effects on pitchers and hitters. A pitcher in Arlington suffers immensely compared to a pitcher in Petco Park, even if they are roughly equivalent in talent.
- We shouldn't really care about the unearned vs. earned run distinction, because giving up an unearned run is no different to an earned one in terms of wins and losses.
- ERA tends to fluctuate wildly year to year, which implies that it is divorced from a pitcher's true abilities.
Essentially, we need to look at what a pitcher can actually control to evaluate how good he is at pitching. Seems obvious, right? How?
What we can measure about a pitcher which the defense has no control over:
- Strikeouts swinging
- Strikeouts looking
- Hit by pitch
- Ground balls
- Line drives
- Fly balls
- Home runs
By only looking at these numbers, defence is more or less entirely eliminated. There are a few robbed home runs every year, but that's neither here nor there. What do we do next? We have to figure out how many runs and outs each of these are worth. This is a long, complicated process, and I'm not going to go into the details here. Suffice to say that every line drive, strikeout, etc over an entire year is examined to see how many runs/outs result, with the average taken after the all of these events have been calculated. And then you can figure out how many runs and outs a pitcher should have given up, without considering his defence, over the course of a year. Much like ERA, tRA is calculated by Runs/Outs*27. Parks are also adjusted for (when measuring the bulleted list above). It's actually a pretty simple concept, and unlike ERA is defence independent, park independent, and much more stable year-to-year (meaning that it's a closer measure of talent - regressing tRA gets you even closer).
There are problems with it, of course - some pitchers' ground balls are hit harder than others', etc, but these issues pale in comparison to the challenges with using ERA to evaluating pitching. Just because ERA is more traditional and fits into the old-school definition of baseball stats better doesn't mean it's right. tRA, as well as other advanced statistics, can be found on StatCorner.
*The Raul Ibanez school of defence.