Predicting the Quality Start

As you well know, fantasy formats have long been eschewing (and ridiculing) the use of the win as a category. It tends to still hold on rather stubbornly in standard Rotisserie 5×5 formats which are also widely panned (yet this author still clings to one of those teams annually). There are a number of logical swaps for the win, and one of them has historically been the “quality start,” which is what this post is all about.

The quality start has also been criticized as being rather useless inasmuch as describing whether a pitcher performed well nor not, and yet if you don’t want to get into weighted metrics in your fantasy league, it’s still preferable to the sometimes arbitrary assignment of wins (and perhaps more on point — the arbitrary lack of assigning a win).

Just so we’re operating with the same definition, in most fantasy circles the quality start still uses the John Lowe Philly Inquirer characterization as being a starting pitcher going six innings without giving up more than three earned runs. We can punch holes in that over beers another day, but that’s our baseline for a quality start going forward.

What this post will seek to do is provide you with some basic raw data on quality starts from 2014. We’ll analyze correlations among a variety of variables with the quality start, and then seek out to develop a predictive model that might help you target pitchers most likely to register a goodly number of quality starts in 2015.

You might want to maximize that browser size to absorb a rather involved chart of data below. I somewhat arbitrarily chose 150 innings as my qualified cut off for starting pitchers in 2014. I say somewhat because in my head I really wanted at least 100 pitchers and I got 98 and 150 innings is nice and tidy, so, Yahtzee. There was far more data in the original chart, but I’ve distilled this down to include many usual suspects plus a couple which require a quick explanation.

“Pitches” is the total number of pitches — we need that to calculate the data point at the end of the chart, “PPI” or pitches per inning. QS is our total raw number of quality starts, but obviously that can be impacted by the total number of starts one had during the season, so next to it we have QS% or the percentage of quality starts per game started. No surprise to see Mr. Kershaw there. QS-W was mainly for my curiosity, which is the number of quality starts to wins, or more crudely put — who got screwed out of a win and who managed to net more wins than actual quality starts, of which there are few (Bud Norris, take a bow). Everything else should be self explanatory.

