Revised prelim after the new DanR WARP data

I'm still trying to sort through the data but many of my rankings went back to where they were in the first place - ie made more sense. John McGraw is saying goodbye to my ballot though he's not as far off.

1) Boggs
2) Luis Tiant
3) Bob Johnson
4) Tommy Bridges
5) Bus Clarkson
6) Graig Nettles (DH penalty goes away)
7) Rick Reuschel
8) Reggie Smith (drops out of top 3)
9) Norm Cash
10) Tommy Leach
11) Virgil Trucks (pitchers moved up in tandem, I may bump them down a little but I like how my PHoM rankings moved
12) Bret Saberhagen
13) Ben Taylor
14) Ron Cey
15) Gavy Cravath

16-20) Dick Redding, Vic Willis, Jim McCormick, Buddy Bell, Alejandro Oms
21-25) Urban Shocker, Dutch Leonard, Lee Smith, John McGraw, Jack Quinn
26-30) Bob Elliott, Dizzy Trout, Tommy John, Bobby Bonds, Frank Tanana

Dawson moves up to 39, Richie Ashburn makes PHoM instead of Vic Willis or Charley Jones

Bump

'Cause win shares is fibbing.

Everybody here knows I've criticized Win Shares for many things over the years.
However, one thing I do like about it is that it divides up the Wins.
(My criticisms are usually about How it divides up those wins.)

Someone earlier on this thread (or the Duffy thread) brought up a comparison between Thompson and Duffy. Let's look at their teams from 1892-1894. The teams played about 410-415 games during those 3 years. By Davenport's WARP numbers, which are built from the playing statistics, Philly was better (further above replacement) than Boston each year by an aggregate total of about 12 games over that time. Yet on the playing field, Boston actually won 41 more games than Philly did.

This is a HUGE difference. A discrepancy of more than 50 wins between stats and reality. 20% of Boston's wins over those 3 seasons are essentially unaccounted for, cannot be explained by the statistics (at least given the current state of their analysis). These extra Wins also caused them to "steal" two pennants. Somebody did something right to get all those extra wins. (Slightly overstating that percent there because in truth Philly appears to have also underperformed by a small amount.)

The HOF Old-Timer's Committee of 1944-45, which consisted of Connie Mack and some other old guys who had actually seen those teams 50 years prior, saw fit to honor both of Boston's "Heavenly Twins", as the local writers had called them back in the 1890s. So there is some precedent for giving Duffy a significant share of that credit. (McCarthy probably would need almost all of it to get some peak votes from this group ;-) Win Shares inflates the victory credit of everybody on those Boston teams, proportionate to Win Shares' view of their statistics.

The criticism of WS divvying up the Wins may be appropriate for a one-year "fluke" kind of a situation. But this is many standard deviations away from the expected, and over an extended period of time. (50+ wins over three years, with about 10 Wins as one SD.)

even tho I've not been a fan of Duffy's nor of his WS-assigned totals, I agree with jimd's larger point.

Wrestling with modern pitcher candidates Reuschel and Saberhagen:

Can some of you post stats that show how their actual W-L records compare with other run-based stats? Pythag W-L record? Avg run-support based records? Rob Wood, you did a lot of work with pitcher records compared with actual game-by-game team run support; what've you got? These two are both looking like ballot borderliners, easily able to move 5 to 10 places, and Sabes might get strong support in his first year from some. Let's especially focus on what we can find on him, with all of the modern era data we have.

Tom: the following is my system. RA+ Pythpat, season-by-season. Arbitrarily set at 9.0 IP per decision, whatever the actual number of decisions. No adjustment for either defensive support or pitcher's own offense. Sorted by yearly equivalent FWP.

104. jimd Posted: September 13, 2007 at 07:36 PM (#2523603)
The criticism of WS divvying up the Wins may be appropriate for a one-year "fluke" kind of a situation. But this is many standard deviations away from the expected, and over an extended period of time. (50+ wins over three years, with about 10 Wins as one SD.)

About half of the difference must be in run-scoring and run-prevention, not in the timing that produces an unusual number of wins or losses from given runs. In 1992 and 1993 --Duffy's first two seasons with the team, when it "steals two pennants"-- Boston is only +8 and +8 wins above Pythagorean expectation (derived from actual runs scored); in 1894 only +5; sum +21.

In the same three seasons Philly is -5 -3 -3 wins; sum -11

In five prime seasons 1894-98 Baltimore is +5 -3 +3 +4 -5; sum +4

66. Howie Menckel Posted: September 11, 2007 at 09:07 PM (#2520716)
HOM by pct of games at each position in the field or DH, thru 2003
. . .
Caveats: Totals treat all careers as equal.
A little off on players like McVey and Sutton due to changing schedule length.
Guesstimates on Negro Leaguers.
<u>Hybrid P-hitters such as Ward, Ruth, Caruthers, Spalding have estimates that attempt to reflect their respective roles</u>.

Howie,

1. (bold)
Does the qualification refer only to the variation in number of games played by teams in one league-season, or is there another problem?
That one is severe only for the NA 1871-75 and for occasional later clubs that drop out mid-season, maybe why you name McVey and Ezra Sutton.

2. (<u>underline</u>)
For pitcher-fielders is you estimate systematic?

For example, how do you score Bob Lemon 1946 and why?

Cleveland AL
154g scheduled
156g played

Bob Lemon
55 games
-- of which
32 pitcher games (28 starts)
12 outfield games (all in center!)
11 pinch games, at least

There is no close relation between the two questions.

how do you score Bob Lemon 1946 and why?
At this stage, I mean
2a. how would you score Bob Lemon if the 1946 season were his entire career, and why?
2b. how about 32 pitcher games, 12 outfield games, and 40 games played?

Paul, the changing schedules only matter where a guy may have played 5 seasons as a regular at one position and 5 at another. In that case, he'd get more "credit" for the longer seasons, which might not be ideal.

Not systematic on the hitter/pitchers. There I roughly went by "seasons as regular," and a small juggle for the rare hybrid that you note.
In the Lemon example, just spitballing, I'd probably say 85 P, 15 OF.

I'd actually be interested if anyone wants to refine these hard-to-manage players, as well as my Negro League estimates. They're in the ballpark, so to speak, but it would be nice to clean them up a little more.

Thanks, Howie.
I thought you were using some "seasons as a regular" measure for everyone. No, so Deacon White should be "mismeasured" more than Cal McVey ---given White's transition in fielding positions over a much longer career commencing at the same time.

I think I see how to do this systematically given a pitcher bonus and a catcher bonus that change through history. Pitcher games and catcher games should (separately) have greater weight than fielding games at other positions, in order to pick up the true significance of Martin Dihigo's pitching and Buck Ewing's catching. But what should be the weight.

Regarding the definition of a catcher bonus, Brian Downing may be a good modern player for consideration (Brian Downing at baseball-reference). The short 1981 season is one of his split-position seasons, easy enough to handle by prorating his fielding games. But what is the proper catcher bonus? How much weight should his 675 catcher games count, where the weight of his 777 outfield games is one? And why that much weight?

(I am thinking that DH games should have weight one, although I can see an argument for weight less than one.)

Dan R., from the Bob Lemon thread:

Joe Dimino, does your system have the entire second tier of 90s pitchers (Brown, Schilling, Mussina, Glavine, Smoltz, Cone) making it in? ...it seems to me that you may be making a mistake by adjusting for usage (by translating IP) but not for standard deviation.

