Learning from past mistakes: who should move to the bullpen?

A few weeks ago, I posted a story on how to calculate WAR for pitchers. Now, I'm going use the move of Z to the bullpen to show how to employ pitching WAR.

Heading into the season, Zambrano was projected to be the Cubs' 2nd-best starter, with an expected FIP of 3.95 that was behind only Ryan Dempster (3.86) and a projected ERA (3.86) only behind Dempster (3.81) and Ted Lilly (3.78). When you consider Z's prowess as a hitting pitcher, he's probably the Cubs' best player to put on the mound on any given day. However, the season started off a little rocky for Zambrano, and for the Cubs' bullpen. Meanwhile, Randy Wells, Tom Gorzelanny, and Carlos Silva all performed admirably while Ted Lilly was out with an injury. When Lilly came back, the team decided to move Zambrano to the bullpen. At that point, they already had enough lefties in the pen (apparently ruling out Gorzelanny and Lilly) didn't want to mess with Silva or Wells, and Dempster was pitching like the team's ace. Meanwhile, Zambrano had a few rough starts and a bloated ERA. So was this decision wise? Well, no. It was pretty damn stupid. More importantly, who should they send to the pen now, with Z returning to the rotation? In my opinion, that should be Carlos Silva. Follow me below the fold to see how I answered these questions using modern pitching metrics...

I've already told you Z was projected to be our best pitcher. Even after his "horrible" April, he was still projected to be one of the three best starters on the roster by a comfortable margin. Now, to knock down the strawman argument of "The projections are wrong! Looks at Z's ERA!" If you look at his Z's peripherals (which ALL teams should be doing these days) he wasn't nearly as bad as his top-line numbers were suggesting. Z was the victim of the team's biggest problem this year - a slap in the face from Lady Luck. More on this later. Besides, the "logic" behind moving him to the bullpen was that he'd also become the best pitcher there, and thus would improve an area of weakness on the roster. HOWEVER, moving him to the bullpen also made the rotation weaker. And that's where the problem lies.

As a setup man in the bullpen a pitcher typically throws somewhere between 75-90 innings/season. 90 innings is usually at or near the league lead in relief innings, and it's usually what the top setup men achieve. Think of 2008, when Carlos Marmol was a "lights-out" setup man who seemed to pitch every day. He accumulated 87 innings that season. Meanwhile, starting pitchers usually more than double that output: 200 innings is often used as an easy to remember criterion for a durable starting pitcher, and a starter that gets hit with a couple injuries can still rack up 150 innings over the course of a season. So moving a pitcher from the rotation to the bullpen essentially gives them less playing time. To be fair, some innings are more important than others (the term we use for this is leverage). To account for this, when projecting the value of relievers we multiply their WAR by a factor of 1.7. But that's not enough to offset the factor of 2 or more in the innings discrepancy. The bottom line is this: you want your best players getting the most playing time. It's really that simple and you don't need stats to understand this principle. Relievers are much less important than starters, so you want your best pitchers starting. Suggesting otherwise is acceptable from fans that don't know the stats behind these decisions and that are creatively throwing ideas at the wall to "turn the team around." It's even an acceptable idea to throw around during a brainstorming session in a front office... but to act out such a suggestion is completely unacceptable from the supposedly serious-minded people running Major League Baseball teams. It's indefensible.

I hope everyone can see why moving Z to the pen was a bad decision: he was and still is one of the team's top-5 pitchers. Besides that, he's the best hitting pitcher in the league, the best competitor I've ever seen in a Cubs uniform, and my favorite Cub. Needless to say - stats or no stats - I'm anxiously awaiting his return to the rotation. So the question we have to ask now is the one that should have been asked a month ago... or at least answered better: which 5 players are the team's 5 best starting pitchers? Here are some numbers to help us decide:

(SKIP THIS PARAGRAPH IF YOU ALREADY KNOW WHAT FIP and xFIP ARE.) Across that table, you see the players ERA, FIP, and xFIP in their career, in 2010, and in their rest-of-season ZiPS projections, as well as their 2010 W-L record. ERA and W-L everyone here should be familiar with. FIP is the runs/game a pitcher should allow based on the number of strikeouts, walks, and home runs they've allowed - it is a way to predict ERA based on those numbers alone; note that this metric doesn't rely on balls in play that fielders have a say in. Discrepancies between FIP and ERA are thus attributable to the quality of the defense playing behind the pitcher, luck in those balls in play turning into hits (and runs), and to some degree, the ability (or lack thereof) to induce easily-fielded balls (think grounders). xFIP makes a further luck adjustment to FIP based on the expected number of HR's given the number of fly balls the pitcher has allowed. So discrepancies between FIP and xFIP are attributable to luck in the number of allowed fly balls that have left the park. And here's the important thing: a simple study by Colin Wyers shows that FIP and xFIP are better predictors of ERA than ERA itself is. In other words, if you want to find out how much a pitcher's success in the current season is repeatable going forward, you're better off using projected FIP, xFIP, and FIP... and NOT ERA, which is a function of luck and defense.

So what can we discern from these data? Well, over the course of their careers, Tom Gorzelanny and Carlos Silva have been - by far - the worst pitchers on this list. In 2010, two things jump off the table for me: 1.) Carlos Zambrano has been very, very as he's pitched much better than his peripherals, particularly ERA, suggest; and 2.) Carlos Silva and Ted Lilly have, conversely, been fairly fortunate in the number of runs they've allowed. Finally... if we look at the projected numbers, which include this year's performances and weight recent years more heavily than distant years, one pitcher is expected to do much worse than his peers: Carlos Silva. Looking at that table, there's only two numbers that support Silva's place in the rotation: 6 and 0. But the W-L record is an outdated, poor attempt at assigning wins to a pitcher's performance. WAR is a much better way to go about this, and WAR is based on FIP (or xFIP or tTA, not shown here), all of which are numbers Silva is the worst at.

In conclusion, if you're trying to win this year, Silva should go to the bullpen for the same reason Zambrano should not have: the best 5 starting pitchers on this team are Zambrano, Dempster, Lilly, Wells, and Gorzelanny. The only reason not to move Silva to the bullpen is if the team is trying to translate his solid but lucky 6-0 start into an opportunity to unload his bloated... contract.