Eno Sarris MLB starting pitching ranks for 2023, with Stuff+ powered projections and spring updat

Publish date: 2024-06-04

We’ve been working, here in this space, on improving the Pitching+ model. Your comments, your reactions, and your questions have all shaped improvements on that model.

Thanks to your help, we’ll reach a couple of milestones today. The Pitching+ model is performing at its best when it comes to predictive strength, and we’re also going to offer projections that use the model to power perhaps the most predictive pitching projections on the market.

In this new update to the Pitching+ model, we’ve lumped cutters in with fastballs to create three mini-models within the overall Pitching+ — you’ll notice some improvement in Stuff+ for pitchers like Graham Ashcraft, Martín Pérez and Nestor Cortes as a result of this aspect of change. We’ve also trained the model on more 2022 data, and more data means better performance. Remember, this model includes data that only Hawkeye can provide, and that was put into place in 2020. This re-training helped some sinkers perform better, so you’ll see Logan Webb, Framber Valdez and Brady Singer improve some.

Pitching+ now comes online faster and is more predictive in small samples than strikeouts minus walks, the most powerful small-sample statistic based on results.

That’s great for in-season predictive power. More important right now, and for year-to-year fantasy work, though, is taking Pitching+ and incorporating it into a projection system that marries these process statistics with some regression and also some of what actually happened on the field. This projection system, created by Jordan Rosenblum and powered by Pitching+, also gets to incorporate things like park factors and aging to improve their predictive ability. Unfortunately, that means a little drop for Nick Lodolo, but it also helps us remember there are physical realities that impact a player’s performance beyond just the quality of their process.

Here’s a look into Rosenblum’s process in making these projections:

“I first built a traditional pitching projection model following the well-known Marcel approach to project the various pitching components (K%, BB%, BABIP, and barrels and home runs). This approach captures the last three years of major league performance, with more recent years weighted more heavily, and adds in the appropriate amount of regression to the mean for each component (rather than regressing each component the same amount, as Marcel does for simplicity). I park-adjusted the statistics using park factors from Baseball Savant. I then added in aging effects and major league equivalencies to capture minor league performance developed in my previous research.

The best traditional model of ERA I could build in terms of predictive accuracy was an xFIP-style model that captures projected K%, BB%, FB percentage, and barrel/FB rather than HR/FB. Whereas xFIP assumes every pitcher allows a league average number of home runs per fly ball, my model captures a pitcher’s regressed barrel per fly ball rate. Barrel/FB is a much stickier metric than HR/FB, requiring only about a third as much regression to the mean, so it is less beneficial to assume every pitcher allows the same league average rate. My xFIP-with-barrels model outperformed PCRA slightly in terms of accuracy of predicting ERA, and both my model and PCRA stood apart from projection models based on other ERA estimators, SIERA, xFIP, FIP, ERA, and kwERA.

With the traditional pitching projection model built, I then added in the Pitching+ metrics. Stuff+ is extremely sticky, almost as sticky as fastball velocity, so I just focused on the most recent year, whereas I considered the last three years of performance (weighted by recency) for Location+ and Pitching+ like I did for the traditional components. I tested out the three Pitching+ metrics in projecting each of the important component metrics and I kept them in the final projection model of a particular component if they helped maximize predictive accuracy. For instance, adding Stuff+ into my traditional model of BABIP led to a significant boost in predictive power: adding Stuff+ improved the correlation between the BABIP forecast and the observed outcome, and reduced the mean absolute error and root mean square error by about 3%.

The end result is a novel pitching forecast that accounts for both traditional results-based pitching components and the powerful suite of pitching+ metrics that capture the physical characteristics, locations, and counts of each pitch.”

That’s pretty cool. If you want the TL;DR version: these projections, using Pitching+ under the hood, are more predictive than models using other available ERA estimators. It beat other pitching projections on the market, too, but because it’s such a baby, and we have only three years of data, we might have to wait another year or two to be sure of those results.

The most controversial part might be the improvement in projecting batting average on balls in play. Baseball research has long held to the idea that pitchers don’t exert much control over the ball in play, and that premise powered great advancements like FIP (Fielding Independent Pitching) and FanGraphs WAR (which is built on FIP). It’s part of why the strikeout rate has surged across baseball. But our work shows that using Stuff+ and Location+ improves our ability to project a pitcher’s BABIP, and that inherently suggests pitchers do have some control over a ball in play, and that this control has something to do with the shapes and velocities of their pitches as well as their ability to put the ball in the right locations. That echoes some of the work that Alex Chamberlain has been pioneering over at FanGraphs, as well.

The end result is that we can hopefully provide the best pitching rankings on the internet, now with Pitching+ projected ERA (ppERA) and strikeout rate (ppK%), projected Stuff+, Location+, and Pitching+, as well as Jeff Zimmerman’s pitching injury percentiles, and some hand-adjusted innings projections that began with the estimations provided by FanGraphs’ depth charts. The full projections, as well as new model results, are here on the Google doc for the deep divers. If you’re looking at a pitcher who is going to change roles on that sheet, you can see the projected innings pitched per appearance as one of the columns — a pitcher going from one inning to five innings per appearance would expect their Stuff+ to go down five and a half points, and their ERA to change according.

Good luck drafting!

(These rankings were updated March 25 to reflect the newest injury news, as well as some Stuff+ information coming in from Spring Training. Spring Stuff+ is in the google doc.)

Eno	IP	ppERA	ppK%	ppStuff+	ppLocation+	ppPitching+	Injury Pct	NFC ADP	League	Team	spring Stuff
1	196	2.34	32.90%	126	103	109	5.6	15	NL	MIL	#N/A
2	199	2.65	34.30%	129	103	111	70.3	14	AL	NYY	132.1
3	140	1.76	39.00%	137	106	115	92.9	26	AL	TEX	#N/A
4	151	2.43	35.60%	134	102	111	64.3	22	NL	ATL	137.4
5	171	2.9	32.90%	125	98	106	70.3	7	AL	LAA	#N/A
6	216	2.87	25.80%	120	105	107	4.2	30	NL	MIA	118.6
7	181	2.89	29.80%	112	103	107	50.8	36	AL	TBR	122.5
8	182	2.97	29.70%	113	107	109	21.9	33	NL	MIL	#N/A
9	175	2.86	28.90%	118	105	108	97.5	33	NL	NYM	112.3
10	175	3.2	29.70%	106	103	105	54.7	37	NL	NYM	102.3
11	202	3.18	28.30%	107	108	108	13.4	30	NL	PHI	99.6
12	187	3.18	31.60%	124	98	106	3.1	51	AL	CHW	#N/A
13	194	3.12	25.70%	102	103	102	26.1	53	AL	SEA	#N/A
14	179	3.19	28.10%	110	105	108	39.5	51	AL	TOR	128.8
15	181	3.17	26.90%	111	106	109	94.3	51	NL	PHI	106.2
16	175	3.21	32.10%	116	99	104	36.3	49	AL	HOU	92.3
17	183	3.33	26.40%	107	104	106	30	70	NL	ARI	105.4
18	190	3.48	25.70%	110	107	109	0.3	56	NL	LAD	#N/A
19	185	3.31	27.30%	113	102	104	90.4	70	NL	SDP	#N/A
20	204	3.49	27.00%	98	102	103	8.8	50	AL	CLE	#N/A
21	156	3.18	31.20%	119	98	104	74.5	106	NL	SDP	#N/A
22	154	2.89	32.60%	114	103	106	89.7	75	AL	NYY	#N/A
23	195	3.44	25.00%	110	97	99	8.1	74	AL	HOU	135.3
24	146	3.44	27.60%	108	102	104	99.6	102	AL	NYY	111.9
25	186	3.33	23.30%	105	103	104	11.3	113	NL	SFG	104
26	186	3.62	22.70%	96	101	102	12	64	NL	ATL	89.6
27	181	3.57	24.50%	102	100	103	7.4	109	AL	SEA	#N/A
28	170	3.42	23.90%	99	106	104	44.8	95	AL	SEA	#N/A
29	160	3.28	24.60%	106	104	107	52.6	143	AL	TBR	119.9
30	199	3.63	24.60%	100	99	100	3.8	66	AL	TOR	86.8
31	190	3.58	27.60%	95	101	100	6.3	85	AL	SEA	#N/A
32	165	3.35	25.60%	103	101	103	15.5	89	NL	SDP	#N/A
33	144	3.36	27.10%	106	103	106	89.3	114	NL	LAD	#N/A
34	156	3.71	31.30%	128	100	106	78	100	NL	CIN	116.2
35	157	3.47	27.00%	107	102	103	24.3	127	AL	NYY	103
36	156	3.55	26.20%	93	104	104	68.9	138	AL	TBR	99.5
37	140	3.73	25.50%	103	102	101	99.2	115	AL	BOS	96.8
38	130	3.61	26.80%	104	100	104	91.5	149	NL	LAD	#N/A
39	156	3.72	#N/A	#N/A	#N/A	#N/A	#N/A	173	NL	NYM	110.9
40	149	3.75	26.50%	102	100	103	92.2	151	NL	ATL	104.5
41	174	3.76	25.60%	103	100	100	11.6	154	AL	HOU	105.8
42	175	3.87	25.70%	98	103	100	23.3	131	AL	MIN	100.1
43	175	3.78	24.20%	99	103	101	86.5	125	AL	CHW	107.4
44	191	3.8	27.10%	109	101	103	5.3	86	AL	CLE	#N/A
45	172	3.47	22.50%	100	103	100	61.1	164	NL	STL	96.6
46	181	3.91	22.70%	103	98	99	14.4	153	NL	ATL	#N/A
47	125	3.05	28.90%	114	100	102	#N/A	182	AL	BAL	119.6
48	125	2.19	37.10%	139	102	110	97.8	157	AL	TBR	#N/A
49	145	3.39	25.20%	108	106	109	94.3	225	AL	TEX	#N/A
50	128	3.41	29.10%	107	99	103	74.9	139	NL	MIL	#N/A
51	161	3.77	23.30%	96	104	102	66.4	159	NL	MIA	103.6
52	172	3.65	24.70%	100	101	100	84.8	198	AL	TEX	97.3
53	155	3.54	22.70%	94	102	101	94.3	204	NL	SFG	#N/A
54	145	3.62	24.70%	100	104	104	27.9	240	AL	MIN	96.4
55	170	3.91	22.70%	100	105	105	87.6	237	NL	CHC	#N/A
56	169	3.7	24.70%	101	101	102	31.8	233	NL	MIA	98.4
57	176	3.87	22.10%	98	102	101	79.8	162	AL	TOR	84.4
58	137	3.3	27.20%	112	100	100	53.3	198	AL	HOU	103.3
59	135	3.6	25.90%	113	100	104	#N/A	274	AL	NYY	111.1
60	155	4.54	24.90%	97	97	97	33.2	159	AL	LAA	#N/A
61	159	3.64	24.10%	108	98	101	61.4	297	AL	BAL	115.3
62	141	3.84	25.00%	103	97	97	62.8	140	NL	MIA	123.6
63	179	3.7	22.80%	98	101	99	43.8	183	AL	KCR	115.6
64	135	3.41	24.90%	104	100	102	78.7	201	AL	MIN	106.3
65	144	3.49	22.70%	98	107	102	75.9	260	AL	TBR	105.5
66	176	4	22.90%	101	103	103	1	239	AL	TOR	97.3
67	163	4.13	23.80%	95	98	98	17.6	193	AL	LAA	#N/A
68	181	4.01	21.10%	103	100	98	30.3	314	NL	PIT	110.8
69	130	3.87	26.90%	97	102	102	95.7	208	AL	TEX	90.1
70	171	3.71	21.90%	104	102	101	66.7	266	NL	CHC	#N/A
71	131	3.57	25.80%	103	104	106	59.7	292	AL	BOS	#N/A
72	160	4.15	28.00%	105	97	99	55.1	131	NL	CIN	84.6
73	140	4.12	#N/A	#N/A	#N/A	#N/A	#N/A	313	AL	MIN	106.6
74	130	3.89	23.20%	102	102	101	48.4	308	AL	CLE	#N/A
75	169	3.94	18.10%	91	106	101	81.2	225	NL	STL	89.4
76	142	3.96	21.90%	98	105	100	57.5	293	NL	SFG	#N/A
77	135	3.87	26.20%	104	95	95	71	213	NL	MIA	115.1
78	126	3.77	22.70%	92	101	99	85.5	349	NL	SFG	91.5
79	125	3.58	26.20%	106	98	102	95.4	330	AL	HOU	#N/A
80	196	4.07	20.70%	96	104	101	9.5	239	NL	ARI	100.8
81	178	4.01	25.10%	92	101	102	1.4	135	AL	CHW	#N/A
82	141	3.89	21.40%	81	103	98	52.2	302	AL	DET	84
83	141	4.02	21.30%	101	104	104	28.6	268	NL	CHC	#N/A
84	154	4.16	22.50%	89	102	99	72	282	NL	SFG	#N/A
85	155	4.01	22.50%	106	105	106	54.4	278	AL	HOU	#N/A
86	140	3.74	23.00%	98	99	99	80.9	224	NL	STL	98.3
87	165	4.08	22.40%	115	101	105	56.5	451	NL	ARI	104.5
88	145	3.98	22.30%	92	104	102	97.1	271	NL	NYM	86.2
89	142	4	23.70%	106	99	100	49.4	289	NL	PIT	99.6
90	155	3.82	23.50%	97	98	98	46.6	298	NL	CHC	#N/A
91	125	3.92	#N/A	#N/A	#N/A	#N/A	#N/A	416	AL	DET	100.4
92	138	3.95	26.60%	113	98	103	90.1	259	AL	CHW	#N/A
93	148	4.25	21.00%	102	103	103	80.5	365	AL	BOS	91.5
94	143	4.09	24.20%	102	98	101	74.2	225	NL	LAD	#N/A
95	134	4.12	21.50%	95	101	101	42.7	388	NL	SDP	#N/A
96	140	4.21	20.20%	109	100	103	73.1	446	NL	CIN	#N/A
97	120	3.53	22.90%	100	103	102	91.8	311	NL	STL	93.2
98	184	4.49	18.70%	93	106	101	48	256	AL	LAA	97.5
99	173	4.3	20.00%	87	101	97	9.1	349	NL	PHI	#N/A
100	121	4.05	22.00%	102	101	101	69.9	487	AL	NYY	110.3
101	171	4.19	24.80%	107	99	102	0.3	404	AL	BOS	115.1