I've got a problem in my consideration on Saberhagen, and what Dan said is part of it. Saberhagen's career straddles the divide between the LOOGY/setup/closer era we now live in and the transitional 80's. His peak lies back in the 80's. But he has a number of late years - years that are part of the latter era - in which he pitched with great effectiveness but for very few innings. (Some of those years were injury limited.) You can see some of that in the chart I posted in #107 above. For a pure peak voter, that wouldn't make a difference, but I've been a career voter for pitchers and those years are helping his case. And they should help his case; the question is, by how much?

I haven't found my way through this yet.

Here is a tangent from my tinkering with fielding positions data
Source : lahman5.4 database

Most games used by a team as pinch batter or runner only - no "position" fielding or DH

_1 1896 NY1 Reddy Foster, New York Giants ( 1-game career )
( tied by two others before broken )
_2 1899 LS3 Harry Croft, Louisville Colonels ( part-season ; later 2 games 2B with Phillies )
( tied by one other before broken )
_6 1907 BOS Deacon McGuire, Boston Red Sox ( part-season manager ; earlier 1 game Catcher with Yankees )
_6 1909 PIT Kid Durbin, Cincinnati Reds ( part-season ; also 1 game with Pirates )
26 1911 BOS Jack Thoney, Boston Red Sox
35 1914 NY1 Mike Donlin, New York Giants
43 1947 SLA Joe Schultz, Saint Louis Browns
43 1948 SLA Joe Schultz, Saint Louis Browns
54 1959 SFN Dusty Rhodes, San Francisco Giants
77 1967 CHA Smoky Burgess, Chicago White Sox

All-time Top 15

Ron Northey 1957 is double-listed in that Top 15 (which is remarkable in itself).

Next all-time is the DH-era leader
28 1991 PHI Ron Jones, Philadelphia Phillies

So I looked up Herb Washington, 1974. BB-ref lists him with 92 games played, with the "fielding" block recording 28 games at DH. However, he had no plate appearances. I would assume that these were occasions in which he pinch-ran for the DH. He would then have officially remained in the lineup unless or until the DH position came around in the order again, in which case he would have been pinch hit for. It is clear that those listings as a DH kept him off of Paul's list, but I think that's a technicality.

I would assume that these were occasions in which he pinch-ran for the DH. He would then have officially remained in the lineup unless or until the DH position came around in the order again, in which case he would have been pinch hit for. It is clear that those listings as a DH kept him off of Paul's list, but I think that's a technicality.

I agree.

To Howie Menckel
and anyone else who works on fielding position without using lahman5.4 or an equivalent database:

Is data in this following format useful? It is a selection from four database tables.

For example, here is Frank Huelsman 1904, the first and only player to change major league teams four teams in one season.



Semicolons in the header line and double-spaces in the data demarcate the four tables in a way and provide a little more readability here. Huelsman in 1904 changed teams four times with his 1st and 3rd of five stints for the Chicago White Sox (columns 5,6). In stint one he played in 3 of the 156 White Sox games (columns 7,8); in stint five he played in 84 of the 157 Senators games. In stint three, a one-game return to the White Sox, he played one game without fielding (column 9 is empty). In stint four he played 20 of the 156 Browns games including 18 in the outfield (columns 9,10); none at any other fielding position, which implies two pinch games.


Next is Bob Lemon 1946 in the same format.


Lemon did not change teams (one stint, column 5) but he did play two positions in the field, 32 as pitcher and 12 in the outfield (cols 9,10). His 32 pitcher games include 5 starts and 282 outs (cols 11,12), which may be useful for some of the purposes at hand.


Next is the entire mlb career of Cal McVey.


In this data there are eight fielding positions (POS): outfield, designated hitter, and the usual six infield and battery positions.

>>
Next is Bob Lemon 1946 in the same format.


Lemon did not change teams (one stint, column 5) but he did play two positions in the field, 32 as pitcher and 12 in the outfield (cols 9,10). His 32 pitcher games include 5 starts and 282 outs (cols 11,12), which may be useful for some of the purposes at hand.
<<

I don't know much about using Excel or any other spreadsheet program, but enough to know that some useful calculations will be convenient given data in this format. For example in 1946 Bob Lemon pitched in 32 of 156 or .205 share of Cleveland games, and 32 of 55 or .582 share of his games played.

It may remain quite inconvenient to work with the number 44 (sum of column 10), an estimate of the number of games he played in the field. For example, granting that any mid-game position switches constitute double-counting, Lemon played 32 of his 44 fielding games, or .727 share of his fielding games as a pitcher.

That calculation is convenient in a database program, which is designed for combining rows in one table, or across several tables, based on a match in part of the data. (For example Bob Lemon 1946, that is a match in the first six columns of the one table given here.)

If I manipulate correctly,
63 teams have used a player in 100 games without fielding. Oakland 1974 is tie for #70-71 with Herb Washington, 92 games.

How many pure designated hitters, without appearing in the field or in the pinch?
Eight, including three teams who used one player as DH in every game.
Here are the top ten.



Among all player-team-stints in mlb history I find
<u>1323</u> no fielding, of which
1079 pinch only
_199 pinch and DH only
__45 DH only (top ten listed above)

Ralph Houk twice and Darrell Johnson are the managers who one truly everyday DH.

Paul,
It is not possible for a "percentage" chart to completely and accurately reflect Negro Leaguers or pre-1871 players, for sure.
Some tend to want to "exclude" them, as if they were inconvenient.
I do not.
I have reasonably accurate numbers to reasonably portray where various players performed in the field.

For whatever reason, I'm less concerned with a "800 G C/800 G OF" guy than I am about changing season-lengths or P/H (that is, 30 G as a pitcher and 30 G as an OF surely are not a "50-50" split). But I'm open to suggestions that tweak the numbers a little bit more accurately.

OK,
Here is summary career fielding data for Cal McVey normalized for the number of games played by each of his teams.
FWIW, this can be calculated from more basic data that I have combined in a single table, an extension table quoted and described just above.

fielding position ; full seasons played (normalized) ; share of total "full seasons played"


to the nearest whole percent, in decreasing order, that is
_C : 1B : OF : 3B : _P : SS : 2B
35 : 26 : 17 : 15 : 05 : 01 : 01

Contrast these shares with the ones in Howie Menckel's table (previous page #66)
_C : 1B : OF : 3B
30 : 31 : 18 : 14 (sum 93)


The normalization is twofold for every player-team-season-stint. It accounts for the number of team games played (which varies by schedule, ties, cancellations, etc) and for multi-position player games (approximately). McVey frequently switched positions mid-game: at least 77 times in his nine seasons, for he played 530 games and with sum of games by position 607.
Cal McVey at baseball-reference
Without accounting approximately for mid-game switches he would have 9.79 "full seasons played" in sum over all positions --in only nine years.

Here there is no extra weight for battery games, the so-called pitcher bonus or catcher bonus. A one-stint player with 10 games each at pitcher, catcher, and first base would be scored 1/3, 1/3, 1/3 at those three positions.

Here there is no accounting for play outside the majors (NA and NL) --for McVey, before 1871 or after 1879. Everyone credits him with at least two more seasons, Cincinnati 1869-1870.

