September 4, 2007
Prospectus Hit and Run
From the You Learn Something New Every Day files... I'm not sure how many times over the years that I've referred to the quality start stat as a Bill James invention. Apparently, I'm mistaken. Coming across an old Rob Neyer column behind ESPN's subscription wall the other day, I was clued into the fact that the stat was defined by John Lowe of the Philadelphia Inquirer (now of the Detroit Free Press). James helped spread it to the masses via his Baseball Abstract series, which is where I first encountered it, but in this instance, he's overshadowed somebody else's worthy contribution. My apologies to Mr. Lowe for any failure to properly credit him in the past and to my readers for spreading such misinformation. Score that E-6.
As defined by Lowe, a quality start is one in which a pitcher goes at least six innings and allows no more than three earned runs. It's a simple and elegant stat that suggests a pitcher did a reasonable job of keeping his team in the ballgame. And while it's possible for a pitcher to earn a quality start with a game ERA of 4.50, such instances are rare. In the aforementioned ESPN column, Neyer found that in 2005, the average quality start featured a game ERA of 2.04, a non-quality start 7.70--that's not a misprint, it's Boeing's next jet--and the 6 inning/3 earned run/4.50 case constituted just 9.2 percent of all quality starts.
Based on this year's numbers, a team getting a quality start wins 68.0 percent of the time, on par with the 67.4 percent Neyer reported based on 1985 and 2005 data. Let's take a look at the leaderboard, with teams ranked by total number of quality starts:
Team Tot W L Win% NYN 77 59 18 .766 SDN 77 52 25 .675 CLE 76 54 22 .711 OAK 73 49 24 .671 ANA 72 56 16 .778 BOS 71 52 19 .732 TOR 70 50 20 .714 ARI 70 49 21 .700 HOU 69 44 25 .638 CHN 68 50 18 .735 ATL 67 49 18 .731 MIN 67 47 20 .701 SFN 67 42 25 .627 BAL 67 39 28 .582 CHA 66 42 24 .636 COL 65 45 20 .692 PHI 63 46 17 .730 PIT 63 42 21 .667 NYA 62 48 14 .774 LAN 61 48 13 .787 CIN 61 38 23 .623 MIL 60 41 19 .683 SLN 58 45 13 .776 DET 56 42 14 .750 TBA 56 36 20 .643 SEA 55 43 12 .782 KCA 55 39 16 .709 WAS 51 29 22 .569 FLO 44 29 15 .659 TEX 43 28 15 .651
The five top-ranked teams all play in parks that favor pitchers, presumably boosting their total number of quality starts. But if we compare this to the adjusted-for-everything metric SNLVAR (Support Neutral Lineup Adjusted Value Above Replacement, explained simply and clearly by Derek Jacques here if you need a refresher) on a team-wide basis, we find a very strong correlation (r = .87) between the two. In other words, they're telling pretty much us the same thing.
A quick look at the individual leaderboards for SNLVAR and Quality Starts (the latter compiled via the highly-recommended Baseball-Reference.com Play Index feature) confirms this. Between the two lists, we finds a high degree of redundancy, with players very closely bunched together on each:
Pitcher Team SNLVAR Pitcher Team QS Jake Peavy SDN 8.0 Danny Haren OAK 25 Tim Hudson ATL 6.7 Jake Peavy SDN 24 Brad Penny LAN 6.5 Brad Penny LAN 23 Kelvim Escobar ANA 6.3 John Smoltz ATL 22 Brandon Webb ARI 6.2 Fausto Carmona CLE 21 Erik Bedard BAL 6.0 Erik Bedard BAL 21 Johan Santana MIN 5.9 C.C. Sabathia CLE 21 Roy Oswalt HOU 5.8 Tim Hudson ATL 21 Dan Haren OAK 5.8 Tom Glavine NYN 21 Chris Young SDN 5.8 Johan Santana MIN 20 John Smoltz ATL 5.6 Kelvim Escobar ANA 19 Aaron Harang CIN 5.4 Mark Buehrle CHA 19 Fausto Carmona CLE 5.4 John Lackey ANA 19 Mark Buehrle CHA 5.4 Roy Oswalt HOU 19 Matt Cain SFN 5.3 Andy Pettitte NYA 19 Roy Halladay TOR 5.3 Matt Cain SFN 18 Tom Glavine NYN 5.2 Doug Davis ARI 18 C.C. Sabathia CLE 5.2 Gil Meche KCA 18 Joe Blanton OAK 5.2 Bronson Arroyo CIN 18 John Lackey ANA 5.1 Brandon Webb ARI 18 Andy Pettitte NYA 5.1 Orlando Hernandez NYN 17 Josh Beckett BOS 5.0 Roy Halladay TOR 17 Adam Wainwright SLN 5.0 Braden Looper SLN 17 John Maine NYN 5.0 Justin Verlander DET 17 Greg Maddux SDN 4.9 Daisuke Matsuzaka BOS 17 Chien-ming Wang NYA 4.9 Tom Gorzelanny PIT 17 James Shields TBA 4.9 Adam Wainwright SLN 17 Daisuke Matsuzaka BOS 4.9 Ian Snell PIT 17 Tom Gorzelanny PIT 4.8 James Shields TBA 17 Brian Bannister KCA 4.8 Livan Hernandez ARI 17 Orlando Hernandez NYN 4.7 Aaron Harang CIN 17 Justin Verlander DET 4.5 Joe Blanton OAK 17
