Baseball for the Thinking Fan

Login | Register | Feedback

btf_logo
You are here > Home > Primate Studies > Discussion
Primate Studies
— Where BTF's Members Investigate the Grand Old Game

Saturday, December 27, 2008

Run Support for Top Starters, 1954-2008

UPDATED 12/30: Fixed a math error in the way that I calculated expected run support, which changed some of the below-expected support counts as well. I also went back and looked at Don Newcombe, but he had only 192 starts in my DB, so I chose (again) to leave him out.

Minimum 300 starts, includes only the starts for which I have event data:

Player		Starts	Supp	ExpSupp	BelowExp	%BelowExp
Schilling, C	436	4.69	4.94	273		62.6%
Brown, K	476	4.46	4.77	295		62.0%
Martinez, P	400	4.88	5.12	244		61.0%
Maddux, G	740	4.43	4.66	441		59.6%
Clemens, R	707	4.76	5.07	419		59.3%
Ryan, N		773	3.85	4.02	457		59.1%
Koufax, S	314	4.22	4.24	184		58.6%
Palmer, J	521	4.40	4.42	304		58.3%
Sutton, D	756	4.08	4.21	441		58.3%
Blyleven, B	685	4.19	4.39	399		58.2%
Ford, W		396	4.66	4.77	230		58.1%
Perry, G	690	3.89	4.09	401		58.1%
Bunning, J	516	4.20	4.28	299		57.9%
Seaver, T	647	3.92	3.89	374		57.8%
Drysdale, D	463	4.09	4.07	267		57.7%
Gibson, B	482	4.08	4.18	277		57.5%
Perry, J	447	4.21	4.37	257		57.5%
Smoltz, J	466	4.68	4.71	268		57.5%
Jenkins, F	594	4.48	4.44	338		56.9%
Mussina, M	536	5.34	5.35	305		56.9%
John, T		700	4.20	4.23	396		56.6%
Johnson, R	586	4.92	5.06	331		56.5%
Tiant, L	484	4.47	4.39	273		56.4%
Guidry, R	323	4.81	4.88	182		56.3%
Roberts, R	395	3.99	4.02	222		56.2%
Stieb, D	412	4.32	4.47	231		56.1%
Carlton, S	709	4.40	4.28	397		56.0%
Spahn, W	357	4.50	4.49	200		56.0%
Glavine, T	682	4.76	4.64	381		55.9%
Cone, D		419	4.83	4.90	234		55.8%
Morris, J	527	4.97	4.93	294		55.8%
Appier, K	402	4.71	4.68	223		55.5%
Niekro, P	710	4.26	4.22	391		55.1%
Hunter, J	476	4.38	4.17	260		54.6%
Kaat, J		625	4.43	4.42	341		54.6%
Marichal, J	457	4.63	4.35	249		54.5%
Gooden, D	410	4.95	4.68	222		54.1%
Saberhagen, B	371	4.58	4.51	195		52.6%

Most pitchers will have more than 50% of their starts with below-average support because they will sometimes be supported by oodles and scads of runs, which drives the average up. A starter who receives expected run support overall will usually have somewhere around 57% of his games below expectations.

Run support, as I calculate it here, is based only on actual innings batted in support of the starter. If a visiting team starter goes exactly six innings, he gets credit for the runs his team scores in innings one through seven. If his team scores 3 runs in those seven innings, then his team gets credit for 3*9/7, or 3.86 RS/9. Expected support is calculated the same way, based on how many runs per nine innings with which the team supported all of their starters. I don’t count runs scored, and innings batted, in support of relievers.

You don’t normally think of guys like Schilling and Pedro as being undersupported by their mates, do you?

The 50s Braves gave their top starters unusually good support. Burdette, who isn’t included here, was the only pitcher with 300 or more starts who got less-than-expected run support in fewer than 50% of his starts, and Bob Buhl also was pretty far down on the list.

Koufax netted out to being above-average in terms of support even though he had a high number of games below expectations because he had a good number of games in which he got a LOT of runs. Saberhagen netted out as below average because he didn’t have a lot of those high-support games; when he did get better than average support it wasn’t usually by much.

What I’m going to post next is a look at how these pitchers did in those games where they received less-than-expected run support.

 

Mike Emeigh Posted: December 27, 2008 at 02:08 AM | 18 comment(s) Login to Bookmark
  Related News:

Reader Comments and Retorts

Go to end of page

Statements posted here are those of our readers and do not represent the BaseballThinkFactory. Names are provided by the poster and are not verified. We ask that posters follow our submission policy. Please report any inappropriate comments.

   1. Benny Distefano's Mitt Posted: December 27, 2008 at 04:05 AM (#3038781)
Fascinating stuff. Thanks.
   2. Dag Nabbit at ExactlyAsOld.com Posted: December 27, 2008 at 05:23 AM (#3038792)
Are you adjusting for park at all?