In the current version of my normalization the accounting for multi-position games or mid-game switches is approximate. Let me indicate the approximation by explaining that it is possible to do better using commonplace season statistics, not to mention game-level data or season innings data that is available for recent seasons.

1.
For anyone whose games at one position equal total games played, every multi-position games includes a switch involving that position.

McVey in 1871 played 29 games, 29 catcher games, 5 outfield games, 1 thirdbase game. (Dave Birdsall played 7 catcher games, so McVey missed part of at least one game.) In 1878 he played 61 games, 61 thirdbase games, 3 catcher games.

2.
For anyone who played only two positions in a season, all of his switches involved those two positions equally. For example recall the fictitious Bob Lemon with 40 games, 32 pitcher games, 12 outfield games. Presuming no pinch games, he played 28 whole and 4 part pitcher games, 8 whole and 4 part outfield games. Rather than use that deduction somehow, my normalization simply allocates his 40 games in the proportion 32:12.

In 1874 McVey played 70 games, 57 outfield games, 23 catcher games. Also in 1878 (above) he played only two positions.

3.
The fielding data for all players on a team includes information about the composition of its switches and substitutions.

For Boston in 1871 the fielding data for all players provides nearly complete information about the numbers of complete games, switches, and substitutions for each position and player.

4.
We know that a mid-game switch in McVey's time was a common method of changing the pitcher or catcher. Those two positions were relatively more likely than others to be involved in a switch.

This indicates a likely bias if the goal is to estimate the share of defensive innings by position. Is it?

Here is summary career fielding data for Deacon White normalized

to the nearest whole percent, in decreasing order, that is
3B : _C : OF : 1B : 2B : SS : _P
39 : 37 : 12 : 08 : 03 : 00 : 00

Contrast these shares with the ones in Howie Menckel's table (previous page #66)
3B : _C : OF : 1B : 2B : SS : _P
51 : 28 : 10 (sum 89)


But "full seasons played" may be more useful because it is easy to revise for play outside the majors. Here it is for White.

fielding position ; full seasons played (normalized)
3B 7.057
_C 6.735
OF 2.208
1B 1.447
2B 0.607
SS 0.066
_P 0.021

So McVey as 35-26-17-15 is presumably closer to accurate than the current 30-31-18-14 for McVey, yes?

If so, I'll take it.

Yet I like the full seasons played idea even better, perhaps.
If we credit McVey for 1869-70, and we believe he played mostly X position, I'd like to credit that, yes.

Question for Sunnyday

You analyzed 30 hitters, without analysing Tommy Leach. As a Winshare supporter, shouldn't a 329 WS / 24.7 rate get some love from you?

Responding to a comment from Bob Lemon thread:
"Don't forget, we've got 60% more teams in the 1980s-1990s than we had back in the day - it hold that the group of electees should also be that much larger."

Half of a modern roster (12/25) is made up of pitchers. 6/15 'starters' (5 SP and one closer) are pitchers, 6/16 in the AL. That's somewhere between 37.5% and 48% of the opportunity to play baseball coming as a pitcher. Being in the top 10 as a starting pitcher (93rd percentile) any given year is similar to being in the top 2 among position players at a given position. Pitchers are also less fungible than players so the replacement value is clearer. You can slide a 3B to RF or a SS over to 3B to cover short stretches. Position players don't slide over to pick up a start anymore like they did in the 1890s. In order to be "fair to all positions" we should be electing a pitcher _every_ year. We don't even have any in the queue!

We've elected 26% pitching so far. I'm only about 28% in my PHoM and I think I am going to take action to correct that to 30%.

So McVey as 35-26-17-15 is presumably closer to accurate than the current 30-31-18-14 for McVey, yes?

Yes for 1871-79.
He was the regular right fielder in 1869-70 so it isn't clear overall.
There is also the question how much to weigh McVey's 5% pitcher games. Their composition is 34 games, 19 starts, 12 complete, 12 finish; average 5.2 decimal innings.

Yet I like the full seasons played idea even better, perhaps.
If we credit McVey for 1869-70, and we believe he played mostly X position, I'd like to credit that, yes.

In my desktop database, or manipulating the big intermediate table in a spreadsheet, the full seasons played is more direct and maybe convenient in other ways. In a spreadsheet one would naturally calculate extra columns that normalize each player-season-stint-position to "full seasons". Focusing on one player, one position (eg using the "Filter" in Excel), a career record of full seasons would be displayed in that column.

For McVey as catcher in 1871,
[ 29 player games / 31 team games ] * [ 29 catcher games / 35 sum of position games ] = 0.775 full seasons
For Birdsall, the only other catcher for Boston 1871,
[ 29 player games / 31 team games ] * [ 7 catcher games / 34 sum of position games ] = 0.192 full seasons ]

The method handles each player-season-stint separately, using no team(mate) data except total games. The calculated "full seasons" as catcher for McVey and Birdsall, hence for Boston 1871, do not add up to one.
At catcher McVey played 24 complete games, Birdsall 2 complete, and they both played 5 part games. At his other fielding positions McVey played fewer complete games (in fact, none) and more part games (all).

--
What do you think, Howie?
I will let this sit a few days. Perhaps the method will attract other comment.

--
Is Excel still limited to 65536 records?
There are almost 130,000 player-season-stint-positions records.

--
In Access, I am storing many queries, nesting few. Should it be easy to improve that and nest more?

I'll rephrase my point this way. We're at 26% pitching. If the 6 remaining backlog slots between now and 2008 are _all_ filled with pitchers we jump up to 27%. If none of the slots goes to a pitcher (very possible) we fall below 25%. Be fair to all eras and all positions and vote for modern-era pitchers.

I make this comment every 20 "years" or so: there are two competing trends with pitchers across the history of the game. First, pitchers pitch fewer innings. Second, the importance of the pitcher (vs. the fielders or hitters) increases during each inning. If we elect roughly the same number (or percentage) of pitchers from each era, we are assuming that these two trends cancel each other out. They might but there is no reason to assume they do. It is perfectly plausible that the "right" percentage of pitchers is 40% in the 1890s and 15% in the 1990s or vice versa. I wish we had better empirical evidence to compare pitchers (and figure pitcher importance) across eras.

Another revised prelim (putting my ballot where my mouth is)

1) Boggs
2) Luis Tiant
3) Bob Johnson
4) Tommy Bridges
5) Bus Clarkson
6) Rick Reuschel
7) Graig Nettles
8) Bret Saberhagen
9) Virgil Trucks
10) Reggie Smith
11) Norm Cash
12) Tommy Leach
13) Ben Taylor
14) Lee Smith
15) Ron Cey
16-20) Gavy Cravath, Dick Redding, Vic Willis, Jim McCormick, Buddy Bell

"I wish we had better empirical evidence to compare pitchers (and figure pitcher importance) across eras."

We have some, but I agree it is incomplete. It is interesting that as defenses improved due to better equipment the value of a strikeout seems to go up which seems contradictory.

The amount general managers are willing to spend on pitching and the expanded number of pitchers on the roster point to pitchers deserving a higher percentage of electees in the modern era.

can someone post a chart of pitchers elected by 5- or 10- or 15- or 20-year blocks of
a) birth year or
b) first pro year or
c) last pro year or
d) middle year or
e) all years they were active?

I'd like to see how the modern (1977ff) era (5-man rotations, expansion, more offense) stacks up against the previous in terms of pitcher electees.

