Getting Dirty With Stats – Pitching Estimators

Getting Dirty with StatsIntro | Batting 1 – Linear Weights | Batting 2 – wOBA and wRC+

Lets continue Getting Dirty With Stats! We’ve already covered an introduction, which is important as I lay out my goals for the series. We’ve also talked about hitters, so now we move on to pitchers.

As with hitters, we’ll start by figuring out what things pitchers are and aren’t responsible for. These include the home park a pitcher plays in, which can vary in size and conditions, as well as his league, which varies in talent level. Both are important variables we accounted for when looking at hitters and we’ll do the same for pitchers.

Another thing we’ll need to think about is the defense behind them. Pitchers spend long periods of time with the same or a similar defense. Since defenses vary in skill levels and shouldn’t affect the way we evaluate pitchers, we’ll need to consider that.

Another key difference between hitters and pitchers is the role of luck on what happens to balls put into play.

Here’s one of the greatest and most controversial discoveries of the recent statistical revolution: pitchers have much less control over balls in play than hitters.

This isn’t intuitive. It seems weird to give a pitcher complete blame for giving up a fly ball that lands just over the fence, but no blame for a ball that lands just in front of it. Although not everyone agrees that pitchers have no control whatsoever over batting average on balls in play, also known as BABIP, lots of people agree that pitchers have much less control than previously thought. So while it isn’t perfect, BABIP is pretty good way to tell how much luck is involved in a pitcher’s results when looking at traditional stats.

So how do we deal with all this? Lets start with the current king of pitching stats, Earned Run Average. Earned runs are attributed to a pitcher based on errors. Those errors are decided by the game’s scorekeeper. Unfortunately, scorekeepers suffer from the same imperfections in judgement that affect all humans. I’ll cover errors in a future post on defense, but  the short explanation is that due to limited range, bad defenders aren’t charged with enough errors.  And since good defenders get to more balls to make a play on, they’re charged with too many errors as a result.

Earned runs also don’t account for differences in ballparks, or differences in leagues. And, ridiculously, earned runs allow relievers to decide the fate of the previous pitcher’s runners left on base. ERA- (or the reverse, ERA+) account for park and league differences but none of the other issues mentioned. There has to be a better way! And there is…

Fielding Independent Pitching, or FIP, looks at looks at things we absolutely know pitchers control: home runs, walks, and strikeouts. While FIP is scaled to look like ERA, the better option is FIP- which considers the pitcher’s park and league. It works the same as wRC+ does for hitters: 100 is average, 80 is really good, and 120 is really bad.

Some pretty smart people think pitchers don’t have much control over how often fly balls turn into home runs. We can use xFIP, or Expected FIP, to deal with this, which replaces the pitcher’s rate of home runs per fly ball with the league average. That fixes the problem of counting balls that land just before/after the fence differently. It otherwise looks and acts like FIP.

Only looking at walks, strikeouts, and home runs might seem weird at first. It did for me. The reason it’s so popular is that while other results like line drives and grounders are still useful, assigning credit and blame for what happens is complicated. We’re still figuring out the best way to deal with them. With FIP, at least we know we’re not wasting our time with stuff that doesn’t matter. So while it may not include all the information, it definitely doesn’t include bad information.

Luckily for baseball fans, people who study baseball aren’t known for giving up. Bits of progress are made by building on top of previous ideas to slowly create a better way to measure the game, which lead to new statistics. Not because previous stats were “wrong”, but because we understand a little more than we did before.

This brings us to SIERA. It’s based on FIP, but includes bits about types of hits we know we can blame on the pitcher, like line drives and ground ball tendencies. I’ll let Fangraphs get into the nitty gritty:

In general, groundballs go for hits more often than flyballs (although they don’t result in extra base hits as often). But the higher a pitcher’s groundball rate, the easier it is for their defense to turn those ground balls into outs. … And if a pitcher walks a large number of batters and also has a high groundball rate, their double-play rate will be higher as well.

Balls in play are complicated, huh? Luckily, as with wRC+, the math is already done for us with SIERA. It’s set to the same scale as ERA, and is park adjusted. SIERA only measures hits as they come off the bat (home runs, grounders, line drives, etc). So the quality of a pitcher’s defense doesn’t mess anything up. Oddly SIERA isn’t league adjusted, I’m not sure why.

Here are the Padres 2013 SIERA leaders, and Jason Marquis.

2013 SIERA
Tyson Ross 3.20
Andrew Cashner 3.70
Eric Stults 4.21
Jason Marquis 5.14

Jason Marquis finished 7th on the team (min 40 IP), but I’m including him because he was a special case. He finished the season at a 4.05 traditional ERA. While that looked good, a lot of the things that ERA doesn’t account for indicated he was perhaps more lucky than it would seem, wich is shown in his 5.14 SIERA. Unfortunately his injury kept us from learning how things would play out, but his season serves as a great example of the differences in looking at pitching.


You are encouraged to comment using an exisitng Twitter, Facebook, or Google account. Upvote comments you find helpful, and only downvote comments that do not belong. The downvote is not a 'disagree' button.