	GS	IP	Pitches	W	TBF	H	R	ER	HR	BB	BB%	SO	K%	K/9	BB/9	WHIP	H/IP	QS	QS%	QS-W	PPI
Clayton Kershaw	27	198.1	2722	21	749	139	42	39	9	31	4%	239	32%	10.85	1.41	0.86	0.70	24	89%	3	13.74
Johnny Cueto	34	243.2	3659	20	961	169	69	61	22	65	7%	242	25%	8.94	2.4	0.96	0.69	29	85%	9	15.05
Jon Lester	32	219.2	3493	16	885	194	76	60	16	48	5%	220	25%	9.01	1.97	1.1	0.89	27	84%	11	15.94
Cole Hamels	30	204.2	3136	9	829	176	60	56	14	59	7%	198	24%	8.71	2.59	1.15	0.86	25	83%	16	15.36
Chris Sale	26	174	2753	12	685	129	48	42	13	39	6%	208	30%	10.76	2.02	0.97	0.74	21	81%	9	15.82
Felix Hernandez	34	236	3434	15	912	170	68	56	16	46	5%	248	27%	9.46	1.75	0.92	0.72	27	79%	12	14.55
Alex Wood	24	156.1	2420	8	625	132	50	45	14	39	6%	151	24%	8.69	2.25	1.09	0.85	19	79%	11	15.50
Sonny Gray	33	219	3295	14	899	187	84	75	15	74	8%	183	20%	7.52	3.04	1.19	0.85	26	79%	12	15.05
Adam Wainwright	32	227	3258	20	898	184	64	60	10	50	6%	179	20%	7.1	1.98	1.03	0.81	25	78%	5	14.35
Corey Kluber	34	235.2	3500	18	951	207	72	64	14	51	5%	269	28%	10.27	1.95	1.09	0.88	26	76%	8	14.88
Julio Teheran	33	221	3271	14	884	188	82	71	22	51	6%	186	21%	7.57	2.08	1.08	0.85	25	76%	11	14.80
Aaron Harang	33	204.1	3394	12	876	215	88	81	15	71	8%	161	18%	7.09	3.13	1.4	1.05	25	76%	13	16.63
Jordan Zimmermann	32	199.2	2924	14	800	185	67	59	13	29	4%	182	23%	8.2	1.31	1.07	0.93	24	75%	10	14.68
Yordano Ventura	30	181.1	2959	13	776	167	70	65	14	69	9%	156	20%	7.74	3.42	1.3	0.92	22	73%	9	16.34
Garrett Richards	26	168.2	2627	13	678	124	51	49	5	51	8%	164	24%	8.75	2.72	1.04	0.74	19	73%	6	15.62
Hyun-Jin Ryu	26	152	2443	14	631	152	60	57	8	29	5%	139	22%	8.23	1.72	1.19	1.00	19	73%	5	16.07
Lance Lynn	33	203.2	3450	15	866	185	72	62	13	72	8%	181	21%	8	3.18	1.26	0.91	24	73%	9	16.98
Dallas Keuchel	29	200	3020	12	808	187	71	65	11	48	6%	146	18%	6.57	2.16	1.18	0.94	21	72%	9	15.10
Doug Fister	25	164	2468	16	662	153	52	44	18	24	4%	98	15%	5.38	1.32	1.08	0.93	18	72%	2	15.05
Jake Arrieta	25	156.2	2416	10	614	114	46	44	5	41	7%	167	27%	9.59	2.36	0.99	0.73	18	72%	8	15.47
John Lackey	31	198	3078	14	833	206	94	84	24	47	6%	164	20%	7.45	2.14	1.28	1.04	22	71%	8	15.55
Tyson Ross	31	195.2	3119	13	811	165	75	61	13	72	9%	195	24%	8.97	3.31	1.21	0.85	22	71%	9	15.98
David Price	34	248.1	3730	15	1009	230	100	90	25	38	4%	271	27%	9.82	1.38	1.08	0.93	24	71%	9	15.03
Stephen Strasburg	34	215	3295	14	868	198	86	75	23	43	5%	242	28%	10.13	1.8	1.12	0.92	24	71%	10	15.33
James Shields	34	227	3632	14	939	224	95	81	23	44	5%	180	19%	7.14	1.74	1.18	0.99	24	71%	10	16.00
Jon Niese	30	187.2	2792	9	786	193	80	71	17	45	6%	138	18%	6.62	2.16	1.27	1.03	21	70%	12	14.91
Jeff Samardzija	33	219.2	3339	7	879	191	86	73	20	43	5%	202	23%	8.28	1.76	1.07	0.87	23	70%	16	15.23
Scott Feldman	29	180.1	2964	8	765	185	86	75	16	50	7%	107	14%	5.34	2.5	1.3	1.03	20	69%	12	16.46
Alfredo Simon	32	196.1	3014	15	818	181	80	75	22	56	7%	127	16%	5.82	2.57	1.21	0.92	22	69%	7	15.37
Wily Peralta	32	198.2	3192	15	838	198	88	78	23	61	7%	154	18%	6.98	2.76	1.3	1.00	22	69%	7	16.10
Rick Porcello	31	202.2	3019	15	828	208	88	77	18	39	5%	129	16%	5.73	1.73	1.22	1.03	21	68%	6	14.93
Tanner Roark	31	198.2	2999	17	798	178	64	63	16	39	5%	138	17%	6.25	1.77	1.09	0.90	21	68%	4	15.13