If you don't adjust for standard deviation, there are a BUNCH of pitchers from the 1990s that I would think would be locks...not just the big four of Clemens, Maddux, Pedro, and the Unit, but the entire second tier (Schilling, Mussina, Brown, Clemens, Glavine, Smoltz). How would letting those ten guys in affect these ratios? (note that I think you SHOULD adjust for standard deviation, but virtually no one here does).

whoops, I named Clemens twice. I meant to say Cone, but I don't think he's quite up there with the other five.

Woo! Another Matlock voter!!!!!

120,000 senior citizens religiously watching the USA Network can't be wrong!

dan, what if you do adjust for sd, then how do the 1990s guys stack up?

Well, that depends on how much you adjust for usage (innings)! Based on the stdev regression I'm currently using for pitchers (30% r^2, similar to that of hitters), I have the projected stdev for pitchers with at least 100 IP at 1.74 wins per 200 IP for the 1994-2004 period. By contrast, I have 1973-87 (excluding 1981) at 1.50. So that suggests that you would have to multiply the wins-above-average-per-inning rate of 1973-87 pitchers by 1.16 to make them comparable to 1990s pitchers, and vice versa. Then again, you'd also have to dock their innings pitched, which is a counterbalance--probably by more than 1.16 for the Wilbur Wood generation and much less for, say, early Clemens.

On number of pitchers for the 1990s:

The election schedule is set for us to elect three players per year now and to continue to do so through all the elections in which 1990s stars will be becoming eligible. That suggeests that for "the 1990s" we ought to be electing around 30 players. It might be 25 or it might be 35 depending upon how this generation compares to those that come before and after, but about 30 are planned for.

So far we have, by my count, elected 62 pitchers (that's adding up several "half-pitchers" like Caruthers, Rogan, Dihigo, and Ward) out of 222 electees. That's 28% pitchers. If we continue at that rate, we should elect around 8.5 pitchers for the 1990s.

To make up that group, we have

The Big Four: Clemens, Maddux, Martinez, Johnson
The Relief Duo: Rivera, Hoffman
The Second Tier: Brown, Cone, Glavine, Mussina, Schilling, Smoltz

To elect all 12 would seem a little high (though less so if relievers are a second category), though not necessarily out of the question.

To elect 8-9, we would essentially draw the in-out line through the middle of the Relief Tier/Second Tier, electing 5 of 8.

My guess is that we will elect more than 8 but not all 12.


While I am posting I should also explain that I have been absent from the discussion (and slow on answering some HoM e-mail!) because I have recently become a father! My wife and I had our first child--a healthy baby boy--a week ago today! As a result, I am spending less time with spreadsheets and more with dirty diapers, a pattern that will continue for a while.

Wow! Congratulations, Chris! That's great news. You didn't name him Tyrus, did you???

I'm going to PHoM 10 of them (Cone and Mussina are still maybes). I don't think 12/30 is too high at all for this era especially when you look at the 80s pitcher trough. Half of the MLB players in this era were pitchers.

Your son has some pretty good players in common with that birthday: Randy Johnson, Ted Kluszewski, Roger Maris, George Kelly. I remember my boys' birthdays as follows:

Me - Babe Ruth
Oldest - Willie Mays
Middle - Joe DiMaggio
Youngest - Shoeless Joe Jackson

My wife and daughter don't have anyone good on their birthdays.

Mussina was victimized by horrible defenses in NY. He was routinely among the league leaders in DIPS ERA.

Congratulations, Chris.

Spreadsheet time will decrease for a while, definitely :)

Mussina was victimized by horrible defenses in NY. He was routinely among the league leaders in DIPS ERA.

Then he lost his fastball....

On number of pitchers for the 1990s:

I'm still working on the pitchers from the 1980s!

Congrats Chris! Teach the kid to be a lefthander immediately for the fastest way to the majors...

Ran Saberhagen through my system and he's running equal to Tiant and Shocker for me. That will get him about #20 to start with on my ranking. I sense that he's the type of candidate that might end up with enough low ballot spots to make some noise this year with our splintered backlog.

Wow! Congratulations, Chris! That's great news. You didn't name him Tyrus, did you???

Not my favorite proper name, though it would be better than Corn-on-the Cobb. :-)

Congratulations, Chris!

The standard deviation/IP knot for pitchers is difficult to untangle.

The danger with adjusting for standard deviations is that these so-called 2nd tier guys were good enough to carry teams to real pennants.

Brown, Schilling, Mussina and Smoltz look like obvious selections to me, if they struck people out and didn't walk many then what are we supposed to do, give credit for winning those games to their fielders? Yes, adjust for changing league norms, but don't take away real wins that clearly should be credited to the pitchers.

Saberhagen and for peak voters Gooden may be ahead of Cone but still I'd put all three above the line. If you're using the peak salary estimator then I would think Gooden has to be on your ballot. Andy Petitte including 2007 I think is now above the line, as well.

Appier is more like the borderline as I see it, Joe's point I think is that he's in the ballpark, not that he's going in easily. Glavine is the one I would pass on, from a DIPS perspective he's the overrated one, getting credit wrongly for pitching on good teams (see also Jim Palmer). Wells doesn't make my cut either.

Santana, Oswalt, and Halliday are going in assuming they last a little longer. Of course with pitchers that's not guaranteed, but still, they are making very rapid progress.

Unfortuately this does mean that the ten most valuable pitching seasons have all been since 1995. And I'll say, yes, that's obviously true. Look, in 1999 Pedro struck out 313 in 213 innings while walking only 37...basically making it impossible for anyone to beat the Red Sox that year when he pitched. In 1998 Kevin Brown struck out 257 and walked 49, giving up only 8 home runs. Don't adjust that away, Brown should get a lot of credit for being hugely better than replacement even if Pedro, Roger, and Randy were also doing similar things. There were a handful of pitchers capable of blowing the other team away, while the vast majority of pitchers were getting rocked. If you didn't have one of those guys on your team, you sure wished you did. Especially with the multiplication of critical playoff games.

Jim Sp, ceteris paribus, if the underlying distribution of talent does not change, then it should take more wins to bring home a pennant in higher standard deviation years than in lower standard deviation ones (this is why there is a strong correlation between the winning percentage of the worst team in the league and standard deviation). When you adjust for standard deviation, you're not taking wins away from anyone, nor are you reapportioning credit to fielders from pitchers--you are merely correcting for changes in the relationship between regular-season wins and pennants.

Glavine isn't terribly impressive from a DIPS point of view, but he is one of those rare pitchers that actually does have the ability to get out of a jam by changing the way he pitches. Between 1990 and the present, he's allowed 124 fewer runs than we would expect from his hits, walks, hit by pitch, and home runs allowed (according to Baseball Prospectus). That is approximately 50 bajillion standard deviations from the mean--the only other guys I can think of who are in triple digits there are Whitey Ford and Lefty Grove. That's a real, and rare, skill of his, and you are wrong to ignore it.

Brown in particular will be very high on my ballot. Like you, I find his 1998 one of the premier pitching seasons of the century.

Gooden definitely falls short for me. Sure, 1985 is a $25M+ season, and '84 and '86 were excellent too (he was very hit-unlucky in 1984), but he was only slightly above league average for the rest of his career--it's just not enough.

David: What about Saberhagen, and in particular, the concerns I raised about how much his later low-usage but effective years should add to his 80's peak?

