I can’t get into specifics (so please don’t ask) but I will say this: there are things that are generally publicly held as sabermetric doctrine—in some cases, crucial underlying assumptions—that are demonstrably false. — Russell A. Carlton (nee Pizza Cutter) for Baseball Prospectus
The first frontier for baseball research was correctly rewarding hitters for their work at the plate. Branch Rickey started work on that subject in the 1950s with an equation that attempted to put a number on offense and defense. The second frontier might have been correctly rewarding pitchers for their work on the mound. Voros McCracken famously created the DIPS theory of pitching by showing that pitchers had very little control over a ball once it was in play.
The current frontier is not so well defined. There’s great work being done on pitching injuries, the effect of the shift, and process-based metrics that attempt to quantify ‘stuff.’ We’re constantly trying to stay on top here at Roto-Relevant Research.
But there is a concept that can serve to unite most of these: refinement. Perhaps the current era is one of refinement and reappraisal. We knew something about the relationship between pitches and injuries, and now we’re refining it. We knew what ‘stuff’ was by sight, but now we’re refining it. We thought we knew some things and now it’s time to re-question what we thought we knew, too.
And maybe we knew, all along, that taking all of the credit for a ball in play away from the pitcher was a drastic move that didn’t feel quite right. In the defense of the sabermetric community, they’ve been working on refining DIPS theory for some time. If FIP (Fielding Independent Pitching, created exclusively using walks, strikeouts and home runs) was the first generation, xFIP helped adjust the home run rate based on fly ball rate and league average home runs per fly ball — and SIERA (or Skill Interactive ERA) gave ground-ball pitchers more credit for avoiding the home run, among other things. A fly ball and a ground ball are fundamentally different, and pitchers do have control over those events. So we’ve adjusted.
FanGraphs made a major adjustment (or refinement) this week. fWAR is based on FIP, and so it removes any credit for balls in play and good work out of the stretch. If there were some that felt it was flawed, the other WAR frameworks had similar issues. Baseball Reference WAR attempts to give some credit on balls in play to the pitcher by judging team defense and awarding that defensive work to each pitcher on the team. But pitchers get uneven performances from the defense behind them, so that approach has its own flaws. Just look at the BIP-wins for pitchers on the same team, like Jason Vargas (1.9 BIP-wins) and Kevin Millwood (-.7 BIP-wins).
In an effort to give readers the ability to see the components of a pitcher’s value, FanGraphs retained their FIP-based WAR, but added BIP-wins and LOB-wins in order to create a second value metric, called FDP-wins. Put the two together, and you have a descriptive statistic for the year gone by.
But we still don’t know how prescriptive the statistics are. And FanGraphs EIC Dave Cameron admitted as much when he rolled out the stats. Of course, the prescriptive nature of the stats is what’s most interesting to fantasy players.
For example, it’s nice to know, from both a real-life and fantasy perspective, that Jason Vargas has been worth over three-and-half wins above replacement if you give him credit for his BABIP and LOB% this season. His fantasy owners will probably echo the thought that he’s been valuable this season, too. But can you count on that to continue? Especially if he leaves his home park?
The reason his home park seems so important becomes clear if you sort the qualified starters of the 2000s by BIP -wins. The whole framework for DIPS was that pitchers couldn’t really control their BABIP, and if you look at the leaderboard, you won’t find many two-time entrants. In the top 50, only Felix Hernandez, Pedro Martinez, Randy Johnson and Tim Lincecum appeared more than twice. Hard to bet your 2012 BIP-winner is one of those guys right off the bat. (And it’s worth mentioning here that SIERA awards high-strikeout guys lower BABIPs, and that might be the case here.)
On the other hand, if you redo the count for teams in the top 50, you get some interesting results. The Cardinals have five entrants. The Red Sox had five. The Giants had five. The Mariners had three. Even if you count the Phillies (three) and the Diamondbacks (four), you’re still talking about some very interesting parks. Each has a quirk.
St. Louis has a pitching-friendly park, but its particularly suppressive of home runs by righties (89 by FanGraphs park factors). Boston has a generally offensive-friendly park but it suppresses home runs by lefties by 8%. We know about the parks in Seattle and San Francisco, but even Philadelphia and Arizona have below-average park factors for one event or another. Maybe we’ll find in the future that certain arsenals play up in certain parks — more than the park-adjusting can account for, generally. The lesson here for fantasy is that there might be a bonus you can give to a pitcher like Jason Vargas next season, provided he continues to pitch in Seattle.
The league strands baserunners at about a 70-72% rate, but each pitcher has his own number. LOB-wins try to capture the value that a pitcher gets from stranding runners, and there is a chance there’s a skill there — some pitchers are legitimately bad or good at stranding baserunners. But do the count, and only Tim Hudson and Roy Oswalt end up on the list more than twice. Jake Peavy, Livan Hernandez and Tom Glavine managed it twice. It’s a list that you’re much more likely to only end up on once. That suggests that there’s less skill involved in stranding runners. Even on the bad side — and a scout will tell you that it’s possible for a pitcher to bad out of the stretch — only Greg Maddux shows up more than twice, and A.J. Burnett, Eric Milton, Jason Schmidt, Mat Cain and Randy Johnson manage it twice.
Looking at cumulative wins doesn’t clear much up. Barry Zito, Ted Lilly and Matt Cain lead the 2000s in BIP-wins. So it must be about fly balls. Except Tim Hudson is fifth. (There is something to the fact, perhaps, that five of the top six (and eight of the top 11) in this category played in pitcher’s parks, though, maybe.) Livan Hernandez, Roy Oswalt and Tom Glavine have more LOB-wins than anyone, how are you going to pair those guys up? It’s interesting that Ricky Nolasco is 26th-worst in BIP-wins, and tenth-worst in LOB-wins, since he’s underperformed his FIP for a while. Michael Jong, at least, thought some of his problems came from working in the stretch, so it’s not a… stretch.
If these things were bonafide skills, wouldn’t there be more pitchers that show that skill at an elite level (or lack of that skill) more often than 1/4 of the years in a 12-year sample? Wouldn’t it be easier to discern what about them made them a leader in the category? It seems that we have a lot to learn.
Maybe, when trying to figure out the keeper value of a pitcher, you can peruse his BIP-wins and LOB-wins anyway. Let’s say you are looking at Johnny Cueto, who has outperformed his FIP two years in a row. You think there’s some skill FIP is missing, and so you check out his value section on FanGraphs. Last year, Cueto had 1.7 BIP-wins from a .249 BABIP, but this year, with a .291 BABIP, he only has .2 BIP-wins. Last year, Cueto had a reasonable 76.4% left on base percentage and only had .1 LOB-wins. Now his LOB% is practically 80% and he has 1.6 LOB-wins. If he has a skill that allows him to beat FIP, it changes from season to season.
Matt Cain on the other hand? He’s had more than a win and half from balls in play in every year save two (0.9 in 2007 and -0.1 in 2008). Maybe he has some sort of skill, or maybe his arsenal fits his home park really well. It looks like folly to bet against him at this point either way.
Thanks to these new numbers, one-stop reality-checking and sleeper-finding got quicker for fantasy players. And a whole new research road became clear as well. That doesn’t mean any of this will be easy.