Through Sunday, Haren and Peavy have just four non-quality starts apiece, Penny and Smoltz five, Carmona six, and Bedard and Escobar seven. That's at least a 75 percent rate of quality starts for each hurler, quite impressive--so much so that these pitchers probably belong in any Cy Young Award discussion, though at least on the NL side, Peavy's dominance on the SNLVAR list suggests he stands head and shoulders above the other candidates.
The pitchers on the SNLVAR list who aren't on the quality start list didn't miss by much; Beckett, Maddux, and Wang have 16 through Sunday, Maine and Young 15, and Bannister 14. Of the pitchers on the quality start list but not on the SNLVAR one, Meche, Looper, Davis, and Snell are all within one win of 30th-ranked Bannister; only Arroyo and Livan Hernandez are significantly further down the list.
As a metric, SNLVAR certainly has its advantages over quality starts. It adjusts for ballpark and opposition strength, strips out things a pitcher can't control like run support and bullpen support, and expresses the result in wins above replacement level. For my money, it's the best metric in the BP toolbox with which to measure starting pitchers, and as such, I use it every week in the Hit List, along with its bullpen sibling, WXRL. However, you can't eyeball SNLVAR over a cup of coffee and a page full of box scores, nor can you impress mixed company with such an unwieldy acronym, one which brings to mind that old Serak the Preparer line: "To pronounce it correctly, I would have to pull out your tongue." The humble quality start is perfect for just such occasions.
Then again, the quality start metric does lack the zazz we at BP like to apply to things, so it's worth passing along a little tidbit from Keith Woolner: our Support Neutral family can provide a sophisticated approximation of quality start rate if we untether ourselves from replacement level and turn towards league average via the per-game stat SNLVA_R (Support Neutral Lineup-adjusted Value Added Rate). Simply put, a pitcher's SNLVA_R + 0.5 is the percentage of the time his team would win a game given average offense and bullpen support. So for Jake Peavy, who's got an SNLVA of 5.3 in 28 starts and thus an SNLVA_R of .189, his team can be expected to win at a .689 clip. That's tops among pitchers with 100 or more innings this season. We'll close with that leaderboard:
Pitcher Team Rate Jake Peavy SDN .689 Kelvim Escobar ANA .645 Chris Young SDN .644 Brad Penny LAN .637 Tim Hudson ATL .636 Erik Bedard BAL .619 Brandon Webb ARI .618 Johan Santana MIN .614 Roy Oswalt HOU .612 John Smoltz ATL .611 Brian Bannister KCA .609 Orlando Hernandez NYN .607 Ben Sheets MIL .605 Fausto Carmona CLE .605 Josh Beckett BOS .604 Roy Halladay TOR .603 Dan Haren OAK .603 Chad Billingsley LAN .602 Mark Buehrle CHA .602 Matt Cain SFN .601 Chien-ming Wang NYA .600 Aaron Harang CIN .597 Jeremy Guthrie BAL .597 Shaun Marcum TOR .596 Adam Wainwright SLN .588 John Maine NYN .586 John Lackey ANA .585 Daisuke Matsuzaka BOS .584 C.C. Sabathia CLE .584 Tom Glavine NYN .581
Again, Peavy dominates the NL, while Escobar opens up a nice cushion in the AL. At the very least, that's something to clip and save for you next Cy Young argument.