You don’t normally think of guys like Schilling and Pedro as being undersupported by their mates, do you?

I do, but then again I crunched the numbers for this years ago. Going purely by memory, prior to going to Boston, Curt Schilling's run support was about as poor as Dazzy Vance and Bobo Newsom. Vance had BY FAR the worst career run support by any HoF. His Boston years has caused his career numbers to rise considerably.

If anyone's curious, the ESPN Encyclopedia put together by Pete Palmer & Gary Gillette includes a stat for pitcher run support. It's centered at 100, with higher better and lower indicating worse than league average. The book is worth buying for that reason alone, IMHO.

Mike, I know it's out of your time frame, but Don Newcombe and Allie Reynolds would also score near Juan Marichal on this.

What I’m going to post next is a look at how these pitchers did in those games where they received less-than-expected run support.

Do you have Jim Perry and Frank Tanana. My hunch is that Perry would do pretty damn well by this approach and Tanana rather poor.
   3. Walt Davis Posted: December 27, 2008 at 05:34 AM (#3038795)
mean vs. median :-)
   4. Rough Carrigan Posted: December 27, 2008 at 03:01 PM (#3038843)
If you watched Pedro circa 2000, yeah, you thought of him as being spectacularly undersupported by his teammates. The guy had a 1.74 ERA in the year 2000, in the middle of the PED era, in the AL, in Fenway Park! And his record was "only" 18-6.
   5. Rough Carrigan Posted: December 27, 2008 at 03:04 PM (#3038846)
Are these #1 starters facing other #1 starters and that's the cause?
   6. Howie Menckel Posted: December 27, 2008 at 03:11 PM (#3038850)
At what pct difference do you consider the spread statistically signifgicant?
   7. Mike Emeigh Posted: December 29, 2008 at 06:06 PM (#3039669)
Are you adjusting for park at all?


Indirectly. I divide by home starts and road starts and look at expected support on that basis.

It's centered at 100, with higher better and lower indicating worse than league average.


I don't agree with figuring run support based on league context - I use team context. As I've said before, a pitcher's job IMO is to work with what his team gives him - what the other teams for their pitchers is immaterial. If Pedro plays on a team that scores 5 runs per 9 innings for their other starters, and he only gets 4.9 runs per 9 innings, he's getting less support than expected for his context - even if the other teams only give their starters 4.7 runs per 9 innings - and it's not appropriate in my view to punish him for playing on a good offensive team, or to reward a guy who is getting more support than expected for his context on a poor offensive team.

Mike, I know it's out of your time frame, but Don Newcombe and Allie Reynolds would also score near Juan Marichal on this.


I probably should have put Newcombe on this list. I've limited this to pitchers who have drawn some mentions in the HoM discussions (as Newcombe has) because I don't particularly care how a pitcher did relative to an average starter; all of these guys should be well above that. I'm interested in how they compare to each other. Newcombe, like Roberts and Spahn, will have a good portion of his career that is NOT included, obviously.

mean vs. median


I know what you are saying, but for most of these guys there is not normally of an impact to make a significant difference, because you can't score fractions of a run and the median and mean usually fall in the same space between, say, 3 runs in seven innings and 4 runs in seven innings.

Are these #1 starters facing other #1 starters and that's the cause?


Not normally. The starting pitching distribution gets skewed pretty quickly after the first couple of weeks of the season; there's no strong tendency for #1 starters to be matched against each other.

At what pct difference do you consider the spread statistically signifgicant?


Not sure what you are asking here. I'm trying to separate things into "above expected" and "below expected". I expect (and it's normally true) that just about any pitcher will win games when he's getting above expected run support - so that the true separation between the great pitcher and the less-great is what they do when they don't get good run support.

-- MWE
   8. Dag Nabbit at ExactlyAsOld.com Posted: December 29, 2008 at 06:28 PM (#3039695)
I don't agree with figuring run support based on league context - I use team context. As I've said before, a pitcher's job IMO is to work with what his team gives him - what the other teams for their pitchers is immaterial. If Pedro plays on a team that scores 5 runs per 9 innings for their other starters, and he only gets 4.9 runs per 9 innings, he's getting less support than expected for his context - even if the other teams only give their starters 4.7 runs per 9 innings - and it's not appropriate in my view to punish him for playing on a good offensive team, or to reward a guy who is getting more support than expected for his context on a poor offensive team.

Huh? It's not punishing anyone for anything. It's noting. If a guy's offense gives him 4.9 runs per game and league average is 4.7, then you're noting what his team gives him.
   9. Mike Emeigh Posted: December 29, 2008 at 06:36 PM (#3039706)
If a guy's offense gives him 4.9 runs per game and league average is 4.7, then you're noting what his team gives him.


But he's not getting above-average support in his team's context - and that's what I want to know. His context is his team - not the league.

-- MWE
   10. Dag Nabbit at ExactlyAsOld.com Posted: December 29, 2008 at 06:42 PM (#3039717)
Ah, I see. This most intersting case I know of in that regard is Marichal & Perry when they were teammates in San Fran.
   11. JPWF13 Posted: December 29, 2008 at 06:57 PM (#3039728)
Using a half-assed method based upon the RS figures given above, I determined what each pitcher's W-L record would have been with "expected" support, and converted that W-L record to Fibbonacci Win Points...

1 Roger Clemens 525.1
2 Greg Maddux 442.8
3 Randy Johnson 378.8
4 Nolan Ryan 321.9
5 Pedro Martinez 313.7
6 Tom Seaver 304.9
7 Robin Roberts 278.4
8 Curt Schilling 276
9 Mike Mussina 274.7
10 Gaylord Perry 272.8
11 Warren Spahn 267.2
12 Bert Blyleven 260.8
13 Don Sutton 253.3
14 Kevin Brown 247.7
15 Tom Glavine 247.3
16 Bob Gibson 241.3
17 Tommy John 207.3
18 Jim Palmer 206.1
19 Fergie Jenkins 204.5
20 Steve Carlton 199.3
21 Jim Kaat 196.9
22 Phil Niekro 191.7
23 Dave Stieb 188.2
24 Jack Morris 186.2
25 David Cone 178.2
26 Luis Tiant 174.3
27 Jim Bunning 174.2
28 Rick Reuschel 170.1
29 Kevin Appier 167.4
30 Sandy Koufax 161.2
31 Bret Saberhagen 149.4
32 John Smoltz 144.7
33 Don Drysdale 134.3
34 Juan Marichal 122.1
35 Catfish Hunter 101.8


You know, considering the extent of his run support Marichal really should have gone better than 243-142...
   12. JPWF13 Posted: December 29, 2008 at 07:11 PM (#3039739)
I don't agree with figuring run support based on league context - I use team context. As I've said before, a pitcher's job IMO is to work with what his team gives him


nevermind post 11

this is the same mistake Bill James made when he first worked on offensive winning percentage

League context is a hell of a lot more important than team context.

You are assuming that pitchers can and do pitch to the score
until that is shown, your statement
- I use team context. As I've said before, a pitcher's job IMO is to work with what his team gives him - what the other teams for their pitchers is immaterial.

is completely useless.

If someone gets 4 runs from a team that usually scores 3.8, he is NOT better supported than a guy who gets 4.2 runs from a team that usually scores 4.4- but your method says/assumes he is.

Case in point- Jack Morris, expected support of 4.8? For the times he played????? No you are just giving him credit for the luck of playing almost exclusively for good offensive teams- and using it as the denominator understates how extraordinarily lucky he was in the run support he received- compared to his contemporaries.
Tom Seaver- received 3.91 runs compared to 3.98 expected? One, just ONE of his Mets teams ever cleared a 100 OPS+, but most easily cleared a 100 ERA+- his "expected runs" is artificially low- he played for good pitching/poor hitting teams for more than half his career.

But he's not getting above-average support in his team's context - and that's what I want to know. His context is his team - not the league.

His context is the league. I want to know how good a pitcher IS, I want to know how well he would do compared to pitchers in the league, not just how well he does compared to his staff mates.
   13. Wes Parkers Mood (Mike Green) Posted: January 09, 2009 at 01:54 AM (#3047160)
This way of looking at expected support can distort value, for example, when comparing Don Sutton with Bert Blyleven. Sutton may have received roughly the same rate of support relative to team that Blyleven did, but his teams were much better.

To get a handle on a starting pitcher's win efficiency while properly factoring out offensive and bullpen support, the easiest place to start is to look at performance by leverage. Blyleven, like almost every other great starter, performed better in low leverage situations than in high leverage ones. The exceptions to this rule from the retrosheet age are Koufax, Seaver and Clemens. Blyleven's low leverage delta was more than typical for a great pitcher, but much less than Sutton's (for instance).
   14. Sunday silence: Play Guess How long season lasts Posted: January 21, 2009 at 03:42 AM (#3056638)
What sort of conclusions would this method show or suggest that cant be obtained using ERA+ or some other similar measure?

I take it this method does not account for park effects, which I would imagine this issue would be even more problematical owing to a smaller sample size vice the entire season.
   15. Barca Posted: January 21, 2009 at 11:37 PM (#3057526)
"I don't agree with figuring run support based on league context - I use team context. As I've said before, a pitcher's job IMO is to work with what his team gives him - what the other teams for their pitchers is immaterial. If Pedro plays on a team that scores 5 runs per 9 innings for their other starters, and he only gets 4.9 runs per 9 innings, he's getting less support than expected for his context - even if the other teams only give their starters 4.7 runs per 9 innings - and it's not appropriate in my view to punish him for playing on a good offensive team, or to reward a guy who is getting more support than expected for his context on a poor offensive team."