Chris: Congratulations. I've been there, but it wasn't recently. In fact, the youngest is a senior in high school and the empty nest is near.

(In #151, "David" = Dan R. Sorry)

Dan R,
I am in favor of adjusting for standard deviations in the sense that you list above. I'm not in favor of using them to wipe out a talent glut at a position, that's my concern (not having seen your numbers).

If you are adjusting pitching standard deviations separately from hitting and fielding standard deviations then you are shifting credit for wins away from the modern standout pitchers. If the position players are getting the same standard deviation adjustment as the pitchers then I don't have a problem with it.

It is not a problem to give more credit over time for wins to standout pitchers, if they deserve them. Increased K rates and K/W ratios, plus increased HR rates indicate that fielders should receive less credit over time and pitchers more. The % of pitchers vs. position players in the HoM does not have to be a constant, if Appier was better able to take advantage of his conditions to create wins than some position player, it doesn't matter that there were a bunch of better pitchers in the league. Which actually I think was your argument with regard to talent gluts at fielding positions...why would pitching be different?

Here's a thought experiment: If the standard deviation for pitchers doubled while the hitting/fielding standard deviations for position players stayed constant, then would salaries rise for the good pitchers? I believe they would, and if the market would value it then why shouldn't we?

We may be in violent agreement here. Saberhagen and Cone are definitely in the "your mileage may vary" area. I understood your earlier comment to be against Brown for the HoM, I thought that could only come from an excessive adjustment. Which I see now is not the case.

I wasn't aware of the Glavine data, interesting. If that's due to him and not his defenses then yes, he should get credit for it. Can you explain more about how you are using the BP data to get to that conclusion?

I have a lot of questions about how to separate defense from pitching, it would be interesting to hear the details of how you propose to do it, even if the system is not complete yet.

OCF,

Well, clearly you would need to reduce his ERA+ and increase his IP for those years. Again, the question is whether those adjustments cancel out or not, and I'm sorry to say my research isn't complete enough to have an answer yet.

Jim Sp,

Remember that since I use regression-projected rather than actual standard deviations, my adjustments don't "know" when there is a talent glut. (Indeed, high residual errors are usually evidence of star gluts or droughts).

The position players do NOT get the same standard deviation adjustment as the pitchers, because the standard deviation for hitters is influenced by different factors than the standard deviation for pitchers. That said, your point is very well taken that pitchers are responsible for a greater proportion of run prevention now than in the past due to the higher percentage of True Outcomes. My standard deviation adjustment to position players is applied to *overall* wins above average. The stdev of fielding has *definitely* declined over time, but that obviously gets swamped in the regression by the much larger effect of the hitting. I can definitely see an argument that I should do one stdev regression just on hitting (the run-scoring half of the game) and another on pitching and defense together (the run-prevention half of the game). If so, that might result in less steep corrections being applied to modern pitchers, as you suggest. (It could also improve my r-squared, since presumably fielding and hitting stdevs are affected by different factors as well). Whenever I get pitcher WARP going and have a fully integrated system, I will definitely see what kind of results that gives me.

I will be the best friend of Kevin Brown if he needs one. From 1996 to 2003 (minus '02) he had one of the best pitching peaks ever, no matter what adjustments you apply, and it was a pretty long one too.

The Glavine data is simply the DR (delta-runs) number from his BP page, which is the gap between runs actually allowed and the number of runs a run estimator would predict should have been allowed. His career DR is one of the best in baseball history. Moreover, I happened to have read an analysis somewhere of how he manages to accomplish it, I think it compared him to Kirk Rueter if you want to Google it...basically he becomes a nibbler from the stretch.

My basic methodology for pitcher WARP is:

1. Count the pitcher's balls in play allowed. Multiply that by the league batting average on balls in play, and then add his BP DH number (the number of hits on balls in play he allowed above or below the number he would have allowed if his BABIP were equal to that of his teammates' BABIP) to get defense-and-park-independent non-HR hits allowed.
2. Adjust his home runs allowed for his home HR park factor.
3. Plug the adjusted non-HR hits, adjusted HR, BB, K, HBP, and batters faced into a run estimator to get expected runs allowed.
4. Add on the pitcher's DR number (from his BP page) to credit him for runner stranding (or failure to do so).
5. Calculate his hitting value above/below the league average pitcher hitting for that year, using the same methodology as for position player WARP.
6. Convert adjusted runs scored and allowed to wins above/below average.
7. Adjust for standard deviation.
8. Subtract replacement level.
9. Adjust for era usage (both seasonal IP and career length).

Steps 1 through 6 I can easily handle now. But figuring out the appropriate methodology for steps 7 through 9 just takes more time than I've got these days, unfortunately.

Prelim is pitcher heavy, the pitchers at the bottom may get bounced.

1) Boggs
2) Saberhagen
3) Tommy John
4) John McGraw--Ultra-dominant player when healthy.
5) Rizzuto--The man lost his age 25, 26, and 27 seasons to the war, right after a very good season in 1942, and 1946 wasn’t a good one for him as well. One of the best fielding shortstops of all time. A 93 career OPS+ is strong for a grade A shortstop, not weak. Great peak season in 1950 (11.4 warp3). PHoM 1977.
6) Campaneris--great non-SB baserunning.
7) Concepcion--Grade A+ shortstop and could hit some too. Weak hitting at the beginning and end, but above average during prime 1973-1982. Warp3 prime: 10.7, 10.2, 10.2, 9.7, 8.8, 8.7, 8.3, 8.0. Note that Win Shares is conservative in assigning fielding credit to the great fielders. PHoM 1994.
8) Nettles--Great fielder with quite a bit of pop in his bat. Best Warp3: 10.7, 10.2, 8.9, 8.4, 8.2. PHoM 1995.
9) Reggie Smith--I’m convinced now. Compare to Wynn.
10) Reuschel--Joe D is on to something here.
11) Quinn--ditto
12) BancroftConvinced now that the BP warp discount is excessive.
13) Paul DerringerDIPS data looks very good in comparison to league.
14) Dutch Leonardditto.
15) Jim WhitneyOPS+ of 112 in the stronger league, plus a K/W ratio of 1571/411. Nice.

Dan R,
How are you accumulating the numbers off the BP web site into usable data?

And where are you getting the HR park factors?

Wow, I'd never heard of Jim Whitney. That 1883 season is Carutherseque.

I'm still working on the methodology before applying it en masse. I imagine I'll just ask David Foss to download and merge it again, as he so kindly did enabling me to produce my position player WARP.

The HR park factors are from the home/road splits for every park-season in history oh-so-helpfully provided to me by KJOK. I am really indebted to various memberes of the electorate for their contributions to my research.

Hi,

Remember back around comment 92 when people were discussing the offensive winning percentage of 19th century players being way too low? I contacted Sean Forman and he wrote me today that correcting it is on the "to do" list, tying OWP to RC vs League RC. No time table, but Sean will get to it. Isn't it great that Sean runs Baseball-Ref and not some big faceless corporation - at least until he sells it for a few million.

Quick question.

Luke Easter has a shot to enter my PHOM this year. What cap should he wear? Indians? Cincinnati Crescents? Titanium Giants?

Chris: congrats on the new baby.

How are you accumulating the numbers off the BP web site into usable data?

