Tuesday, May 21, 2013


OK, we all know what an Earned Run Average (ERA) is; the number of earned runs given up per nine innings, determined by a formula so simple we're surprised the nuns didn't use to teach math - earned runs/innings pitched X nine. It's what sabermetric statheads like to call a "counting number," that is just a raw, unweighted figure that's subject to a jillion variables - the league, the park, starting v relieving, the defense behind a pitcher, teams faced, yada yada.

Most baseball fans accept it for what it is, and use it simply to determine how effective a guy has been on the mound at a given point of the season. But all those outside factors gave a prickly rash to the computer gamers' sensibilities; they hate numeric chaos, even if it is somewhat organized. So they went to work on developing a better indicator. All sorts of different numbers were churned out of databases, with different emphasis on different aspects of the game.

The simplest and most popular was an average that measured the two-man game between the pitcher and hitter. It used the three true outcomes of baseball lore - homers, walks (+ hit batters - intentional walks), & strikeouts - and eliminated the fielding factor entirely, assuming that everyone was pitching in front of a league average defense in a league average park giving up a league average amount of knocks. It became known as Fielding Independent Pitching (FIP).

The results were weighed in that order - HR, BB/HBP, K - and produced an ERA-like number that the boys in the basement believed not only better represented a pitcher's current performance, but was a strong indicator of future performance as well. Its formula is also fairly straightforward: 13HR+3BB-2K/IP+ 3.2 (a constant to make it an ERA-compatible number) =FIP. You're sorry you slept through that junior high algebra class right about now, aren't you? It gets trickier.

In some hurlers' cases, it was fairly close to ERA; in others, it indicated that regression, up or down, was in the cards. And hey, it gave agents and GMs another number to throw out at contract time. But guess what? That home run variable was a nagging red flag; it reeked of randomness, just like regular balls in play. And viola - Expected Fielding Independent Pitching (xFIP) was born.

It replaced the home runs, a variable that can fluctuate considerably year-to-year, with a number calculated by taking the pitcher's fly ball rate X the league average % of homers per fly ball. It was closer to eliminating any outside influences, but it didn't wholly scratch the sabermetric itch, either. Yep, another stat was on its way.

This one is Skill-Interactive ERA (SIERA). We'll not share its formula because unless you're a budding Einstein, it's pretty dang complicated. SIERA's co-efficients have co-efficients. But it measures a lot of stuff - strikeouts, walks, types of contact, situational results and consists of a whole lot of hardball number crunching; it's one thorough system. It's the only one of the three that actually factors in balls in play, which are 70% of a game's action.

In a nutshell, ERA is what it is. FIP is based on the three true outcomes that a pitcher controls; the other variables are reduced to league average. xFIP swaps out HR for a number based on fly ball rate of a pitcher v league average fly ball rate for HR, and SIERA is a comprehensive calculation that considers and weighs the variables of just not the three true outcomes but balls in play by pitch type, park and situation.

What's it mean? Well, FIP, xFIP and SIERA are all predictive, so they have use that way; SIERA does the best job of matching what is with what should be, providing a handy present value. But we're kinda old school; we like counting numbers because they're real and now, and use the rest of the stuff as a guidepost to what to expect down the road.

If you're curious, here's a sampling of Pirate pitchers with their various lines. In some cases, the numbers are pretty similar, and for others, doom and despair are around the corner, though it's good news for beleaguered Tony Watson, who some advanced stats say has equaled (or bettered) Justin Wilson.

AJ 2.57 2.43 2.68 2.78
Wandy 3.25 4.68 3.99 3.93
J Locke 2.73 4.47 4.44 4.75
J Gomez 2.78 5.26 4.41 4.55
J Grilli 0.92 0.64 1.88 1.20
M Melancon 0.78 1.73 1.66 1.20
J Wilson 1.40 2.95 3.97 3.71
T Watson 5.23 4.29 3.94 3.23

(Numbers from Fangraphs)

No comments: