A week ago, when I was ticking off the ways in which baseball experts used to assure us that baseball players could be expected to perform better at some times than others, I overlooked one of the big ones:  Hot streaks and slumps.

            If a starting pitcher has been pitching well in his recent starts, is he more likely to pitch well today?   It is hard to believe that he isn’t, isn’t it?   If a pitcher has pitched shutout baseball in three of his last four starts, don’t you have to believe that he’s more likely to pitch well than if he’s been getting beat up?

            A pitcher being “hot” is a little different than a hitter being hot.   A hitter being hot. . .what is that?   Self-confidence?   Karma?   Extra batting practice?   A new titanium necklace?  A pitcher can be hot or cold for meaningful and apparent reasons—minor injuries, arm fatigue, a new pitch that’s working for him, that the league hasn’t figured out yet.

            I have done three new studies to try to figure out whether a “hot” starting pitcher is more likely to pitch well, in his next start, than a pitcher who has not been pitching well.  In the first of those studies, what we will call the Bob Gibson Study, I studied the performance of “hot” pitchers versus the performance of equally good pitchers who were not hot.   In the second study, which we will call the A. J. Burnett/Josh Beckett study, I studied the question of whether there was any tendency for good games and bad games to form clusters.

 

I.  The Bob Gibson Study

 

            I took all pitchers from the years 1960 to 1969, and divided them into two groups, twice.   The first division was among pitchers as to the quality of the season’s performance.  I ranked all pitchers by the Season’s Score, and then divided them into eight groups with (essentially) equal numbers of games started in each group.    There were 31,730 Games Started in major league baseball from 1960 to 1969, so that’s about 4,000 starts in each group.

            Then I divided the 4,000 starts in each set into eight groups, separated by how “hot” the pitcher was coming into the game.

            The highest quality pitchers, coming into the game red hot, would be group AA.   The highest quality pitchers, coming into the game having been pitching very badly, would be group AH.

            The lowest quality pitchers, coming into the game having been pitching very badly, would be group HH.   The lowest quality pitchers, coming into the game pitching very well, would be group HA.

            There were 64 groups of starting pitchers, with about 500 starts in each group.  

 

            This is coloratura, but you may be wondering who is an “A” quality starting pitcher, who is a “B”, etc.   The Season Score is based on the pitcher’s Wins, Losses, Innings Pitched, ERA, Strikeouts and Walks, and also on Saves if there are any. Here is a chart of randomly selected “A”, “B”, “C”, pitchers, etc., which should give you a sense of what kind of pitcher we are dealing with in each group.

 

		First	Last	Year	G	IP	W	L	WPct	SO	BB	ERA
A		Denny	McLain	1968	41	336	31	6	.838	280	63	1.96
		Jim	Kaat	1966	41	305	25	13	.658	205	55	2.75
		Tom	Seaver	1968	36	278	16	12	.571	205	48	2.20
		Vern	Law	1960	35	272	20	9	.690	120	40	3.08
		Steve	Barber	1963	39	259	20	13	.606	180	92	2.75
B		Don	Drysdale	1960	41	269	15	14	.517	246	72	2.84
		Don	Mossi	1961	35	240	15	7	.682	137	47	2.96
		Mickey	Lolich	1965	43	244	15	9	.625	226	72	3.43
		Larry	Jackson	1963	37	275	14	18	.438	153	54	2.55
		Lew	Burdette	1961	40	272	18	11	.621	92	33	4.00
C		Dave	McNally	1965	35	199	11	6	.647	116	73	2.85
		Fred	Newman	1964	32	190	13	10	.565	83	39	2.75
		Ray	Sadecki	1968	38	254	12	18	.400	206	70	2.91
		Hank	Aguirre	1965	32	208	14	10	.583	141	60	3.59
		Bill	Stafford	1962	35	213	14	9	.609	109	77	3.68
D		Jim	Merritt	1968	38	238	12	16	.429	181	52	3.25
		Jim	Coates	1961	43	141	11	5	.688	80	53	3.45
		Joe	Niekro	1967	36	170	10	7	.588	77	32	3.34
		Al	Downing	1965	35	212	12	14	.462	179	105	3.40
		Dick	Ellsworth	1964	37	257	14	18	.438	148	71	3.75
E		Chris	Short	1962	47	142	11	9	.550	91	56	3.42
		Ken	Johnson	1964	35	218	11	16	.407	117	44	3.63
		Frank	Baumann	1962	40	120	7	6	.538	55	36	3.38
		Ray	Sadecki	1963	36	193	10	10	.500	136	78	4.10
		Wade	Blasingame	1964	28	117	9	5	.643	70	51	4.23
F		Cal	Koonce	1965	38	173	7	9	.438	88	52	3.69
		Ray	Washburn	1965	28	119	9	11	.450	67	28	3.63
		Tom	Murphy	1969	36	216	10	16	.385	100	69	4.21
		Ernie	Broglio	1964	29	170	7	12	.368	82	56	3.82
		Eli	Grba	1962	40	176	8	9	.471	90	75	4.55
G		Barry	Latman	1966	31	103	2	7	.222	74	35	2.71
		Bill	Monbouquette	1966	30	103	7	8	.467	61	22	4.72
		Jim	Shellenback	1969	38	102	4	7	.364	57	52	3.88
		Lew Jr.	Krausse	1967	48	160	7	17	.292	96	67	4.28
		Jack	Fisher	1962	32	152	7	9	.438	81	56	5.09
H		Arnold	Earley	1963	53	116	3	7	.300	97	43	4.73
		Marcelino	Lopez	1967	8	27	1	2	.333	21	19	4.72
		Larry	Miller	1965	28	57	1	4	.200	36	25	5.05
		Jack	Kralick	1965	30	86	5	11	.313	34	21	4.92
		Mike	Marshall	1969	20	88	3	10	.231	47	35	5.13

