Adjusted Quality Starts

Ah, the quality start. It’s one of several stats (along with Bill James’ Game Score and even the venerable, much-maligned pitcher win) designed to answer an age-old question: How well did the starting pitcher do his job? Most would agree that a pitcher’s job is to keep his team in the game and give his offense a good chance to win, and most would agree that even the bare-minimum quality start (6 IP, 3 ER) is at least an acceptable performance.

That said, the criteria for a quality start are pretty arbitrary, and they’ve invited a bit of criticism. Why is a six-inning, three-run performance a QS, but an eight-inning, four-run outing is not? Why do we say a pitcher had a “quality” performance if he pitched to the tune of a 4.50 ERA? What if a pitcher has a quality start through six, then blows it in the seventh inning or later?

The usual response to those criticisms is that, hey, they tend to get worked out in the aggregate. Overall, pitchers actually do post very good numbers in their quality starts. The 4.50 ERA “quality” pitcher is a myth.

Then again, if we want something that works out in the aggregate, why bother with QS? Most of the issues with plain old pitcher wins get worked out over time, too. Aren’t they a good enough proxy for quality performances?

Of course not. The idea behind quality starts is a good one. All we need is a clearer look at the question the stat is designed to answer, and we’ll have a better definition for the stat itself.

The question: “How many times did the pitcher give his team a good chance to win?”

The definition: A pitcher is awarded an Adjusted Quality Start (AQS) if he:

  1. Starts the game.
  2. Pitches at least six innings.
  3. Posts a run average (RA9) no worse than the league average.

#1 is, well, a requirement for something with the word “start” in its name. Moving on.

#2 is admittedly still arbitrary, but it’s a pretty good criterion, I think. A six-inning performance leaves only three innings to the bullpen, which isn’t all that much strain for most teams.

#3 is the change that gets to the heart of the issue. If the starter gives his team a decent number of at least league-average innings, then his teammates (assuming an average offense and average bullpen) should have at least a 50/50 shot at winning.

I use RA9 rather than ERA partly because of the well-documented issues with the definition of “earned” runs and partly because, as far as winning is concerned, it doesn’t especially matter whether a run is “earned” or not. A team that loses 4-3 because of four unearned runs still loses.

So, let’s put this metric to the test. In the American League in 2013, the league-average RA9 was 4.29, yielding three ways to post an AQS.

1) Pitch at least 6 innings, give up 2 or fewer runs.

This is by far the most common AQS because it includes all the zero-, one- and two-run starts. The top 10 in this sort of AQS were:

James Shields 20
Max Scherzer 20
Hisashi Iwakuma 19
Felix Hernandez 19
Bartolo Colon 19
Derek Holland 17
Ervin Santana 16
Anibal Sanchex 16
Justin Masterson 16
5 tied with 15

2) Pitch at least 6.1 innings, give up 3 runs.

Chris Sale was the master of the exactly-three-run AQS, as he did it 8 times in 2013. The top 10:

Chris Sale 8
Justin Verlander 7
James Shields 7
CC Sabathia 6
Jarrod Parker 6
Doug Fister 6
Yu Darvish 6
C.J. Wilson 5
7 tied with 4

3) Pitch at least 8.1 innings, give up 4 runs.

Unsurprisingly, this was by far the least common sort of AQS. It’s not often that a starter who gives up that many runs is allowed to pitch into the ninth. In fact, it only happened twice in the AL last year: CC Sabathia’s four-run, complete-game victory on June 5, and Corey Kluber’s 8.2-inning, four-run win on July 31.

Overall, your 2013 AL leaders in AQS:

James Shields 27
Max Scherzer 23
Hisashi Iwakuma 22
Chris Sale 22
Bartolo Colon 21
Doug Fister 21
Jarrod Parker 21
Justin Verlander 21
Yu Darvish 20
Felix Hernandez 20

Over in the National League, the average RA9 was a tick lower at 4.04. That’s not going to affect the two-and-fewer starts, but it’ll set the bar for the three- and four-run AQS a little higher.

1) Pitch at least 6 innings, give up 2 or fewer runs.

I doubt anyone will be surprised by the name at the top of the list.

Clayton Kershaw 22
Cole Hamels 20
Patrick Corbin 20
Jordan Zimmermann 19
Travis Wood 19
Madison Bumgarner 19
Zack Greinke 18
Gio Gonzalez 18
6 tied with 17

2) Pitch at least 7 innings, give up 3 earned runs.

Adam Wainwright 6
Cliff Lee 6
Mike Minor 4
Kris Medlen 4
Clayton Kershaw 4
Kyle Lohse 3
Cole Hamels 3
Andrew Cashner 3
Bronson Arroyo 3
13 tied with 2

(As an aside, the 6.2-inning, 3-run start just missed the cut-off in the NL, as that would be a 4.05 RA9 against a league average of 4.04. There were 22 starts that met those criteria in the NL last year, and I debated including them, but it would only make minor changes to the leaderboard. Mat Latos takes home the Just Missed Award with three such starts.)

3) Pitch at least 9 innings, give up 4 earned runs.

Only one NL pitcher pulled this one off in 2013. That was Brandon McCarthy, who gave up four in a complete-game loss on September 2.

Finally, your 2013 NL leaders in AQS:

Clayton Kershaw 26
Cole Hamels 23
Adam Wainwright 23
Cliff Lee 22
Madison Bumgarner 21
Patrick Corbin 21
Travis Wood 21
Jordan Zimmermann 21
Matt Cain 18
Zack Greinke 18
Gio Gonzalez 18
Lance Lynn 18

(If we include the 6.2-inning, 3-run starts, Gonzalez and Lynn take sole possession of ninth and tenth place with 20 and 19 AQS, respectively.)

There’s more that could be done with Adjusted Quality Starts – park factors, for instance – but that’s probably too much work for a stat that’s supposed to answer a pretty narrow question. If you want to know how often a starter kept his team in the game, this is a good, er, start.

newest oldest most voted

I think the QS or AQS would best be done at full inning intervals, since they are by far the most common times that a pitcher comes out of the game. Being 8.1 IP rules out a lot of pitchers who are removed after only one fewer out in favour of the closer.

I think overall the number would be more reflective of chances to win if the bar was set at 6/2, 7/3, 9/4 for both leagues. Just my personal preference I guess.

Pat Sopko
Pat Sopko

I like whole IP-RA>3

Pat Sopko
Pat Sopko

I use this as both a quantitave(pts= whole ip-ra) & quantitative (pts/gs), and weight each equally. My top14 for last yr, just for giggles….Kershaw, Wainwright, Lee, Iwakuma, Shields, Harvey, Scherzer, Colon, Fernandez, Zimmerman, Henandez, Sale, Bumgarner, Darvish


I think it’d be nice to adjust the average runs to the park. I do like the idea, though!

Nick C

My problem with adjusted QS is that the league average RA9 changes throughout the seasons does it not? If we were to use RA9 we would have to decide if we use the current league average, the previous seasons average, or some sort of combination. If we used a current average, a pitcher could in theory be awarded a AQS for a, let’s say a 7 inning 3 run game in April, but not in july if the average RA9 has changed. A fix would be to only record the statistic after the season, but for fans who like looking… Read more »