Kyle Lohse	31	198.1	3002	13	817	183	87	78	22	45	6%	141	17%	6.4	2.04	1.15	0.92	21	68%	8	15.15
R.A. Dickey	34	215.2	3513	14	914	191	101	89	26	74	8%	173	19%	7.22	3.09	1.23	0.89	23	68%	9	16.32
Max Scherzer	33	220.1	3638	18	904	196	80	77	18	63	7%	252	28%	10.29	2.57	1.18	0.89	22	67%	4	16.53
Hiroki Kuroda	32	199	3097	11	820	191	91	82	20	35	4%	146	18%	6.6	1.58	1.14	0.96	21	66%	10	15.56
Zack Greinke	32	202.1	3210	17	821	190	69	61	19	43	5%	207	25%	9.21	1.91	1.15	0.94	21	66%	4	15.88
Jose Quintana	32	200.1	3346	9	830	197	87	74	10	52	6%	178	21%	8	2.34	1.24	0.98	21	66%	12	16.72
Zack Wheeler	32	185.1	3308	11	794	167	84	73	14	79	10%	187	24%	9.08	3.84	1.33	0.90	21	66%	10	17.87
Jered Weaver	34	213.1	3352	18	888	193	87	85	27	65	7%	169	19%	7.13	2.74	1.21	0.91	22	65%	4	15.73
Bartolo Colon	31	202.1	3011	15	846	218	97	92	22	30	4%	151	18%	6.72	1.33	1.23	1.08	20	65%	5	14.90
Collin McHugh	25	154.2	2486	11	619	117	53	47	13	41	7%	157	25%	9.14	2.39	1.02	0.76	16	64%	5	16.12
Madison Bumgarner	33	217.1	3372	18	873	194	81	72	21	43	5%	219	25%	9.07	1.78	1.09	0.89	21	64%	3	15.53
Jarred Cosart	30	180.1	2947	13	766	173	80	74	9	73	10%	115	15%	5.74	3.64	1.36	0.96	19	63%	6	16.36
Matt Garza	27	163.1	2538	8	680	143	77	66	12	50	7%	126	19%	6.94	2.76	1.18	0.88	17	63%	9	15.56
Alex Cobb	27	166.1	2611	10	681	142	56	53	11	47	7%	149	22%	8.06	2.54	1.14	0.85	17	63%	7	15.72
Gio Gonzalez	27	158.2	2623	10	653	134	66	63	10	56	9%	162	25%	9.19	3.18	1.2	0.85	17	63%	7	16.58
Phil Hughes	32	209.2	3046	16	855	221	88	82	16	16	2%	186	22%	7.98	0.69	1.13	1.06	20	63%	4	14.56
Scott Kazmir	32	190.1	2983	15	777	171	81	75	16	50	6%	164	21%	7.75	2.36	1.16	0.90	20	63%	5	15.69
Jake Peavy	32	202.2	3225	7	852	196	91	84	23	63	7%	158	19%	7.02	2.8	1.28	0.97	20	63%	13	15.95
Chris Archer	32	194.2	3160	10	822	177	85	72	12	72	9%	173	21%	8	3.33	1.28	0.91	20	63%	10	16.27
Yovani Gallardo	32	192.1	3216	8	817	195	86	75	21	54	7%	146	18%	6.83	2.53	1.29	1.02	20	63%	12	16.74
John Danks	32	193.2	3298	11	855	205	106	102	25	74	9%	129	15%	5.99	3.44	1.44	1.06	20	63%	9	17.07
Chris Tillman	34	207.1	3411	13	871	189	83	77	21	66	8%	150	17%	6.51	2.86	1.23	0.91	21	62%	8	16.47
Miguel Gonzalez	26	157	2513	10	662	153	59	56	24	50	8%	110	17%	6.31	2.87	1.29	0.97	16	62%	6	16.01
Ervin Santana	31	196	2987	14	817	193	90	86	16	63	8%	179	22%	8.22	2.89	1.31	0.98	19	61%	5	15.24
Edinson Volquez	31	190.2	2949	13	802	165	75	65	17	71	9%	138	17%	6.51	3.35	1.24	0.87	19	61%	6	15.50
Hisashi Iwakuma	28	179	2542	15	709	167	70	70	20	21	3%	154	22%	7.74	1.06	1.05	0.93	17	61%	2	14.20
Mike Leake	33	214.1	3215	11	902	217	93	88	23	50	6%	164	18%	6.89	2.1	1.25	1.01	20	61%	9	15.02
Nathan Eovaldi	33	199.2	3198	6	854	223	107	97	14	43	5%	142	17%	6.4	1.94	1.33	1.12	20	61%	14	16.05
Jason Vargas	30	187	2611	11	790	197	82	77	19	41	5%	128	16%	6.16	1.97	1.27	1.05	18	60%	7	13.96
Henderson Alvarez	30	187	3003	12	772	198	65	55	14	33	4%	111	14%	5.34	1.59	1.24	1.06	18	60%	6	16.06
Mark Buehrle	32	202	3082	13	857	228	83	76	15	46	5%	119	14%	5.3	2.05	1.36	1.13	19	59%	6	15.26
Tom Koehler	32	191.1	2941	10	803	177	84	81	16	71	9%	153	19%	7.2	3.34	1.3	0.93	19	59%	9	15.39
Jason Hammel	29	173.1	2765	10	705	151	70	68	23	44	6%	154	22%	8	2.28	1.13	0.87	17	59%	7	15.97