See post #49 of the Rosenheck data thread. BP's URL's have the playerid in them so you can create a list of all their player pages and then download them all those to your hard drive in one batch command of 'wget'. From there, the HTML can be parsed in ASCII files using some sort of WYSIWYG HTML-to-ASCII converter. From there, you can parse the ascii files and split the lines into columns using a whitespace delimit (BP kindly fills in zeros when data is missing).

I'm still working on the methodology before applying it en masse. I imagine I'll just ask David Foss to download and merge it again, as he so kindly did enabling me to produce my position player WARP.

Did they update the site again?

Not to my knowledge...I'm just saying whenever I'm ready to take on pitchers.

Wow, I'd never heard of Jim Whitney. That 1883 season is Carutherseque.

Let's see: you started voting in 1925, left after 1939, and came back in 1994 (with a couple of other missing years here or there). So you have an extensive history with the project, but aren't quite an old-timer. I'm not a founder, but as someone who started voting in 1904, I guess I am an old-timer.

By 1910, we were mostly done with pre-1893 pitchers, having elected Clarkson, Keefe, Galvin, Spalding, and Radbourn, as well as the pitching half of Ward. At that pont, there existed support in the electorate for other such pitchers, but there was division as to who the favorite candidate was. There were really five splitting that support: Bob Caruthers, Tony Mullane, Jim McCormick, Jim Whitney, and Mickey Welch.

I'm looking at my 1916 ballot tally, the first I recorded. There were 46 voters, and it was an elect-2 year, with Cy Young unanimous first and Fred Clarke a very strong second. Where were the pre-60'6" pitchers"

Caruthers was in 11th place with 310 points, appearing on 25 ballots.
Welch was in 24th place with 78 points, appearing on 7 ballots.
McCormick was in 28th place with 48 points, appearing on 6 ballots.
Whitney was in 35th place with 18 points, appearing on 2 ballots.
Silver King had 10 points, appearing on 1 ballot.
Mullane did not appear on any 1916 ballots but had apparently been on someone's 1915 ballot.

As you may know, a major long-term campaign was waged on behalf of Caruthers, and he was eventually elected in 1930. The McCormick and Whitney voters, most of whom felt they had good reason for regarding either McCormick or Whitney to be superior to Welch, eventually gave up these pitchers in favor of more recent candidates; more of the Welch voters have hung on.

I was in the McCormick camp, myself, but I also last voted for him in 1915. (He was on my ballot from 1905 through 1915, topping out at 9th in 1906 and 1908.)

The Welch voters in 1916: Daryn, karlmagnus, MattB, Rusty Priske, Jeff M, yest, Adam Schafer.

The McCormick voters in 1916: Rusty Priske, Marc, Tom H, Clint, Mark McKinniss, Devin McCullough.

The Whitney voters in 1916: Marc, jimd. Both of the Whitney voters also had Caruthers on their ballots in a higher position.

In 1916, Jim Sp (who still had a full last name) had Caruthers in 12th as his only pre-93 pitcher.

I was one of the leaders of said Caruthers campaign. I remember gleefully counting the votes ahead of time to see if he got in. No one I idiosyncratically support now is close.

It's very odd that there aren't Caruthers-type players today, because cricket, equally long-lasting and professionalized/specialized (though not quite as lucrative) has lots of them -- players worth their place in the team as either hitter or pitcher (batsman/bowler.) Particularly with the DH, I don't see why a pitcher can't play full time, DHing when not pitching, and obviously his team would benefit greatly if he did so (certainly in terms of WS.)

Thge "all-rounders" are not necessarily the absolute best at either specialty -- not Bonds or Pedro -- but if you imagine a player who combined Josh Beckett with Dave Kingman, you're close to the England 80's player Ian Botham, and that's pretty useful to a team.

The general view is that both hitting and pitching at the MLB level is just so damn hard that you can't make it if you don't devote yourself to one specialty or the other. And then there's Rick Ankiel, who's just plain weird.

If hitting and piching at an ML level are so hard, why aren't batting and bowling? And why couldn't Ankiel have had his hitting and pitching success simultaneously, rather than successively? Could he have learned to hit ML pitching while remaining a pitcher, perhaps in the bullpen?

For example, if a decent ML setup man is worth $3mm, and a decent DH is worth $7m, why aren't there players who get $10m for doing both? Setup and DH aren't all that physically taxing, so it ought to be possible to combine the two.

In the historical record, 1893 is the great divide here. After that time, the players who hit and pitched both (not necessarily at the same point in their careers) were significantly better hitters when they stopped pitching. Even Babe Ruth. Along those lines, I doubt Ankiel would be the hitter he is today if he were still pitching.

Under current AL rules, DH and relief pitching are essentially incompatible duties, as are pinch hitting and relief pitching. If you use him as a pinch-hitter in a game, you've "burned" him and he's not available to pitch. And if he starts the game as DH, if you then choose to bring him into the game to pitch, then you've lost the DH, and the next pitcher will have to bat.

NCAA rules are different. They decouple the position of DH from that as pitcher, and in so doing enable the position of DH/P. Someone like that can start the game as both a DH and a pitcher and be relieved as a pitcher while remaining the DH. There are a number of people who have had successful collegiate careers doing that - the Mets' Jason Vargas is a good example. (Better hitter than Troy Tulowitzki when they were teammates.) The opportunity to continue with that sort of career would lie in the NL, as a starting pitcher who is available to pinch hit on the days when he's not pitching.

Well, supporting Caruthers wasn't really idiosyncratic if he got elected ;-) Here are some guys I supported in the first 19 years (conveniently located on one page of my records) who were never elected.

Jim McCormick--13 of 19 years with a peak of #4
Harry Wright--10 years/peak #5
Tommy Bond--10 years/peak #7
Ed Williamson--8 years/peak #11
Pete Browning--8 years/#5
Tony Mullane--4 years/#14
Jimmy Ryan--4 years/#6
GVH--2 years/#12
Billy Nash--2 years/#15
Hugh Duffy--2 years/#7
Denny Lyons--1 year/#15
Jim Creighton--1 year/#10

Still support Bond and Williamson and of course Browning. I'm back to supporting Duffy. McCormick is also pretty bona fide. The rest, not so much.

Speaking of old timers, retrosheet just added box scores for 1872 and 1874. We're done with those guys here, but its fascinating stuff.

DavidFoss, does that include anything that would help me with run estimation for those years? (e.g. reached on error, CS, GIDP data)

Replying to TomH, here is my estimate of how many wins each pitcher contributed to his
team over the course of their career, on a game-by-game basis, given his team's offensive
support, relative to a *league average* pitcher (note these figures include estimates
of the 2007 season):

80.1 Roger Clemens (a)
67.9 Greg Maddux (a)
55.9 Randy Johnson (a)
54.9 Tom Seaver
53.1 Pedro Martinez (a)
44.9 Jim Palmer
41.1 Mike Mussina (a)
39.5 Tom Glavine (a)
38.1 Curt Schilling (a)
38.1 Bert Blyleven
37.3 Gaylord Perry
36.7 John Smoltz
35.0 Steve Carlton
34.1 Kevin Brown
33.6 Fergie Jenkins
31.1 Nolan Ryan
30.6 Phil Niekro
28.3 Bret Saberhagen
28.2 Don Sutton
27.6 Kevin Appier
27.4 David Cone
26.7 Tommy John
25.3 Jimmy Key
25.0 Chuck Finley
24.5 Luis Tiant
24.3 Ron Guidry
22.9 Rick Reuschel
22.8 Dave Stieb
21.1 Orel Hershiser
20.6 David Wells (a)
18.8 Jack Morris
18.6 Dwight Gooden
18.5 Vida Blue
17.8 Bob Welch
17.1 Frank Viola
16.5 Jerry Koosman
14.2 Dennis Martinez
11.4 Frank Tanana
9.2 Charlie Hough
7.8 Dave Stewart
7.5 Catfish Hunter
7.4 Doyle Alexander
1.3 Mike Torrez
-0.5 Jerry Reuss
-2.9 Joe Niekro