 

 

 

            To determine how “Hot” a pitcher was coming into the game, I used a method which is conceptually simple, but which will entangle us in an undesirable number of details.   It was based on Game Scores.   Game Scores are a method that “score” each start by a starting pitcher essentially on a zero-to-one-hundred scale.  To convert this into a “Hot Pitcher Scale”, each pitcher’s score after each game (and thus, heading into his next start) was 20% of his score from his last start, plus 80% of whatever his score was prior to his last start.

 

            For illustration, Bob Gibson, heading into his start of June 6, 1968, had a “Heat Index” of 69.4, which means that he had been pitching very, very well, indeed up to that time.  On June 6 he pitched a 3-hit shutout, Game Score of 84, which increased his Heat Index to 72.3--.80 times 69.4, plus .20 times 84.    On June 11 he pitched a 5-hit shutout, striking out 4, for a Game Score of 79, increasing his Heat Index to 73.6.   On June 15 he pitched a 4-hit shutout, striking out 13 batters and walking no one, for a Game Score of 92.   This increased his Heat Index to 77.3   On June 20 he pitched a 5-hit shutout, striking out 6, Game Score of 82.   This increased his Heat Index to 78.3   On June 26 he pitched a 4-hit shutout, striking out 7, Game Score of 86.   This increased his Heat Index to 79.3. 

            On July 1, 1968, he pitched a complete game, but gave up 9 hits and a run, for a Game Score of “just” 67.   This knocked his Heat Index down to 77.2, but he followed that up on July 6 with a 6-hit shutout, striking out 9, Game Score of 80, which pushed him back up to 77.8.   His Game Scores in his next four starts were 85, 86, 82 and 80 (four complete games in there, with a total of two runs allowed), which pushed his Heat Index up to 80.7.

            On August 4 he pitched his worst games in months.   He pitched 11 innings and struck out ten batters, but he did give up 5 runs, 4 of them earned, Game Score of 62.   This cut his Heat index down to 77.0.   Then, however, he began to rebuild it, pitching two more shutouts in his next three starts, Game Scores of 84, 71 and 92, which put him at 79.9.

            On August 24 he struck out 15 batters but gave up 6 runs, Game Score 70, which cut him back down to 77.9.   On August 28, however, he pitched a 4-hit shutout, striking out 14 batters, Game Score 90.   This pushed him up to 80.3   On September 2 he pitched a ten-inning, 4-hit shutout, Game Score 89.   This put him at 82.1. 

 

            This frankly incredible run of games—12 shutouts in 19 starts—made Gibson at that moment, heading into his next start, the hottest pitcher of the 1960s, and probably one of the hottest pitchers in the history of baseball.    But the 1960s were a pitching-dominated era in which pitchers completed games, and there were many hot streaks by pitchers in the 1960s which were comparable to this—none equal, but many close.    The hottest starting pitchers of the 1960s, not making redundant mentions of Gibson or other pitchers on the same streak, were:

 

1.  Bob Gibson, September 6, 1968	82.1
2.  Juan Marichal, May 31, 1966	81.1
3.  Luis Tiant, July 7, 1968	80.9
4.  Larry Dierker, September 17, 1969	79.3
5.  Gaylord Perry, September 15, 1967	79.3
6.  Sandy Koufax, July 12, 1962	78.5
7.  Sandy Koufax, August 18, 1965	77.4
8.  Ray Culp, September 29, 1968	77.3
9.  Don Drysdale, June 8, 1968	76.8
10.  Sandy Koufax, June 14, 1966	73.5

 

            The dates given are their highest scores entering a game, thus their highest scores AFTER the previous game.   The coldest pitchers of the 1960s were:

 

1.  Lew Burdette, July 30, 1965	28.7
2.  Galen Cisco, September 21, 1962	30.6
3.  Craig Anderson, August 7, 1962	31.7

 

            The ugly and annoying details of this system have to do with the start of a player’s career, or the start of a season.   I started each pitcher out at 50.00 each season if

            a)  he made his first start of the season before May 1, and

            b)  had made a start in the previous season.

            If he had no major league starts the previous season, or if he made his first start of the season on May 1 or later, then he started the season at 40.00.   Trust me, there is research that justifies the distinction, and it doesn’t really matter after about eight starts; the “starting point” disappears pretty quickly, taking a 20% hit in each start.

            Also, to avoid getting non-representative scores for early-season, 1960, I included 1959 data into the study for the purpose of establishing the “Heat Indexes”, although I did not include the games played in 1959 in the study for purposes of output data.

 

 

            Anyway, at the conclusion of this I had 64 groups of pitchers, coded AA, AB, AC, AD, AE, AF, AG, AH, BA, BB. . ..HE, HF, HG, HH.    AA was high-quality pitchers who came into the game hot; HH was low-quality pitchers who came into the game pitching badly, even by their own standards.   We had about 500 starts in each group of games.   The essential question was whether and to what extent pitchers would pitch better, relative to the quality of their overall performance, when they were “hot” than when they were “cold”.

            They did not pitch better.   The extent to which they pitched better was “zero”, or, actually, negative.

            The top-end pitchers (the “A” quality pitchers) actually did pitch better when they were “hot” than when they were “cold”.   This group of pitchers made 486 starts when they were “hot” (Group AA), of which they won 260, lost 150 (.634 percentage), struck out 3,133 batters, walked only 893, and posted a 2.35 ERA.   When they were “cold” (Group AH), they made 506 starts, struck out only 2,557, walked 998 and posted a 2.63 ERA, although they did post an even better won-lost record when “cold” (AH) than when “hot” (AA).  They were 285-112 (.718) when they came into the outing cold. 

            This chart gives their average Game Score, by group:

 

 

 

 

            As you can see, their Game Scores were higher when they were “hot”. 

            This, however, was the exception within the study.  In all other groups, the pitchers pitched better when they came into the start “cold” than when they came in “hot”.

            There is a “natural” or “grouping” effect that would explain this phenomenon.  By grouping pitchers based on their end-of-season stats, we put them on a fixed-point course.  If they had pitched well in the past, they had to pitch less well in the future in order to reach their fixed point by the end of the season, and vice versa.

            I knew that this would occur, but it was my judgment that this effect was would be small enough that the tendency of hot pitchers to continue to pitch well—the tendency of starts to cluster—could fairly easily shine through it if that effect was of any significance.   Maybe I was right about that; maybe I was wrong.   In any case it didn’t.   Consistently, the pitchers in each group pitched better when they came into the game “cold” than when they came into the game “hot”.

            (The aberration in Group A could be explained by the fact that this was an open-ended group, a group with a fixed bottom but an open top, so that the difference between the top and bottom pitchers in the group was larger than in any of the other groups.)

 

 

II. The Josh Beckett/ A J Burnett Study

 

            Josh Beckett and A. J. Burnett were teammates and starred together on the Florida Marlins years ago, and for the last several years have pitched for rival teams in the American League East.   They have similar career records and had somewhat similar seasons in 2009, Beckett going 17-6 with a 3.86 ERA, Burnett going 13-9 with a 4.04 ERA.

            Their seasons, however, were very different in this respect:  that Beckett was obviously streaky during the 2009 season, while Burnett was up and down.  This chart gives the Game Scores for Beckett and Burnett start by start, underlined in red for hot or blue for cold:

 

 

A. J. BURNETT
			33 Starts
51	76	50	19	49	54	44	55	41	69
56	28	73	62	82	62	49	44	60	74
18	77	45	61	17	65	22	65	41	62
55	66	49

 

 

 

 

JOSH BECKETT
			32 Starts
76	43	47	14	21	44	50	61	74	71
76	76	29	84	71	48	57	88	58	58
68	74	68	25	37	41	49	55	58	63
46	39

 

 

            If you subtract the value of each Josh Beckett start from the value (Game Score) of the previous start, you get a total of 473.   If you do the same for Burnett, you get a total of 786. 

            We can extend that measurement another step.   If you subtract the Game Score for every Josh Beckett start from the Game Score for every other Josh Beckett start, you have 496 possible comparisons, which average 21.44.  But if you subtract only the consecutive starts, you get a lower figure, 15.26.   The second figure is lower because each start tends to be like the start before it.

            If you do the same for Burnett, you get an average of 19.66 for all matches, but 24.56 for the consecutive-game matches.   The consecutive-game average is higher because Burnett tended to follow a good start with a bad start.

            Beckett’s games formed clusters.   Burnett’s did not.   If starting pitchers in fact have any tendency to get “hot” and “cold”—as Beckett did—then their “consecutive game” average will be lower than their “random match” average (understanding that we are not actually dealing with randomly matched games, but with the average for all games that could be randomly matched.)  

If there is no tendency for pitchers to be hot or cold, then the averages will be the same.

            I figured these averages for all starting pitchers for the years 2000 to 2009.   Let’s deal only with the pitchers who made 25 or more starts, since that’s complicated enough.    The streakiest starting pitcher of those ten years was Dan Haren in 2007, with the A’s.   Haren finished the season 15-9 with a 3.07 ERA.    Through his first 14 starts, Haren posted a 1.58 ERA.    For his last 20 starts, his ERA was 4.22.   A hot period; a cold period.   These were the ten “streakiest” pitchers of the last decade, in single seasons:

 

 

 

 

            While these were the ten “most inconsistent” or “least streaky” pitchers:

 

 

 

 

            If the pitchers’ Game Scores have any tendency to form clusters, this method will result in a ratio less than 1.000.  If there is no such tendency, this method will result in a figure higher than 1.000 as often as a figure lower than 1.000, and the overall figure will be 1.000.

            OK, we come then to the question:  Is there, in general, any tendency for Game Scores to form clusters?

            None whatsoever.

 

            Well, OK, we can’t say that it is zero.   There may be some very small tendency for games to form clusters.   In the decade as a whole the average for comparisons of consecutive starts, based on a total of 45,588 observations, was 18.79.    The average for randomly matched starts by the same pitcher, based on a total of 581,121 observations, was 18.92.   The overall ratio isn’t 1.000; it’s 0.993. 

            Among pitchers with 25 starts or more, the observed effect was even smaller than that.   There were 983 pitchers in the years 2000 to 2009 who made 25 or more starts.   488 of them had “clustering ratios” less than 1.00, indicating some streakiness.   495 had clustering ratios higher than 1.00.    For these pitchers, the overall ratio was not 1-.993, but 1-.9992.    Whether this difference is statistically significant would not seem to matter since, even if it is statistically significant—which I doubt—the difference is so small that it could still easily be explained by some “loading” factor such as ERAs being higher in mid-summer.

 

 

III. The Matched Set Study

 

            My third study compared pitchers with identical or near-identical year-to-date records, but one of whom came into the start hotter than the other.   For example, John Maine after his start of September 5, 2007, was 14-9.   Roy Oswalt, after his start of September 6, 2008, was also 14-9.   Maine had made 28 starts; Oswalt had made 28 starts.   Maine had an ERA of 3.80; Oswalt, 3.72.   Maine had 146 strikeouts; Oswalt had 144.   There records were, for all practical purposes, the same.

            However, Oswalt was at that moment red hot.   In his last eleven starts, he was 8-1, and had cut his ERA during that stretch from 4.77 to 3.72.   In his last start (the September 6 start) he had pitched a 1-hit shutout.   In the start prior to that, he had pitched 8 and a third innings of shutout baseball.    In the start before that, he had given up one run in 7 innings.   Two starts before that, he had pitched 8 innings of one-hit, shutout baseball, striking out ten.   He was sizzling.

            Maine was not.   In his September 5 start he had given up 9 hits and 6 runs in four and a third innings.   Two starts prior he had given up 6 runs in five and a third innings.   Over his last eleven starts he was 4-5, and his ERA had gone up from 2.71 to 3.80.   He was struggling.   His overall record was the same as Oswalt’s, but his recent performance was very different. 

            From the years 2000 to 2009, I identified 504 “matched sets” like this in which two starting pitchers had nearly-identical records, but one was hot and the other was not.   Details:

            In order to be considered a set, the pitchers had to have exactly the same won-lost records—same wins, same losses.

            The difference between them in ERA had to be no greater than 20 points (actually, no greater than 0.205). 

            The difference between them in game started could be no greater than two.

            The difference between them in strikeouts could be no greater than 10% (actually, 10% of the higher figure.   If one pitcher had 180 strikeouts and the other had 163, that would qualify because the difference is less than 10% of the higher figure.)

            Also, the starts had to occur at the same time of the season, with a difference of no greater than 10 as the “team game number” for the season.   (In other words, I didn’t want to compare a pitcher from September with a pitcher with an identical record in July, since this might introduce other issues into the comparison.)

            I didn’t use any comparisons in which pitchers had made fewer than ten starts.

            I figured how “hot” each pitcher was in the same way as the earlier study, except that I started everybody off at 50.00, rather than starting some guys off at 40.00 as in the previous study.   (Since every pitcher was at least ten starts into the season, the initial starting point is no longer of much relevance.   If I had started pitchers off at 75.00, I’d have gotten basically the same results.)   To qualify for the study, one of the matched-set pitchers had to have a “Hot Score” at least seven points higher than the other one, going into his next start. 

 

            Also, a pitcher couldn’t be matched against himself in the same year. . .in other words, Roy Oswalt as of September 6, 2008, couldn’t be matched against Roy Oswalt as of September 1, 2008.   Pitchers could, however, be matched against themselves in a different season, if they had near-identical records in different seasons, but were “hot” in one season and “not hot” in another.   This did happen numerous times—that pitchers wound up matched against themselves in other seasons.   For example:

 

            Randy Johnson as of September 5, 2000, had made 30 starts with a won-lost record of 17-6, 2.45 ERA, 299 strikeouts.

            Randy Johnson as of September 7, 2001, had also made 30 starts with the same won-lost record (17-6), same ERA (2.45), but 320 strikeouts.  But he was ten degrees hotter at that time in 2001 than he was in 2000.

 

            Greg Maddux as of September 7, 2001, was 17-8, 30 starts, 2.93 ERA, 162 strikeouts.  

            Greg Maddux as of September 13, 2000, was 17-8, 32 starts, 3.00 ERA, 168 strikeouts.   But he was eight degrees hotter in 2001 than he was in 2000.

 

            Russ Ortiz, as of August 14, 2001, had made 25 starts and was 13-6, 3.46 ERA, 116 strikeouts.  

            Russ Ortiz as of August 15, 2004, had made 25 starts and was 13-6, 3.31 ERA, 124 strikeouts.   But Ortiz was nine degrees hotter in 2004 than he was in 2001. 

 

            I should explain also that these records are the pitchers’ records as starting pitchers.   A handful of these pitchers may also have made a relief appearance or two, creating some small differences in ERA or occasionally even a win or a loss.   Anyway, here are 38 more randomly selected matches:

 

 

Month	Day	Year	First	Last	GS	IP	W	L	WPct	SO	BB	ERA
9	19	2001	Matt	Morris	31	197.7	20	7	.741	161	47	3.10	Hot
9	12	2008	Brandon	Webb	31	205.7	20	7	.741	168	59	3.28	Not
Month	Day	Year	First	Last	GS	IP	W	L	WPct	SO	BB	ERA
9	23	2000	Greg	Maddux	34	244.3	19	8	.704	183	40	2.91	Hot
9	24	2002	Roy	Oswalt	33	229.0	19	8	.704	202	61	2.95	Not
Month	Day	Year	First	Last	GS	IP	W	L	WPct	SO	BB	ERA
9	17	2005	Cliff	Lee	30	187.3	17	4	.810	136	49	3.75	Hot
9	10	2004	Mark	Mulder	29	210.0	17	4	.810	131	76	3.90	Not
Month	Day	Year	First	Last	GS	IP	W	L	WPct	SO	BB	ERA
9	17	2004	Roy	Oswalt	32	216.3	17	9	.654	189	57	3.49	Hot
9	30	2001	Tim	Hudson	34	229.0	17	9	.654	175	70	3.42	Not
Month	Day	Year	First	Last	GS	IP	W	L	WPct	SO	BB	ERA
9	25	2001	Wade	Miller	30	200.0	16	8	.667	168	70	3.56	Hot
9	15	2002	Matt	Morris	30	197.3	16	8	.667	163	62	3.42	Not
Month	Day	Year	First	Last	GS	IP	W	L	WPct	SO	BB	ERA
9	27	2005	Tim	Wakefield	32	220.3	16	11	.593	150	68	3.96	Hot
9	24	2000	Livan	Hernandez	32	232.0	16	11	.593	162	71	3.84	Not
Month	Day	Year	First	Last	GS	IP	W	L	WPct	SO	BB	ERA
9	17	2001	Robert	Person	30	190.3	15	6	.714	168	74	4.02	Hot
9	14	2009	Jered	Weaver	30	194.0	15	6	.714	164	61	3.85	Not
Month	Day	Year	First	Last	GS	IP	W	L	WPct	SO	BB	ERA
9	21	2008	Bronson	Arroyo	33	193.0	15	11	.577	158	66	4.66	Hot
9	28	2001	Ryan	Dempster	33	210.7	15	11	.577	170	106	4.66	Not
Month	Day	Year	First	Last	GS	IP	W	L	WPct	SO	BB	ERA
8	28	2004	Pedro	Martinez	27	180.7	14	5	.737	188	46	3.69	Hot
9	2	2009	Josh	Beckett	27	181.3	14	5	.737	172	48	3.87	Not
Month	Day	Year	First	Last	GS	IP	W	L	WPct	SO	BB	ERA
9	15	2006	Bronson	Arroyo	32	221.3	14	9	.609	172	55	3.17	Hot
9	18	2008	Cole	Hamels	32	220.3	14	9	.609	189	52	3.10	Not
Month	Day	Year	First	Last	GS	IP	W	L	WPct	SO	BB	ERA
7	26	2006	Justin	Verlander	20	130.3	13	4	.765	88	36	2.69	Hot
7	18	2002	Bartolo	Colon	20	146.3	13	4	.765	91	43	2.64	Not
Month	Day	Year	First	Last	GS	IP	W	L	WPct	SO	BB	ERA
8	22	2001	Javier	Vazquez	28	193.7	13	11	.542	183	41	3.76	Hot
8	28	2007	Daisuke	Matsuzaka	27	176.3	13	11	.542	174	66	3.88	Not
Month	Day	Year	First	Last	GS	IP	W	L	WPct	SO	BB	ERA
8	3	2009	Matt	Cain	22	148.0	12	3	.800	117	54	2.25	Hot
7	26	2007	Dan	Haren	22	149.0	12	3	.800	118	39	2.42	Not
Month	Day	Year	First	Last	GS	IP	W	L	WPct	SO	BB	ERA
8	27	2004	Derek	Lowe	26	151.7	12	10	.545	85	58	5.22	Hot
8	21	2001	Mike	Hampton	26	164.3	12	10	.545	91	66	5.26	Not
Month	Day	Year	First	Last	GS	IP	W	L	WPct	SO	BB	ERA
7	19	2001	Freddy	Garcia	20	135.3	11	2	.846	79	45	3.46	Hot
7	9	2006	Tom	Glavine	19	119.0	11	2	.846	82	36	3.48	Not
Month	Day	Year	First	Last	GS	IP	W	L	WPct	SO	BB	ERA
7	15	2006	Curt	Schilling	20	134.3	11	3	.786	124	16	3.42	Hot
7	15	2006	Mike	Mussina	20	128.3	11	3	.786	113	24	3.30	Not
Month	Day	Year	First	Last	GS	IP	W	L	WPct	SO	BB	ERA
9	8	2006	Livan	Hernandez	30	189.0	11	12	.478	115	65	5.10	Hot
9	3	2003	Jason	Jennings	30	170.3	11	12	.478	113	84	5.13	Not
Month	Day	Year	First	Last	GS	IP	W	L	WPct	SO	BB	ERA
6	29	2009	Roy	Halladay	15	109.0	10	2	.833	95	14	2.56	Hot
6	24	2004	Roger	Clemens	15	95.7	10	2	.833	101	38	2.73	Not
Month	Day	Year	First	Last	GS	IP	W	L	WPct	SO	BB	ERA
8	30	2000	Darren	Dreifort	26	154.7	10	7	.588	132	69	4.36	Hot
9	1	2000	Randy	Wolf	27	174.3	10	7	.588	136	70	4.34	Not
Month	Day	Year	First	Last	GS	IP	W	L	WPct	SO	BB	ERA
9	11	2004	Sidney	Ponson	29	189.7	10	14	.417	103	60	5.31	Hot
9	15	2006	Ramon	Ortiz	30	178.3	10	14	.417	97	62	5.30	Not
Month	Day	Year	First	Last	GS	IP	W	L	WPct	SO	BB	ERA
7	19	2002	Kyle	Lohse	20	110.7	9	5	.643	71	43	4.80	Hot
7	23	2004	Esteban	Loaiza	20	133.7	9	5	.643	78	43	4.85	Not
Month	Day	Year	First	Last	GS	IP	W	L	WPct	SO	BB	ERA
9	3	2008	Barry	Zito	28	153.7	9	16	.360	98	89	5.45	Hot
9	3	2004	Darrell	May	27	160.3	9	16	.360	97	49	5.61	Not
Month	Day	Year	First	Last	GS	IP	W	L	WPct	SO	BB	ERA
5	26	2001	Curt	Schilling	11	84.3	8	1	.889	93	11	2.77	Hot
5	26	2002	Randy	Johnson	11	79.0	8	1	.889	100	22	2.73	Not
Month	Day	Year	First	Last	GS	IP	W	L	WPct	SO	BB	ERA
8	6	2004	Mike	Maroth	23	149.7	8	7	.533	70	43	4.45	Hot
8	5	2007	Jason	Marquis	23	133.3	8	7	.533	73	55	4.39	Not
Month	Day	Year	First	Last	GS	IP	W	L	WPct	SO	BB	ERA
9	19	2001	Albie	Lopez	30	183.7	8	18	.308	118	70	5.10	Hot
9	27	2001	Bobby J.	Jones	32	192.3	8	18	.308	113	35	5.05	Not
Month	Day	Year	First	Last	GS	IP	W	L	WPct	SO	BB	ERA
6	22	2001	C.C.	Sabathia	14	76.0	7	2	.778	55	37	4.38	Hot
6	10	2008	Chien-Ming	Wang	14	90.0	7	2	.778	51	35	4.30	Not
Month	Day	Year	First	Last	GS	IP	W	L	WPct	SO	BB	ERA
6	15	2007	Derek	Lowe	15	102.3	7	6	.538	72	31	3.08	Hot
6	19	2002	Al	Leiter	15	93.7	7	6	.538	74	28	3.07	Not
Month	Day	Year	First	Last	GS	IP	W	L	WPct	SO	BB	ERA
8	19	2000	Steve	Parris	25	145.0	7	14	.333	86	50	4.78	Hot
8	23	2005	Brett	Tomko	24	145.7	7	14	.333	84	48	4.94	Not
Month	Day	Year	First	Last	GS	IP	W	L	WPct	SO	BB	ERA
5	30	2007	Chris	Young	11	67.0	6	3	.667	60	26	2.42	Hot
5	29	2001	Kevin	Brown	10	65.7	6	3	.667	64	17	2.60	Not
Month	Day	Year	First	Last	GS	IP	W	L	WPct	SO	BB	ERA
7	31	2001	Glendon	Rusch	22	119.3	6	6	.500	101	29	4.60	Hot
7	30	2005	Jose	Contreras	21	125.7	6	6	.500	92	56	4.58	Not
Month	Day	Year	First	Last	GS	IP	W	L	WPct	SO	BB	ERA
8	8	2008	Joe	Blanton	24	149.0	6	12	.333	76	43	4.71	Hot
8	10	2005	Brian	Lawrence	24	150.0	6	12	.333	79	38	4.80	Not
Month	Day	Year	First	Last	GS	IP	W	L	WPct	SO	BB	ERA
7	29	2003	Darrell	May	20	127.7	5	5	.500	65	34	3.81	Hot
8	4	2007	Sergio	Mitre	20	117.7	5	5	.500	67	26	3.67	Not
Month	Day	Year	First	Last	GS	IP	W	L	WPct	SO	BB	ERA
6	3	2000	Kenny	Rogers	11	78.7	5	5	.500	32	17	4.00	Hot
5	30	2003	Jeff	Suppan	11	67.0	5	5	.500	35	18	4.03	Not
Month	Day	Year	First	Last	GS	IP	W	L	WPct	SO	BB	ERA
8	22	2006	Jeff	Weaver	24	131.0	5	13	.278	83	35	6.11	Hot
8	22	2000	Masato	Yoshii	25	147.3	5	13	.278	78	44	5.99	Not
Month	Day	Year	First	Last	GS	IP	W	L	WPct	SO	BB	ERA
6	5	2001	Jarrod	Washburn	10	64.7	4	4	.500	39	26	4.45	Hot
5	27	2005	Tim	Wakefield	10	62.3	4	4	.500	39	30	4.48	Not
Month	Day	Year	First	Last	GS	IP	W	L	WPct	SO	BB	ERA
7	8	2005	Nate	Robertson	17	102.0	3	7	.300	59	38	3.35	Hot
7	3	2002	Chris	Reitsma	17	94.0	3	7	.300	55	32	3.45	Not
Month	Day	Year	First	Last	GS	IP	W	L	WPct	SO	BB	ERA
9	14	2008	Jeremy	Sowers	20	109.0	3	8	.273	60	34	5.70	Hot
9	6	2000	John	Snyder	20	110.7	3	8	.273	61	69	5.86	Not
Month	Day	Year	First	Last	GS	IP	W	L	WPct	SO	BB	ERA
6	12	2000	Frank	Castillo	12	65.0	2	5	.286	50	34	4.71	Hot
6	9	2002	Shawn	Estes	12	69.0	2	5	.286	52	35	4.83	Not