Tim Hudson	31	189.1	2784	9	789	199	86	75	15	34	4%	120	15%	5.7	1.62	1.23	1.05	18	58%	9	14.72
Wade Miley	33	201.1	3217	8	866	207	103	97	23	75	9%	183	21%	8.18	3.35	1.4	1.03	19	58%	11	16.00
Ian Kennedy	33	201	3402	13	846	189	85	81	16	70	8%	207	24%	9.27	3.13	1.29	0.94	19	58%	6	16.93
Justin Verlander	32	206	3409	15	893	223	114	104	18	65	7%	159	18%	6.95	2.84	1.4	1.08	18	56%	3	16.55
Dan Haren	32	186	3096	13	776	183	101	83	27	36	5%	145	19%	7.02	1.74	1.18	0.98	18	56%	5	16.65
Hector Noesi	27	164.2	2587	8	694	166	88	81	27	54	8%	116	17%	6.34	2.95	1.34	1.01	15	56%	7	15.76
Chris Young	29	163	2696	12	682	143	70	67	26	60	9%	106	16%	5.85	3.31	1.25	0.88	16	55%	4	16.54
Jorge de la Rosa	32	184.1	3067	14	768	161	90	84	21	67	9%	139	18%	6.79	3.27	1.24	0.87	17	53%	3	16.66
Brandon McCarthy	32	200	3044	10	836	222	100	90	25	33	4%	175	21%	7.88	1.49	1.28	1.11	16	50%	6	15.22
Josh Collmenter	28	170.1	2595	10	683	155	73	67	17	35	5%	110	16%	5.81	1.85	1.12	0.91	14	50%	4	15.26
Jeremy Guthrie	32	202.2	3235	13	864	215	100	93	23	49	6%	124	14%	5.51	2.18	1.3	1.06	16	50%	3	16.00
Trevor Bauer	26	153	2591	5	663	151	76	71	16	60	9%	143	22%	8.41	3.53	1.38	0.99	13	50%	8	16.93
J.A. Happ	26	153	2600	11	647	154	75	70	21	46	7%	128	20%	7.53	2.71	1.31	1.01	13	50%	2	16.99
Kyle Gibson	31	179.1	2800	13	757	178	91	89	12	57	8%	107	14%	5.37	2.86	1.31	0.99	15	48%	2	15.63
Wei-Yin Chen	31	185.2	2977	16	772	193	77	73	23	35	5%	136	18%	6.59	1.7	1.23	1.04	15	48%	-1	16.07
C.J. Wilson	31	175.2	3108	13	761	169	95	88	17	85	11%	151	20%	7.74	4.35	1.45	0.96	15	48%	2	17.74
Francisco Liriano	29	162.1	2714	7	691	130	68	61	13	81	12%	175	25%	9.7	4.49	1.3	0.80	14	48%	7	16.74
Ryan Vogelsong	32	184.2	3058	8	780	178	86	82	18	58	7%	151	19%	7.36	2.83	1.28	0.97	15	47%	7	16.60
Clay Buchholz	28	170.1	2741	8	737	182	108	101	17	54	7%	132	18%	6.97	2.85	1.39	1.07	13	46%	5	16.11
Charlie Morton	26	157.1	2504	6	666	143	76	65	9	57	9%	126	19%	7.21	3.26	1.27	0.91	12	46%	6	15.94
Shelby Miller	31	182	2841	10	759	158	78	76	22	72	9%	127	17%	6.28	3.56	1.26	0.87	14	45%	4	15.61
Jake Odorizzi	31	168	3028	11	719	156	79	77	20	59	8%	174	24%	9.32	3.16	1.28	0.93	14	45%	3	18.02
A.J. Burnett	34	213.2	3472	8	935	205	122	109	20	96	10%	190	20%	8	4.04	1.41	0.96	15	44%	7	16.29
Eric Stults	32	176	2833	8	763	197	93	84	26	45	6%	111	15%	5.68	2.3	1.38	1.12	14	44%	6	16.10
Travis Wood	31	173.2	3045	8	781	190	110	97	20	76	10%	146	19%	7.57	3.94	1.53	1.10	13	42%	5	17.58
Kyle Kendrick	32	199	3102	10	865	214	108	102	25	57	7%	121	14%	5.47	2.58	1.36	1.08	13	41%	3	15.59
Bud Norris	28	165.1	2746	15	687	149	68	67	20	52	8%	139	20%	7.57	2.83	1.22	0.90	11	39%	-4	16.63
Roberto Hernandez	29	162.2	2701	8	713	154	84	75	19	72	10%	102	14%	5.64	3.98	1.39	0.95	11	38%	3	16.65
Drew Hutchison	32	184.2	3051	11	786	173	92	92	23	60	8%	184	23%	8.97	2.92	1.26	0.94	12	38%	1	16.56
Ricky Nolasco	27	159	2641	6	695	203	96	95	22	38	5%	115	17%	6.51	2.15	1.52	1.28	10	37%	4	16.61
Vidal Nuno	28	157.1	2497	2	655	148	82	75	23	43	7%	124	19%	7.09	2.46	1.21	0.94	10	36%	8	15.89
Roenis Elias	29	163.2	2661	10	693	151	77	70	16	64	9%	143	21%	7.86	3.52	1.31	0.93	8	28%	-2	16.31
Colby Lewis	29	170.1	2802	10	762	211	107	98	25	48	6%	133	17%	7.03	2.54	1.52	1.24	8	28%	-2	16.47