Relative to a *league replacement level*:

127.9 Roger Clemens (a)
115.1 Greg Maddux (a)
103.7 Tom Seaver
96.4 Randy Johnson (a)
91.3 Gaylord Perry
89.9 Nolan Ryan
89.7 Bert Blyleven
89.0 Steve Carlton
88.0 Phil Niekro
86.7 Tom Glavine (a)
85.4 Don Sutton
84.9 Jim Palmer
81.4 Pedro Martinez (a)
80.7 Tommy John
79.9 Fergie Jenkins
74.1 John Smoltz
73.7 Mike Mussina (a)
71.8 Curt Schilling (a)
69.6 Kevin Brown
63.2 Rick Reuschel
63.0 Luis Tiant
60.5 Chuck Finley
59.7 David Wells
59.6 Dennis Martinez
59.3 David Cone
58.9 Jack Morris
58.2 Frank Tanana
58.1 Jerry Koosman
57.3 Orel Hershiser
57.2 Kevin Appier
56.8 Bret Saberhagen
56.5 Jimmy Key
54.8 Vida Blue
54.5 Dave Stieb
53.6 Bob Welch
52.7 Charlie Hough
49.9 Dwight Gooden
49.6 Ron Guidry
48.7 Frank Viola
44.6 Doyle Alexander
43.8 Catfish Hunter
42.6 Jerry Reuss
39.7 Joe Niekro
38.3 Dave Stewart
36.6 Mike Torrez

Recall that this metric does not reflect the team's defense behind the pitcher.

DavidFoss, does that include anything that would help me with run estimation for those years? (e.g. reached on error, CS, GIDP data)

There's CS data, but that's already at bb-ref. It not PBP data and its patchy like the 1912 data is (missing XBH in some games, etc). You do see all the errors (lots of passed balls), but no ROE data.

Looking at October 5th, 1872 games, you see a three hit shutout by Spalding over the Atletics:

RedSox vs Athletics

The ten errors a bit of a giveaway as is the pitcher's slot in the batting order, but otherwise that almost looks like it could happen today.

Then on the same day you have Baltimore beating the Atlantics 39-14. The Canaries Scott Hasting committed 4 E's and 6 PB's in a winning cause. He went 7-9 at the plate to atone for his bad hands. Atlantic commited 18 errors (3B-SS-2B committed 5-4-4 errors).

Baltimore Brooklyn Box from post #174

Rob Wood, thanks much. Good timing, too... I sit here with a nasty, nasty head cold, so instead of infecting my office mates, I'm spreading germs onto my keyboard.

congrats to Chris on being a daddy. just think, in the next 22 years, you'll drop a cool quarter million or so on this child :)

Some of our modern hurlers from Rob's data

wins above avg
28.3 Bret Saberhagen
22.9 Rick Reuschel
22.8 Dave Stieb

wins abve replacement
63.2 Rick Reuschel
56.8 Bret Saberhagen
54.5 Dave Stieb

With the understood caveat 'defense not included'. My Questions are two, one for Rob, one in general:

1) Do these numbers include a pitcher's entire career? For example, are below average years at the twilight of a career indluded, dragging "down' the "above avg" total?
2) How do the wins above average here compare with the WAA metric? The difference between the two I would find interesting; as it ought to reflect a measure of clutch wins based on W-L record, run support, and possibly a difference in handing unearned runs.

Here's my RA+-Pythpat records for many of the pitchers in Rob Wood's post #173, with the following changes: I reached slightly further back in time to include Seaver, GIbson, and Koufax. And I haven't worked up most active or recently retired pitchers yet, so you can't see where they fit. (That probably should be my next project, to catch up with this). In the listing below, the first number is the FWP associated with the equivalent record indicated after the name.

Things not adjusted for: pitcher's own offense; defensive support; league strength. (Except I did adjust Palmer for defensive support.)

333 Seaver 330-201
279 Blyleven 322-230
269 Perry 337-258
262 Gibson 265-166
261 Carlton 328-252
261 Palmer 268-171 * see below
253 Niekro 334-266
240 Jenkins 287-213
* 235 Palmer 260-179 [Defense - adjusted]
231 Ryan 326-273
228 Sutton 320-267 ["Ryan without the no-hitters"]
194 Marichal 226-164
191 Eckersley 215-150 [Many things not adjusted for]
189 Tiant 224-164
187 John 281-244
172 Stieb 190-133
171 Koufax 163-95
170 Drysdale 209-157
170 Reuschel 221-174
169 Saberhagen 174-111
168 Koosman 233-193
158 Kaat 262-241
155 Tanana 245-220
155 Key 171-117
150 Martinez 231-203 [That would be Dennis]
148 Larry Jackson 200-162
148 Morris 226-189
145 Guidry 158-108
144 Pappas 195-159
142 Blue 202-169
141 Lolich 215-189
140 Hershiser 191-157
138 Hunter 206-178
137 Viola 177-138
135 Gooden 174-137
133 Friend 212-190
133 Perry 196-169 [The other brother: Jim]
130 Hough 219-203
130 Candelaria 160-121
124 Langston 178-151
116 McDowell 154-123
116 Wood 163-136 [Wilbur the knuckler]
114 Cuellar 167-144
111 Tudor 120-80
103 Reuss 204-203
97 Valenzuela 168-158
91 J. Niekro 196-202
83 Stewart 149-143
73 McLain 112-98
72 Forsch 153-157
70 Richard 97-81
67 Scott 118-112
65 E. Wilson 117-111

Because FWP has gives positive value to some sub-.500 seasons, this ought to be closer to the second list in Rob's post. Look for differences in our lists - I think many of them can be explained if we think about it.

I do have lines in there for Clemens, Martinez, and Maddux. I'll have to look up when I did those - my best guess is that this includes the 2005 season but not 2006 or 2007

[Incomplete data]
365 Clemens 326-173
320 Maddux 297-167
248 Martinez 185-70

from posts 173 and 177

pitcher .... WoodWAA OCFPythpat diff
Saberhagen 28.3 ....... 31.5 ...... -3.2
Tiant.......... 24.5 ....... 30.0 ...... -5.5
Reuschel.... 22.9 ....... 18.5 ........ 4.4
Stieb.......... 22.8 ....... 28.5 ...... -5.7

If I read this right, Reuschel is the only one who gains from the game-by-game run-support analysis.

That sounds like a reasonable conclusion, and I would guess that the largest part of that is defensive support. Saberhagen, Tiant, and Stieb generally pitched in front of good, or at least average, defensive teams; Reuschel didn't.

but I think def support is not included in EITHER metric.

Rob Wood #173, point of reference #176-181
here is my estimate of how many wins each pitcher contributed to his team over the course of their career, on a game-by-game basis, given his team's offensive support, relative to a *league average* pitcher

What is the game-by-game data?
opponent, ballpark, and runs score?