102	149	4.29	22.00%	105	101	102	95.7	298	NL	PHI	93.1
103	135	3.73	25.30%	101	99	100	38.8	440	AL	OAK	#N/A
104	147	4.08							AL	OAK	104.1
105	135	4.36	#N/A	#N/A	#N/A	#N/A	#N/A	572	AL	OAK	94.3
106	157	4.78	18.50%	88	104	101	96.4	310	NL	LAD	#N/A
107	118	4.45	21.60%	100	98	98	67.4	400	NL	WSN	92.5
108	111	3.7	22.70%	106	104	108	78.4	595	AL	BAL	101.1
109	81	4.05	21.40%	115	97	102	63.6	475	NL	PIT	128.4
110	111	4.15	22.70%	91	104	101	31.4	319	AL	MIN	93.9
111	88	4	23.80%	92	96	94	50.5	475	NL	NYM	93
112	94	3.67	25.60%	104	101	102	73.4	473	NL	NYM	91.6
113	93	3.6	26.80%	102	98	100	75.2	252	NL	MIL	#N/A
114	64	3.33	29.30%	112	101	105	#N/A	543	AL	BAL	108.1
115	69	3.76	23.80%	91	102	101	48.7	484	AL	DET	#N/A
116	80	3.72	25.70%	106	97	100	#N/A	442	AL	BOS	#N/A
117	101	3.83	21.30%	0	0	0	#N/A	378	NL	ATL	#N/A
118	148	3.92	21.60%	94	101	98	#N/A	524	AL	DET	90.9
119	156	3.99	17.80%	88	101	97	83	631	AL	OAK	80.6
120	182	3.97	19.40%	94	101	98	4.5	496	AL	BAL	#N/A
121	159	4.08	21.60%	91	101	98	47.3	564	NL	PIT	95.5
122	113	4.12	22.80%	96	99	98	56.8	338	AL	BOS	97.3
123	154	4.35	23.10%	94	100	97	45.9	277	NL	MIL	#N/A
124	85	4.13	22.60%	102	101	105	42.4	680	AL	TBR	108.4
125	107	4.31	20.90%	94	100	97	25.4	402	NL	MIA	92.5
126	140	4.1	19.90%	95	100	98	93.2	526	AL	DET	99.2
127	115	4.15	25.50%	107	97	99	25.7	618	AL	TOR	97.9
128	181	4.68	15.60%	76	104	97	0	475	AL	OAK	#N/A
129	75	4.24	20.90%	93	101	97			AL	MIN	#N/A
130	171	4.25	19.00%	83	102	96	85.1	323	AL	TEX	#N/A
131	101	4.29	19.00%	98	98	98	20.8	744	AL	KCR	102.2
132	124	4.13	18.70%	96	101	98	59.3	493	AL	DET	93.2
133	169	4.49	18.60%	90	100	96	3.5	341	AL	CLE	89.8
134	172	4.56	23.70%	98	98	98	13.7	428	NL	WSN	67.6
135	161	4.21	19.90%	95	100	99	20.4	563	AL	BAL	96.1
136	75	4.03	24.40%	92	98	96	#N/A	578	NL	LAD	#N/A
137	131	4.32	23.50%	0	0	0	99.5	534	AL	BOS	101.5
138	71	4.34	23.60%	105	101	104	60.4	728	AL	BOS	#N/A
139	62	4.24	20.50%	92	102	101	59	713	AL	BAL	90.8
140	95	4.15	23.40%	104	98	99	68.1	608	NL	CHC	#N/A
141	55	3.81	25.30%	103	102	105	60	303	AL	NYY	#N/A
142	83	3.42	26.40%	119	101	104	#N/A	647	NL	SDP	#N/A
143	67	4.21	21.00%	108	86	101	98.5	685	AL	TBR	90.8
144	60	3.83	27.90%	102	99	99	76.3	462	AL	CLE	#N/A
145	135	3.59	20.20%	106	104	102	86.2	357	NL	STL	76.9
146	75	4.14	16.50%	92	101	97	#N/A	751	NL	STL	92.6
147	48	3.91	27.00%	0	0	0	#N/A	548	NL	SFG	#N/A
148	121	4.28	20.40%	92	104	100	7.7	662	NL	WSN	86.9
149	37	4.23	22.90%	100	99	101	#N/A	412	NL	ARI	105
150	134	4.41	15.10%	88	103	99	81.6	673	AL	KCR	104.4
151	110	4.02	21.80%	91	97	96	40.2	539	AL	OAK	88.7
152	194	4.42	20.20%	95	99	98	5.6	479	NL	COL	90
153	128	4.12	20.60%	96	102	101	85.8	561	NL	SFG	#N/A