I ran correlations with the percentage of quality starts across many variables, including things like Pace, line drive rate, ground ball rate, and fly ball rate. Because, why not? It’s fun to see what relationships develop even if you can explain them away. In fact, ground ball rate and fly ball rate possessed a significance at the .05 threshold, but that relationship didn’t really translate when developing a model later.

Below what you have is kind of the “no kidding” variables that would impact a quality start — highlighted by IP/GS, which, duh. A quality start and the number of innings you pitch ought to be related — that is they trend together, and if memory serves this is called multicollinearity and my old instructor of statistics actually liked to see that in correlations just to make sure your variables pass the smell test. If I had more statistical chops, I’d tease this point out — but suffice to say that multicollinearity in our correlations doesn’t do anything negative to the instructive nature of our correlations here, and certainly does not impact the model below because I cast them aside. So there.

		QS%	H/IP	WHIP	IP/GS	Pitches Per Inning	K%	BB%
QS%	Correlation	1	-.504**	-.659**	.744**	-.457**	.422**	-.329**
	Sig. (2-tailed)		.000	.000	.000	.000	.000	.001
	N	98	98	98	98	98	98	98
H/IP	Correlation	-.504**	1	.760**	-.347**	.205*	-.631**	-.165
	Sig. (2-tailed)	.000		.000	.000	.043	.000	.104
	N	98	98	98	98	98	98	98
WHIP	Correlation	-.659**	.760**	1	-.643**	.608**	-.583**	.515**
	Sig. (2-tailed)	.000	.000		.000	.000	.000	.000
	N	98	98	98	98	98	98	98
IP/GS	Correlation	.744**	-.347**	-.643**	1	-.669**	.404**	-.516**
	Sig. (2-tailed)	.000	.000	.000		.000	.000	.000
	N	98	98	98	98	98	98	98
Pitches Per Inning	Correlation	-.457**	.205*	.608**	-.669**	1	-.093	.646**
	Sig. (2-tailed)	.000	.043	.000	.000		.360	.000
	N	98	98	98	98	98	98	98
K%	Correlation	.422**	-.631**	-.583**	.404**	-.093	1	-.056
	Sig. (2-tailed)	.000	.000	.000	.000	.360		.587
	N	98	98	98	98	98	98	98
BB%	Correlation	-.329**	-.165	.515**	-.516**	.646**	-.056	1
	Sig. (2-tailed)	.001	.104	.000	.000	.000	.587
	N	98	98	98	98	98	98	98

**Correlation is significant at the 0.01 level (2-tailed)
*Correlation is significant at the 0.05 level (2-tailed)

So in looking over this set of data, I had a variety of thoughts, but allow me to condense for clarity: I wanted a simple model, something that could be actually usable. I didn’t want a model controlling for buckets of variables, and fortunately, we don’t have to. Strikeouts and walks are related to pitches per inning, walks very much so, and strikeouts in a negative way, which makes sense (although, no, not statistically significant — but it’s there). WHIP becomes particularly handy because by definition, it’s assisting us with walks and hits. So in an effort to paint with a broader brush, can we explain a decent degree of the variance in a quality start from just WHIP and average pitches per inning? Sort of.

Using basic multivariate regression with WHIP and PPI as our independent variables and with quality start percentage as our dependent variable, we get the following:

Model Summary
Model	R	R Square	Adjusted R Square	Std. Error of the Estimate
1	.663a	.439	.427	9.73458%

An adjusted R-squared of .427 ain’t bad — and as far as the model fit goes, again, this is where we get into what might pass the smell test. Frankly, if I wanted to simply drive up the R-squared then I’d use ERA as a dependent variable and that would yield something super impressive and perhaps there would be much rejoicing and we could all go get a beverage. But that’s like predicting cloud coverage in a rain storm, and really doesn’t help us. Because I can say, unequivocally, if you want quality starts pick guys who ought to have a low ERA. But if we want to drill down a little, well, how about efficiency.

Coefficients
Model		Unstandardized Coefficients		Standardized Coefficients	t	Sig.
		B	Std. Error	Beta
1	(Constant)	156.769	19.191		8.169	.000
	Pitches Per Inning	-1.400	1.506	-.090	-.929	.355
	WHIP	-59.783	9.582	-.604	-6.239	.000

Because they’re negative impacts on quality start percentage, we’re left with a bit of a wacky pair of coefficients working backwards towards the predicted quality start, but in this particular model, WHIP is statistically significant as is the constant, which is good news for the overall fit of the model since pitches per inning doesn’t quite meet the significance standard. My interpretation would be this may suggest some collinearity, but it doesn’t render the model as junk given the “goodness of fit” we find from the model summary. Plowing forward, our model would then look like this: xQS% = 156.769 + (-1.4)*PPI + (-59.783)*WHIP.

Therefore, if you wildly speculate a pitcher with 15 pitches per inning on average who is predicted to have a 1.15 WHIP, the model would predict a 68% quality start rate. Since those two explain just about half of the variance, you could conceivably target the arms projected to have low ERA’s (or FIP, if you prefer) to make up for some of the added variance and you might be on to something. But for the fun of it, applying the model to last year’s starters, we can plug in xQS and then see what the model thinks of their actual quality start percentage:

	GS	IP	Pitches	WHIP	QS	QS%	PPI	xQS	xQS-QS
Clayton Kershaw	27	198.1	2722	0.86	24	89%	13.74	86.1%	-3%
Johnny Cueto	34	243.2	3659	0.96	29	85%	15.05	78.3%	-7%
Jon Lester	32	219.2	3493	1.1	27	84%	15.94	68.7%	-16%
Cole Hamels	30	204.2	3136	1.15	25	83%	15.36	66.5%	-17%
Chris Sale	26	174	2753	0.97	21	81%	15.82	76.6%	-4%
Felix Hernandez	34	236	3434	0.92	27	79%	14.55	81.4%	2%
Alex Wood	24	156.1	2420	1.09	19	79%	15.50	69.9%	-9%
Sonny Gray	33	219	3295	1.19	26	79%	15.05	64.6%	-14%
Adam Wainwright	32	227	3258	1.03	25	78%	14.35	75.1%	-3%
Corey Kluber	34	235.2	3500	1.09	26	76%	14.88	70.8%	-6%
Julio Teheran	33	221	3271	1.08	25	76%	14.80	71.5%	-4%
Aaron Harang	33	204.1	3394	1.4	25	76%	16.63	49.8%	-26%
Jordan Zimmermann	32	199.2	2924	1.07	24	75%	14.68	72.3%	-3%
Yordano Ventura	30	181.1	2959	1.3	22	73%	16.34	56.2%	-17%
Garrett Richards	26	168.2	2627	1.04	19	73%	15.62	72.7%	0%
Hyun-Jin Ryu	26	152	2443	1.19	19	73%	16.07	63.1%	-10%
Lance Lynn	33	203.2	3450	1.26	24	73%	16.98	57.7%	-15%
Dallas Keuchel	29	200	3020	1.18	21	72%	15.10	65.1%	-7%
Doug Fister	25	164	2468	1.08	18	72%	15.05	71.1%	-1%
Jake Arrieta	25	156.2	2416	0.99	18	72%	15.47	75.9%	4%
John Lackey	31	198	3078	1.28	22	71%	15.55	58.5%	-12%
Tyson Ross	31	195.2	3119	1.21	22	71%	15.98	62.1%	-9%
David Price	34	248.1	3730	1.08	24	71%	15.03	71.2%	1%
Stephen Strasburg	34	215	3295	1.12	24	71%	15.33	68.4%	-2%
James Shields	34	227	3632	1.18	24	71%	16.00	63.8%	-7%
Jon Niese	30	187.2	2792	1.27	21	70%	14.91	60.0%	-10%
Jeff Samardzija	33	219.2	3339	1.07	23	70%	15.23	71.5%	2%
Scott Feldman	29	180.1	2964	1.3	20	69%	16.46	56.0%	-13%
Alfredo Simon	32	196.1	3014	1.21	22	69%	15.37	62.9%	-6%
Wily Peralta	32	198.2	3192	1.3	22	69%	16.10	56.5%	-12%
Rick Porcello	31	202.2	3019	1.22	21	68%	14.93	62.9%	-5%
Tanner Roark	31	198.2	2999	1.09	21	68%	15.13	70.4%	3%
Kyle Lohse	31	198.1	3002	1.15	21	68%	15.15	66.8%	-1%
R.A. Dickey	34	215.2	3513	1.23	23	68%	16.32	60.4%	-7%
Max Scherzer	33	220.1	3638	1.18	22	67%	16.53	63.1%	-4%
Hiroki Kuroda	32	199	3097	1.14	21	66%	15.56	66.8%	1%
Zack Greinke	32	202.1	3210	1.15	21	66%	15.88	65.8%	0%
Jose Quintana	32	200.1	3346	1.24	21	66%	16.72	59.2%	-6%
Zack Wheeler	32	185.1	3308	1.33	21	66%	17.87	52.2%	-13%
Jered Weaver	34	213.1	3352	1.21	22	65%	15.73	62.4%	-2%
Bartolo Colon	31	202.1	3011	1.23	20	65%	14.90	62.4%	-2%
Collin McHugh	25	154.2	2486	1.02	16	64%	16.12	73.2%	9%
Madison Bumgarner	33	217.1	3372	1.09	21	64%	15.53	69.9%	6%
Jarred Cosart	30	180.1	2947	1.36	19	63%	16.36	52.6%	-11%
Matt Garza	27	163.1	2538	1.18	17	63%	15.56	64.4%	1%
Alex Cobb	27	166.1	2611	1.14	17	63%	15.72	66.6%	4%
Gio Gonzalez	27	158.2	2623	1.2	17	63%	16.58	61.8%	-1%
Phil Hughes	32	209.2	3046	1.13	20	63%	14.56	68.8%	6%
Scott Kazmir	32	190.1	2983	1.16	20	63%	15.69	65.5%	3%
Jake Peavy	32	202.2	3225	1.28	20	63%	15.95	57.9%	-5%
Chris Archer	32	194.2	3160	1.28	20	63%	16.27	57.5%	-5%
Yovani Gallardo	32	192.1	3216	1.29	20	63%	16.74	56.2%	-6%
John Danks	32	193.2	3298	1.44	20	63%	17.07	46.8%	-16%
Chris Tillman	34	207.1	3411	1.23	21	62%	16.47	60.2%	-2%
Miguel Gonzalez	26	157	2513	1.29	16	62%	16.01	57.2%	-4%
Ervin Santana	31	196	2987	1.31	19	61%	15.24	57.1%	-4%
Edinson Volquez	31	190.2	2949	1.24	19	61%	15.50	60.9%	0%
Hisashi Iwakuma	28	179	2542	1.05	17	61%	14.20	74.1%	13%
Mike Leake	33	214.1	3215	1.25	20	61%	15.02	61.0%	0%
Nathan Eovaldi	33	199.2	3198	1.33	20	61%	16.05	54.8%	-6%
