On quality starts and dominant starts

There are many methods of measuring pitcher performance, and how good a pitcher really is. Can you really draft for fantasy leagues based on won-loss records and ERA? Experience tells me, not really. So what can you look at?

How useful is the definition of a “quality start”? In 1985, John Lowe sought to define what a good starting pitcher ought to achieve in order to give his club a high chance of victory. If a pitcher lasts at least six innings while allowing three earned runs or less, he gets credit for a quality start.

David W. Smith showed back in 1992 that teams who got quality starts won about 70% of the time. So if you could get all quality starts all year long, you should finish about 113-49. Not bad.

But the quality start statistic has also received criticism from many, primarily because of the minimum requirements. Three earned runs in six innings of work corresponds with an ERA of 4.50, which doesn’t sound like a “quality” pitcher. And what of the pitcher who tosses a complete game but allows four runs? He doesn’t get credit for a quality start, yet his 4.00 ERA is better than the guy who meets the minimum requirement for a QS, and gives his team a better chance to win. Bill James has pointed out that these cases are indeed problematic, but few and far between, rendering them statistically negligible.

Checking recent years, from 2004-2007, I found that Smith’s findings still hold generally. Getting a quality start from your pitcher gets you a win 68.2% of the time.

Year    Total QS   	AL QS  		NL QS

2004	1566-711 (.688)	712-295 (.707)	854-416 (.672)
2005	1651-795 (.675)	755-354 (.681)	896-441 (.670)
2006	1566-713 (.687)	737-296 (.713)	829-417 (.665)
2007	1312-618 (.680)	618-292 (.679)	694-326 (.680)

Totals	6095-2837(.682)	2822-1237(.695)	3273-1600(.672)

The National League averages about 200 more quality starts per year than the American League. You can rationalize this with the presence of the designated hitter, who replaces the easy-out pitcher in the lineup with one of the best hitters in the lineup. Interestingly, the winning percentage in the AL is consistently higher than the NL for a quality start. Perhaps this reflects the fact that a quality start in the AL is different than a quality start in the NL. Or it may be due to overall bullpen quality. Or the liberal use of pinch hitters, bunts and small ball.

Looking at the quality starts leaders for this season, and the percentage of starts which are considered quality, we see some very good pitchers:

QS	QS%	Name

25	89.3	Dan Haren
23	82.1	Brad Penny
23	85.2	Jake Peavy
21	80.8	John Smoltz
21	72.4	C.C. Sabathia
21	75.0	Tim Hudson
21	80.8	Fausto Carmona
21	75.0	Erik Bedard
20	71.4	Johan Santana
20	71.4	Tom Glavine
19	67.9	Andy Pettite
19	67.9	Roy Oswalt
19	70.3	John Lackey
19	76.0	Kelvim Escobar
18	64.3	Brandon Webb
18	66.7	Matt Cain
18	69.2	Mark Buehrle

The number of quality starts does give you a sense of how good a pitcher is, but it’s not the whole story. You might say that it tells you more about consistency than dominance. I’d like to propose another related measure, called a dominant start.

The Dominant Start

What defines dominance? Unhittableness? Strikeout rates? How good of a chance your team has to win? It’s difficult to extract that kind of information from standard pitching stats. What you get is a conglomerate of all of a pitcher’s performances, which rewards consistency far more than one dominant outing. So if we analyze outings individually, we get a truer sense of potential, because we see peaks that otherwise get smoothed out by totaling and averaging.

So how should we define dominance? I think a pitcher should go at least six or seven innings, and allow 2 runs or less. Why two runs? Because a pitcher can dominate a whole game, but walk one and allow a home run, and that’s two earned runs. Or a walk and two hits in a bad inning can yield two runs, but typically not three. The starter should also strikeout at least a batter an inning to dominant. Why do K’s count? Because strikeouts lessen the effects of lucky hits and subpar defense out of the equation, and speak to the skill and deception of the pitcher on a given night.

To decide between 6 or 7 innings, I tested the two cases for the years 2004-2007:

League	6 innings	7 innings
AL	848-251 (.772)	428-105 (.803)
NL	1078-359(.750)	560-148 (.791)

Based on the above, I would propose that at least seven innings pitched while allowing two earned runs or less, AND striking out seven or more hitters constitutes a dominant outing. It works out nicely that a quality start gives you a 70% chance of winning the game, while a dominant start puts you right near 80%.

And now, without further ado, the dominant start leaders of 2007:

AL Dominant starts, 2007 

DOM	DOM%	Name
10	35.7	Erik Bedard
9	31.0	C.C. Sabathia
8	28.6	Johan Santana
8	29.6	Daisuke Matsuzaka
8	32.0	Josh Beckett
6	21.8	James Shields
6	21.8	Dan Haren
5	20.0	Kelvim Escobar
4	15.4	Justin Verlander
4	15.4	Javier Vazquez
4	16.0	Jeremy Bonderman
3	12.5	Felix Hernandez
3	11.5	Roy Halladay

NL Dominant starts, 2007

DOM	DOM%	Name
9	32.1	Brandon Webb
9	33.3	Jake Peavy
9	32.1	Aaron Harang
7	28.0	Cole Hamels
6	26.1	Chris Young
5	19.2	John Smoltz
5	17.9	Brad Penny
5	23.8	Tim Lincecum
5	18.5	Matt Cain
4	14.8	Barry Zito
4	15.4	Wandy Rodriguez
4	15.4	John Maine
4	15.4	Rich Hill
4	18.2	Orlando Hernandez

I’d rather draft this second group of pitchers over the quality starts group, any day. Dominant starts gives you very good pitchers, but filters out the lower fantasy performers, and seems to incorporate some of that all-important “upside”.

So together the quality start percentage and dominant start percentage tell you how consistent a pitcher’s been, versus how dominant he’s been in specific outings. You could easily use the dominance measure to data mine for future aces and breakout pitchers for fantasy play. It would also be interesting to see how quality start percentage and dominant start percentage come into play over a pitcher’s career. But we’ll leave that until next time.