 

 

            Obviously, Jeremy Sowers in mid-September, 2008, wasn’t too hot, or he wouldn’t have been 3-8 with a 5.70 ERA, but he was hotter than Snyder.    Sowers’ ERA on July 19, 2008 was 6.44.    Snyder’s ERA had been 4.76 on July 8, but it had gone up to 5.86.   You get my point.   One was hotter than the other.

 

You may be wondering whether the 504 qualifying matches that I found are all there are.   The answer to that is “no”; you could find 2,000 matches from the years 2000 to 2009 as good as these, if you worked at it hard enough.   In this study, each group of 504 pitchers had an aggregate won-lost record of 4637-3779, and each group had an aggregate ERA of 4.12.

 

            Anyway, I then looked at how these pitchers performed in their next start, and also compared their final season stats.

 

            In this study the pitchers who were “hot” did out-perform the pitchers who were not hot in their next starts, and over the balance of the season—not by a huge amount, but they did outperform them.   The “hot” pitchers, in their 504 “next starts”, had a won-lost record of 199-175, an ERA of 4.28, and an average Game Score of 50.62.

            The “cold” pitchers, in their 504 next starts, had a won-lost record of 177-177, an ERA of 4.74, and an average Game Score of 47.94.

            At season’s end the “hot” pitchers had an average season score of 119.1.   The pitchers who were “cold” had an average season score of 112.8.

 

            Of course, I should point out. . ..I never know whether to point out things that any intelligent person could see for himself.   I always feel like I’m insulting your intelligence when I point out things like this, but when I don’t point them out, somebody always explains it to me as if I was the idiot who didn’t get it.

            I should point out that this difference, even assuming it to be statistically significant, is not necessarily evidence of a “hot streak/cold streak” phenomenon.   Players’ levels of ability do change over time.   The ten best pitchers in baseball today are not the same ten guys that you would have listed five years ago.   It may be that, when you measure equal performers but split according to recent performances, some of the pitchers who have pitched well recently have actually improved, while others have actually declined.

            In any case, suppose that you are going to a ballgame tomorrow, and both starting pitchers are 11-7 with ERAs of 3.45, but one of them is hot and the other is cold.   Is the one who is “hot” more likely to win the game?

            Yes.