Jason Vargas	30	187	2611	1.27	18	60%	13.96	61.3%	1%
Henderson Alvarez	30	187	3003	1.24	18	60%	16.06	60.2%	0%
Mark Buehrle	32	202	3082	1.36	19	59%	15.26	54.1%	-5%
Tom Koehler	32	191.1	2941	1.3	19	59%	15.39	57.5%	-2%
Jason Hammel	29	173.1	2765	1.13	17	59%	15.97	66.9%	8%
Tim Hudson	31	189.1	2784	1.23	18	58%	14.72	62.6%	5%
Wade Miley	33	201.1	3217	1.4	19	58%	16.00	50.7%	-7%
Ian Kennedy	33	201	3402	1.29	19	58%	16.93	56.0%	-2%
Justin Verlander	32	206	3409	1.4	18	56%	16.55	49.9%	-6%
Dan Haren	32	186	3096	1.18	18	56%	16.65	62.9%	7%
Hector Noesi	27	164.2	2587	1.34	15	56%	15.76	54.6%	-1%
Chris Young	29	163	2696	1.25	16	55%	16.54	58.9%	4%
Jorge de la Rosa	32	184.1	3067	1.24	17	53%	16.66	59.3%	6%
Brandon McCarthy	32	200	3044	1.28	16	50%	15.22	58.9%	9%
Josh Collmenter	28	170.1	2595	1.12	14	50%	15.26	68.5%	18%
Jeremy Guthrie	32	202.2	3235	1.3	16	50%	16.00	56.7%	7%
Trevor Bauer	26	153	2591	1.38	13	50%	16.93	50.6%	1%
J.A. Happ	26	153	2600	1.31	13	50%	16.99	54.7%	5%
Kyle Gibson	31	179.1	2800	1.31	15	48%	15.63	56.6%	8%
Wei-Yin Chen	31	185.2	2977	1.23	15	48%	16.07	60.7%	12%
C.J. Wilson	31	175.2	3108	1.45	15	48%	17.74	45.2%	-3%
Francisco Liriano	29	162.1	2714	1.3	14	48%	16.74	55.6%	7%
Ryan Vogelsong	32	184.2	3058	1.28	15	47%	16.60	57.0%	10%
Clay Buchholz	28	170.1	2741	1.39	13	46%	16.11	51.1%	5%
Charlie Morton	26	157.1	2504	1.27	12	46%	15.94	58.5%	12%
Shelby Miller	31	182	2841	1.26	14	45%	15.61	59.6%	14%
Jake Odorizzi	31	168	3028	1.28	14	45%	18.02	55.0%	10%
A.J. Burnett	34	213.2	3472	1.41	15	44%	16.29	49.7%	6%
Eric Stults	32	176	2833	1.38	14	44%	16.10	51.7%	8%
Travis Wood	31	173.2	3045	1.53	13	42%	17.58	40.7%	-1%
Kyle Kendrick	32	199	3102	1.36	13	41%	15.59	53.6%	13%
Bud Norris	28	165.1	2746	1.22	11	39%	16.63	60.5%	21%
Roberto Hernandez	29	162.2	2701	1.39	11	38%	16.65	50.4%	12%
Drew Hutchison	32	184.2	3051	1.26	12	38%	16.56	58.3%	21%
Ricky Nolasco	27	159	2641	1.52	10	37%	16.61	42.6%	6%
Vidal Nuno	28	157.1	2497	1.21	10	36%	15.89	62.2%	26%
Roenis Elias	29	163.2	2661	1.31	8	28%	16.31	55.6%	28%
Colby Lewis	29	170.1	2802	1.52	8	28%	16.47	42.8%	15%

So it’s buying Clayton Kershaw (thank goodness) but not Aaron Harang. Further down the scale, it’s not surprising to find that pitchers who seemed particularly unlucky (like Nuno and Elias) in QS% are predicted to have higher figures, although not altogether inspiring percentages. Some exceptions are C.J. Wilson and Travis Wood, presumably nailed for their high WHIP and PPI. One name that jumped out here is Hisashi Iwakuma who registered a fairly respectable 61% QS but his xQS is over 74% which puts him near the top in overall xQS.

So if the question is can we predict quality starts, well that answer is no, not really. But can we come up with a simple framework using reasonable projections and historical data in order to get us about half way towards guessing who might rack up more quality starts? Yeah. We can.

(Note, based on a recommendation from a Twitter follower, they asked if team defense might impact a quality start and I thought that was a nice idea since defense might very well impact the number of pitches per inning a starter throws. But in running correlations against QS, WHIP and PPI, team defense is weakly correlated and not statistically significant. Correlations with QS was .093, WHIP was -1.05, and PPI was -.074.)