In Tuesday’s piece, I wrote about why batting average and RBIs aren’t accurate indicators of a batter’s performance (read it here).
Today we talk about Earned Run Average.
First, let’s get this out of the way: wins and losses are awful stats. A pitcher has control only of preventing runs, and he only has limited control over that (see below). Wins happen as a result of the pitcing and a result of the offense the team produces. They’re entirely lopsided in favor of teams that have better offenses. As it turns out, there’s no pitcher that pitches to keep his team in the game. If he could do that, why wouldn’t he give up fewer doubles and home runs? Fewer runs overall?
OK, now to ERA.
Sparing you the details laid out in the previous post, ERA was also created by Henry Chadwick in the late 19th century. It’s components are earned runs and the number of innings a pitcher pitches. (ER/IP) x 9. That’s the simple formula.
Chadwick didn’t have years and years of data to pull from and as it turns out, now that we do, earned run average doesn’t correlate so well with pitching success. It turns out its a better stat than wins and RBIs, but around the quality of batting average.
But it’s not good enough. That’s for two major reasons. The first is it only counts earned runs. Hypothetical situation: let’s say your team has a phenomenal defensive short stop and a great groundball pitcher. They work very well in tandem. With two outs, a hitter hits a ground ball right between 3B and SS. The shortstop, because he has superior range to the average shortstop, gets to it, but he bobbles the ball on transfer and a stadium official gives him an error. The next batter hits a home run and neither one is earned.
If the shortstop were, ironically enough, a worse defender, the ball would’ve sailed through the hole and both runs would be earned. In the end, the shortstop’s defense looks bad and the pitcher’s ERA is unaffected despite giving up what should have been a base hit. But because of his superior range, he makes 50 more outs on the season than a fielder with worse range and fewer errors.
And that leads us to problem no. 2 of ERA, which is it doesn’t consider the fielders’ range. A guy with a great defense behind him, like say Felix Hernandez behind the M’s D last year, will get more outs per pitches thrown than someone like Zack Greinke or Clayton Kershaw, who play in front of the two worst defensive squads in the majors right now.
In addition to giving up more hits, that means Greinke and Kershaw have to throw more pitches overall and tire quicker. It skews the number of innings they pitch, which skews their ERA.
Long story short, attempts to create a better pitcher evaluation are getting somewhere, but aren’t as close to where they are on offense. That’s partially because offense is independent of most things. The hitter is responsible for getting on base and scoring runs. A pitcher is reliable on his defense.
Pitching stats that tried to evolve past ERA started first with outs and runs scored that are independent of fielders–this was called Fielding Independent Pitching, or FIP. FIP measures solely on home runs, walks, hit by pitches and strikeouts. It’s formula: (HR*13+(BB+HBP-IBB)*3-K*2)/IP.
xFIP came after that and it normalizes home run rates based on flyball rates. Home runs are a factor of allowing flyballs; the more flyballs a pitcher allows, the more home runs he’ll give up. Traditionally, 11% of flyballs are home runs and that’s mostly a constant; xFIP is still in its experimental stages.
Right now we have groundball, flyball and line drive rates. They’re available on player pages at Fangraphs.com. Basically a pitcher can help himself in three ways: raise K-rates, lower walk-rates, get more groundballs. Groundball pitchers succeed in the majors. Tommy John and Derek Lowe have made incredible careers off of it and Joel Piniero is doing a bang-up job as well. Some examples:
2008 Cliff Lee: 2.54 ERA/2.83 FIP/3.57 xFIP
2009 Joel Piniero: 3.49 ERA/3.27 FIP/3.68 xFIP
2010 Ubaldo Jimenez: 0.93 ERA/2.42 FIP/3.33 xFIP
2008 Cliff Lee: 21.1 LD% / 46.5 GB% / 32.4 FB% / 6.85 K/9 / 1.37 BB/9
2009 Joel Piniero: 15.7 LD% / 60.5 GB% / 23.8 FB% / 4.42 K/9 / 1.14 BB/9
2010 Ubaldo Jimenez: 16.1 LD%/55.4 GB%/28.6 FB% / 9.12 K/9 / 3.72 BB/9
So 2008 Cliff Lee is a consummate control pitcher. There’s more to what makes Cliff Lee, but he and Piniero had such great 2008/2009 seasons because of their very, very low walk rates. Lee also snapped off almost seven strikeouts per nine innings, which gave him an awesome K/BB. His K-rate balanced out his average groundball percentage and high line drive percentage, which is what Piniero had going for him, and that gave them about equal xFIP.
Jimenez, though, is just amazing. Low line drive rate, high groundball rate, high K-rate. Though his one flaw is that walk-rate, which is why his xFIP is just below Lee and Piniero.
tRA is a different method, as Andrew Martin at Purple Row explains.
tRA is based on the concepts that wOBA is based on. Each offensive event is worth a certain amount of runs, and each defensive event is worth a certain numbers of outs. Tom Tango gave us this data by poring through years of game data and determined how many outs and how many runs occurred after each event. He also formulated how many runs each base/out situation was worth. Using that matrix there will give you a good idea on what run expectancies you should anticipate based on how many are on and how many outs are remaining.
For example, a strikeout is worth 1 out. A home run is worth 0 outs. But how many outs is, say, a fly ball worth? Based on 2008 data, a fly ball (to the outfield) resulted in an out 83% of the time, so we can infer that an outfield fly ball is worth 0.83 outs. However, just because an OFB is worth 0.83 outs does not mean that it’s worth 0.17 runs. An OFB, in fact, is worth only 0.046 runs.
So each possible offensive/defensive outcome is accounted for as league average. Because the sample from which Tango took from was so large, tRA has removed ballparks, defense and luck.
That creates a problem though because park and defense do have a real effect. Put a San Diego left-hander benefiting from Petco’s home run power outage in Boston and you’ll see his doubles spike greatly. That’s why tRA is best used to evaluate pitchers.
SIERA is the latest attempt and I think I’ll just link to the creators of the stat itself (maybe break it down later when I get a second).
The biggest problem with all of these is that there’s not a good enough defensive metric to assert how much a defense can help. UZR/150 is currently the best we have and we’re still not 100% on top of it.
Regardless, what should you do about this?
I think, much like with slash stats, the bigger picture you get the better. Looking at a pitcher’s ERA first, then his FIP for what he’s doing with non-fielded balls, then his tRA to see what effect fielded balls have and you get a good idea of just how unlucky or lucky the pitcher has been.
If you have any questions, please post them in the comments and hopefully I can answer them for you.