Yes, my figures include a pitcher's entire career, so that any bad final years
could well drag down his career number. This is most true for the metric relative
to league average. Replacement level pitcher is pretty crappy so I wouldn't expect
bad final year(s) to drag down the career number relative to replacement very much
if at all.

To answer Paul's query, my method estimates the win probability value of the
pitcher's contribution in the game, given how many runs (earned + unearned) he
gave up and how many runs his team scored in the innings he pitched. It explicitly
takes into account the ballpark but not the specific opponent.

I was emailed something today about Diamond Mind signing a deal with MLB. Is that the thing some of you guys play?

yes it is howie - what was the deal about?

congrats to chris on fatherdom!

I'll post my list when I get back.

Rob, don't be surprised even with a low replacement level at how much bad years can drag a pitcher down.

This is pretty much it here. I think mine had an offer to do simulations, etc.
But I think this gives the basic 'public' stuff.

http://bdd.blogharbor.com/blog/_archives/2007/9/6/3212867.html

one more request; can someone with Sinin's BB encyc post the RSAA ##s for the 4 guys below? Thanks.

pitcher .... WoodWAA OCFPythpat
Saberhagen 28.3 ....... 31.5
Tiant.......... 24.5 ....... 30.0
Reuschel.... 22.9 ....... 18.5
Stieb.......... 22.8 ....... 28.5

#'s from SBE by Sinnis

pitcher .... RSAA
Saberhagen ..241
Tiant........172
Reuschel.... 202
Stieb........236

Thanks Mike. What I see from this is:

Stieb (and his team) got fewer wins from his efforts than the other 3 did from theirs; he is last in Rob's WAA measure, despite being a close 2nd in RSAA. In fact, Tiant gains the most from comparing the two sets of data.

This doesn't mean I advocate using WAA or RSAA as the main tools for HoM-iness. But whatever tools you use, I suggest in terms of 'clutch wins' or whatever you may wish to call it, Tiant has a small edge over Reuschel and Saberhagen, and if anything we may have rushed Stieb into our Hall before comparing him fully to two of his other contemps.

I personally think people just fell for Stieb's ritzy ERA+ in 1985, which was an unearned runs illusion.

But whatever tools you use, I suggest in terms of 'clutch wins' or whatever you may wish to call it, Tiant has a small edge over Reuschel and Saberhagen,

Tom, how do you reach this conclusion from the data presented so far, and what do you mean? Do you mean that Tiant got more wins per run saved than the other pitchers?

That may be true, but in comparing Saberhagen to Tiant, Saberhagen still got more wins for his team: he was not as "clutch," but he was flat-out better.

As to "rushing" Stieb, it's not that we rushed him. Tiant is not Stieb's contemporary. He peaked in the late 1960s and early 1970s, along with, roughly, Gibson, Seaver, Carlton, Perry, Niekro, Blyleven, Ryan, Jenkins, Palmer, and Sutton. There is suspicion (perhaps backed up by Dan R's standard deviation numbers) that it was easier for pitchers to dominate during that era than it was in the 1980s, and it was certainly easier for pitchers to rack up lots of IP during that era. Reuschel was Stieb's contemporary, but he was also a 1970s pitcher during the high-innings period.

Using wins above average without adjusting either for usage patterns or for ease of domination (or league quality, another mark against Tiant) will underrate Stieb.

No, he doesn't do well by a "clutch pitching" measure, but Wins above average isn't the whole story.

I personally think people just fell for Stieb's ritzy ERA+ in 1985, which was an unearned runs illusion.

Well, I use a wins-based measure, in which Stieb's 1985 shows up as nothing special, and I supported Stieb's election. He's near the bottom of the pitching HoM, but he's in.

Well, since I'm the RA+ guy, I'm immune to "unearned runs illusions." In RA+ equivalent, Stieb had a nice flat-topped 4-year peak from 1982 through 1985. I have his RA+ for those four years as 132, 141, 156, 154, and equivalent records of 20-12, 20-11, 20-9, and 20-9. And the 1981 strike year (129, 12-8) wasn't too far behind.

And after one off year in 1986, he came back and was effective in 1987-91, albeit in fewer innings.

I have the MLB projected standard deviation for pitchers between 1969 and 1973 at 1.92, and from 1982 to 1986 at 1.84. It was slightly, but not substantially easier for pitchers to have very high ERA+ scores in the early 70s than in the mid 80s. The much bigger difference is in the amount of innings they compiled.

Chris, completely agree the WAA is not the whole story - I tried to say that specifically.

Yes, Tiant maybe shouldn't be in this list of contemps, but since he overlapped a little and is one of our bordelriners, I put him in. Yes, I mean that he got more wins per run saved. As to Saberhagen being "better"; well, yes, their careers were different shapes. Better vs Longer.

If we're merely comparing Stieb/Saber/Reuschel, Stieb loses a little in my book by 'clutch pitching'; relative to the other two. He's still not a bad HoM selection; somebody has to be the borderline guy, right? :)

Since we're in Saberhagen discussions, going back to what Chris Cobb posted in #140 when talking about 1990s pitchers:

If we continue at that rate, we should elect around 8.5 pitchers for the 1990s.

To make up that group, we have

The Big Four: Clemens, Maddux, Martinez, Johnson
The Relief Duo: Rivera, Hoffman
The Second Tier: Brown, Cone, Glavine, Mussina, Schilling, Smoltz

To elect all 12 would seem a little high (though less so if relievers are a second category), though not necessarily out of the question.

To elect 8-9, we would essentially draw the in-out line through the middle of the Relief Tier/Second Tier, electing 5 of 8.

My guess is that we will elect more than 8 but not all 12.

Now to the above list add Hershisher, Appier and Saberhagen in the second tier. I'm adding Saberhagen since even today he's 43 years old - or younger than Clemens. How does he rate amongst Brown, Cone, Glavine, Mussina, Schilling, Smoltz, Hershisher and Appier? I want to be sure we're not rushing him to election just because he retired earlier than all the rest.

In fact, he might be the second coming of Tiant - in the grey area in an era where others were better.

Now to the above list add Hershisher, Appier and Saberhagen in the second tier. I'm adding Saberhagen since even today he's 43 years old - or younger than Clemens. How does he rate amongst Brown, Cone, Glavine, Mussina, Schilling, Smoltz, Hershisher and Appier? I want to be sure we're not rushing him to election just because he retired earlier than all the rest.

Excellent point, Al.

It might be me, but it looks like Saberhagen is being tossed into the Stieb group of pitchers instead of the Clemens group.

Using DRA+ (base is R Allowed, not ER Allowed), adjusting for bullpen support, defensive support, strength of league relative to the NL and era IP norms, I get:



The peak is similar to Red Ruffing but a little better.

It's incredibly consistent. Ruffing is the only one with two similar 'best two' seasons (of 14 pitchers) that has close to as good of a best 3rd and 4th best years.

He's not a terrible choice, although I didn't push him hard. I do like him a little better than Walters and Newcombe, but I liked Shocker and Bridges more.

I don't understand at all rating players based on their age. Saberhagen being 43 today is irrelevant. When he was in his prime from 1985-94, most of those other guys were pretty irrelevant or just starting out.

I have no problem rating him with Hershiser and maybe Appier - but most of the others are a stretch to call his contemporaries.

I didn't see Cone in the list either, of course I'd include him.