154	106	4.29	20.90%	100	100	101	90.8	631	NL	PIT	87.1
155	126	4.49	20.80%	92	105	99	93.9	478	NL	CHC	85.4
156	62	4.2	20.20%	99	102	101	47.7	374	NL	ARI	100.2
157	131	4.59	15.90%	85	102	97	2.8	682	AL	KCR	78.4
158	166	4.67	17.50%	89	100	98	22.9	611	AL	KCR	#N/A
159	90	4.02	21.40%	87	98	95	33.2	609	AL	OAK	88.2
160	58	4.7	25.20%	102	93	96	37.8	606	NL	LAD	#N/A
161									NL	ATL
162	142	4.47	19.10%	87	99	96	49.1	672	AL	OAK	#N/A
163	153	4.18	20.90%	84	101	97	34.9	361	NL	NYM	90.1
164	75	4.55	17.90%	87	98	94	34.2	670	NL	STL	#N/A
165	98	4.66	21.20%	83	104	98	18	494	NL	PHI	81.8
166	111	4.31	23.00%	102	100	101	84.4	730	NL	PIT	#N/A
167	73	4.11	21.30%	101	102	102	58.3	676	NL	SFG	#N/A
168	38	4.45	19.10%	106	95	98	29.3	750	AL	KCR	117.9
169	176	4.76	15.50%	81	103	96	2.4	555	AL	SEA	87.9
170	55	4.32	21.70%	97	95	95	24.7	718	AL	TEX	102.2
171	38	4.17	20.80%	96	102	101	36.3	736	AL	MIN	#N/A
172	161	4.53	20.90%	90	102	100	10.2	419	AL	LAA	#N/A
173	#N/A	#N/A	#N/A	#N/A	#N/A	#N/A	#N/A	#N/A	AL	TOR	#N/A
174	159	5.01	15.80%	74	105	97	10.9	618	NL	CHC	#N/A
175	156	4.69	16.80%	81	103	97	96.8	512	NL	MIA	76.9
176	44	3.63	23.00%	104	103	107	82.3	639	AL	BAL	#N/A
177	101	4.23	18.30%	97	102	99	83.7	617	NL	MIL	#N/A
178	20	4.26	20.70%	92	96	95	36.3	746	NL	CHC	88.7
179	122	4.54	20.70%	91	104	101	87.2	461	NL	SDP	#N/A
180	155	4.63	18.90%	91	102	99	17.3	582	AL	CLE	#N/A
181	24	4.4	20.80%	102	93	93	#N/A	751	AL	KCR	#N/A
182	43	4.53	17.90%	85	100	98	#N/A	748	NL	STL	87.3
183	76	4.38	20.00%	95	100	100	35.3	745	AL	TOR	#N/A
184	88	4.5	20.10%	98	97	96	23.3	668	NL	PIT	98.7
185	105	4.98	20.00%	98	92	100	87.6	433	AL	CHW	#N/A
186	61	4.45	17.10%	88	104	99	10.2	#N/A	AL	DET	92.5
187	22	4.91	18.30%	89	96	95	33.9	#N/A	NL	SDP	103.1
188	30	4.44	21.30%	90	100	97	57.9	749	AL	TEX	#N/A
189	49	5.02	19.70%	0	0	0	#N/A	750	AL	NYY	114.9
190	142	4.91	16.90%	87	100	95	44.5	705	NL	ARI	#N/A
191	85	4.46	21.40%	99	97	96	98.2	679	NL	WSN	#N/A
192	69	4.62	19.50%	79	98	93	4.9	726	AL	TEX	#N/A
193	41	4.64	17.70%	86	100	96	15.9	750	NL	PIT	85.3
194	28	4.65	19.50%	91	98	96	28.2	751	AL	TEX	#N/A
195	97	4.55	14.30%	82	98	93	80.2	745	NL	STL	91
196	106	4.53	17.30%	91	100	97	41.3	743	NL	MIL	#N/A
197	100	4.36	18.10%	93	100	96	1.7	688	AL	SEA	#N/A
198	46	4.12	22.30%	0	0	0	#N/A	749	NL	NYM	#N/A
199	55	4.72	20.00%	99	96	95	64.6	750	NL	ARI	113.4
200	58	4.35	18.30%	96	101	101	67.1	746	AL	DET	115.8

(Top photo of Corbin Burnes: Benny Sieu / USA Today)

ncG1vNJzZmismJqutbTLnquim16YvK57k2tnb2hgZnxzfJFsZmlrX2eEcL%2FTmqmtoZ6cerG105yfoqaXYr%2BiusqsZKyZoqe2tHmRaWlsZw%3D%3D

VivaX

Eno Sarris MLB starting pitching ranks for 2023, with Stuff+ powered projections and spring updat