Feng: What do predictive stats reveal about Tigers' bullpen?

By Ed Feng, Special to The Detroit News
Tigers closer Joakim Soria

How will the bullpen do this season?

It's the biggest question with Tigers fans this season. When Brad Ausmus walks to the mound in the seventh inning of a one-game playoff against Kansas City, can the bullpen hold that 4-3 lead? It didn't hold up last season in the series against Baltimore.

To evaluate the bullpen, you can't trust ERA. This dinosaur statistic considers too many factors over which a pitcher has no control, such as the official scorer, who assigns errors, and the clustering of hits.

ERA has even bigger problems when looking at the Tigers' pitchers from last season to the current season. The defense has made a remarkable improvement with the addition of SS Jose Iglesias and CF Anthony Gose. A good pitching statistic should isolate a pitcher's skill from these fielding improvements.

Defense independent pitching statistics

To develop a better statistic than ERA, we rely on the insights of Voros McCracken, who played a large role in Michael Lewis' iconic book "Moneyball."

McCracken asked a basic question: What does a pitcher have control over? Not surprisingly, he found that pitchers control the rate at which they strike out and walk batters. David Price struck out 9.5 batters per nine inning pitched last season, and this implies he'll likely have a high strikeout rate again this season.

However, McCracken discovered pitchers have no control over what happens when a ball gets hit into the field of play. He looked at batting average on balls in play (BABIP), and found that most pitchers allow 30 percent of these balls to go for hits.

Moreover, he found that BABIP was random from season to season for a pitcher. In 2013, Max Scherzer allowed a remarkable .259 average on balls in play, much lower than the .300 average. This hit suppression propelled him to 21 wins and the Cy Young Award. However, this BABIP .259 has almost no ability to predict his future ability to prevent hits on balls in play. In 2014, Scherzer had a BABIP of .315.

For those of you that uncomfortable with baseball numbers, this is the stupidest thing you've ever heard. Of course a pitcher controls how often he gives up hits on balls in play. When he throws a fastball over the middle of the plate, the hitter crushes it for a double. Your eyes assign fault to the pitcher for making a bad pitch based on the outcome.

However, numbers tell a different story. Pitchers have little control over BABIP. When a pitcher throws that fastball over the middle of the plate again, it doesn't always get crushed into the gap. Sometimes, it becomes a sharp grounder to the shortstop. Randomness plays a huge role when the bat strikes the ball.

Fielding Independent Pitching

Based on the pioneering work of McCracken, the baseball community has developed a statistic called fielding independent pitching (FIP) to evaluate pitchers. It relies primarily on walk and strikeout rate to come up with a number you can interpret like ERA.

If you're going to remember anything about this article, it should be this: FIP is better than ERA in representing the skill of a pitcher and hence has a better ability to predict future performance.

Fangraphs has an enhanced version of this statistic called xFIP. The original version of FIP used home run rate as an input, but the number of home runs allowed depends on the fly ball rate of the pitcher. xFIP uses data on fly balls to make a better pitcher assessment.

How are the Tiger relievers doing this season? Not well. The bullpen ranks 12th of 15 AL teams with xFIP of 4.28, ahead of only the Mariners and Twins. This suggests the bullpen will struggle the remainder of the season. Tigers relievers have a better ERA of 4.04, which suggest some help from defense and luck.

In my next post in two weeks, I'll use these defense independent pitching statistics to break down each of the Tigers' relievers.

Ed Feng has a Ph.D. in chemical engineering from Stanford and runs the sports analytics site The Power Rank.