I would have liked to see how pitchers compared with pitchers from different teams, even from different eras. I think it should be ball park adjusted, but pitchers with good hitting teams shouldn't be rewarded. Also, Mike Cuellar shouldn't be penalized because Jim Palmer is on the team.
   16. Sunday silence: Play Guess How long season lasts Posted: January 24, 2009 at 05:22 AM (#3059549)
As I've said before, a pitcher's job IMO is to work with what his team gives him - what the other teams for their pitchers is immaterial. If Pedro plays on a team that scores 5 runs per 9 innings for their other starters, and he only gets 4.9 runs per 9 innings....


If this argument has any vitality then shouldnt we be thinking of a pitcher's ERA w/ respect to his team and not the league? That seems odd, as it is nearly always thought of as a statistic w/ respect to the league and not the team. What am I missing here?
   17. Paul Wendt Posted: January 25, 2009 at 05:11 PM (#3059986)
Mike Emeigh quoted Walt Davis,
> mean vs. median

I know what you are saying, but for most of these guys there is not normally of an impact to make a significant difference, because you can't score fractions of a run and the median and mean usually fall in the same space between, say, 3 runs in seven innings and 4 runs in seven innings.


I don't know what that is saying but I suspect it is merely a label for Mike's explanation, "because they will sometimes be supported by oodles and scads of runs, which drives the average up."

--
OCF
>>At what pct difference do you consider the spread statistically significant?
<<

Mike Emeigh
>Not sure what you are asking here. I'm trying to separate things into "above expected" and "below expected". I expect (and it's normally true) that just about any pitcher will win games when he's getting above expected run support - so that the true separation between the great pitcher and the less-great is what they do when they don't get good run support.
<

Mike's purpose here seems limited, not to rely on the frequency of below-average support (the last column), so it doesn't need much interpretation or analysis. Evidently the purpose is to set the stage for part two. The last column states the share of career starts that will be featured in part two. The point will concern exclusively his work as a pitcher, perhaps pitching and fielding but not batting and baserunning, so it doesn't matter whether the pitcher contributed a lot or a little to his run support.

OCF hopes to interpret the last column, ?lowExp, and probably too the difference between pitcher mean and team mean, Supp - ExpSupp. For example, what do those statistics say about whether a pitcher was "lucky"? Some other comments show great interest in that.

The interpretation or further analysis of these statistics must account for the pitcher's contribution to run support. Maybe we can say whether a the size or frequency of a pitcher's above-benchmark run support was statistically significant after accounting for his own batting/baserunning in the benchmark.

There may be some useful or interesting "quick and dirty" calculation given two more numerical variables, a pitcher's OPS+ as a batter and his share of career starts in DH leagues.
   18. Paul Wendt Posted: January 25, 2009 at 05:48 PM (#3059996)
Because this project uses team benchmarks, any quick and dirty adjustment based on pitcher OPS+ is sure to be very "dirty". If Glavine and Smoltz were good batters, and Maddux above average (all while they were teammates and compared to mlb average), then above average pitcher batting is "expected" for them here.

--
Mike Emeigh, from the preface,
Koufax netted out to being above-average in terms of support even though he had a high number of games below expectations because he had a good number of games in which he got a LOT of runs. Saberhagen netted out as below average because he didn’t have a lot of those high-support games; when he did get better than average support it wasn’t usually by much.

The table shows that Koufax and Saberhagen (my emphasis) enjoy relatively high mean support and relatively low mean support in comparison with most of their neighbors in this layout. That is, Koufax high and Saberhagen low team-relative mean support relative to their team-relative median support. Neither one "reverses the sign" Supp - ExpSupp in the way that Mike's remark means to me.

Reading the table from top to bottom, from the superficially unlucky to lucky,
the pitchers who enjoyed better than team-average run support Supp > ExpSupp are none from the top third; Seaver, Drysdale, Jenkins, Tiant from the middle; Carlton and everyone below him except Cone.
From the top and bottom thirds of the table, only Cone reverses the sign.

Everyone from Koufax or Palmer to Appier or Niekro seems to be in the unremarkable middle of the lot by frequency (the last column). Essentially that is what OCF hopes to see quantified statistically, whose run support (?lowExp) is statistically unremarkable?

You must be Registered and Logged In to post comments.

 

 

<< Back to main

BBTF Partner

Dynasty League Baseball

Support BBTF

donate

Thanks to
HowardMegdal
for his generous support.

Bookmarks

You must be logged in to view your Bookmarks.

Syndicate

Page rendered in 0.3640 seconds
40 querie(s) executed