Fixing Pitcher Wins

August 29, 2008
 
 
On August 1 – the first game after the Manny Ramirez trade – the Red Sox beat the A’s, 2-1. Tim Wakefield pitched 6.1 shutout innings, but he ended up with a no-decision. This is from MLB.com’s game summary:
 
“Wakefield's string of being unable to win despite quality performances continued. With two outs in the eighth, Jack Cust hit a game-tying solo shot against lefty Hideki Okajima that just cleared the Green Monster… It was the sixth time this season he [Wakefield] has left a game with a lead but hasn't had a win to show for it.”
 
This is a pretty common thing – starter leaves a game in line for the victory, bullpen blows the lead, offense comes back, and some reliever gets credited with the win. That exact thing has happened to Wakefield four times this year alone. Really, it’s a loophole in the way pitcher wins are determined. It makes no sense to reward a bullpen for failing to hold a lead. Relievers don’t care about getting wins anyway; if anything, they care about saves, and I always find it ironic to see a reliever record a blown save and a victory in the same game.
 
I’d like to propose that we correct this, that we close that loophole. It’s not the massive overhaul of the pitcher wins statistic that some have called for; rather, it’s a minor revision, an update to a very old measurement. We’ll call the new wins “Added Wins.”
 
An Added Win, then, is any game in which
 
1.      The starter leaves a game eligible for the win,
2.      He is not credited with the win, and
3.      His team wins the game.
 
This does not – not – make any kind of contextual adjustments. It doesn’t create wins that aren’t there – we’re not taking Tim Wakefield, putting him on a team with a neutral offense, neutral bullpen, etc., and then figuring out a theoretical won-lost record. What we are doing is quite simply taking undeserved wins away from Red Sox relievers and returning them to Wakefield – who would have got them anyway if the relievers had done their job in the first place. The wins themselves are very much real wins, i.e., the Red Sox did indeed win those games.
 
By no means is this a perfect system. All the rest of the flaws in the pitcher wins statistic are still there. You still have guys giving up 8 runs in 5 innings and getting credited with the win. You still have guys pitching 8 shutout innings and walking away with a no-decision. All I’m doing is correcting one thing, closing that one loophole. To put it another way, this is not an analytical tool so much as it is an accounting adjustment.
 
I looked at 121 pitchers who spent all (or the vast majority) of their careers in the “Retrosheet era,” i.e. 1957 to the present. These included all the big winners, a bunch of active players, and some random others. One note – I haven’t done a reverse of this study, looking at how many wins individual relievers stole from their starters. A number of the pitchers in this study spent significant time as relievers, and I haven’t gone back and taken away those undeserved relief wins. With that disclaimer out of the way, here are the pitchers in the study with the most career Added Wins:
 
Pitcher
Wins
AW
Total
Roger Clemens
354
26
380
Tommy John
288
24
312
Greg Maddux
353
21
374
Kevin Brown
211
21
232
Don Sutton
324
19
343
Bert Blyleven
287
18
305
Orel Hershiser
204
18
222
Jamie Moyer
240
17
257
David Wells
239
17
256
Luis Tiant
229
17
246
Jerry Reuss
220
17
237
Kenny Rogers
218
17
235
John Smoltz
210
17
227
Mike Flanagan
167
17
184
 
With that one minor change to the rule book, Tommy John and Bert Blyleven are 300-game winners and easy Hall of Famers. The other guy in the group is Jim Kaat; he jumps from 283 wins to 295. You’ve got to figure a 295-game winner would eventually make the Hall of Fame. Luis Tiant’s chances are also improved. And Clemens and Maddux would rank 3-4 all time in career wins, rather than 8-9 as they do now.
 
Some other items of interest:
 
·       Randy Johnson should have notched win #300 with his last victory in 2007. He’s got 16 Added Wins, for an adjusted career total of 310.
 
·       Mike Mussina currently has 265 career wins. Throw in his Added Wins and he’s at 281, and looking at 300 either next year or early 2010. Not only that – Mussina, who has never won twenty in a season, should have reached that mark three times in his career. Our rule change gives him 20-win seasons in 1995, 1996, and 2002. Rather than a borderline Hall of Famer, he’d already be a no-doubter.
 
·       Ten pitchers in the study have had multiple 20-win seasons taken from them by their bullpens. Here’s the list:
 
Pitcher
20W
Added
Total
Greg Maddux
2
4
6
Randy Johnson
3
3
6
Mike Mussina
0
3
3
Ken Holtzman
1
3
4
Roger Clemens
6
2
8
David Wells
1
2
3
Mark Langston
0
2
2
Tim Hudson
1
2
3
Roy Halladay
1
2
3
Dan Petry
0
2
2
 
In other words, Maddux has two 20-win seasons, but he should have four more, for a total of six. We’ve been hearing for, what, twenty years now about the demise of the 20-game winner? Make this rules adjustment, and the 20-game winner is no longer an endangered species. By my count, with this adjustment, we’d have sixteen additional 20-game winners since 2000 alone (one per league per year), plus four pitchers – Randy Johnson, Barry Zito, Roy Halladay, and Curt Schilling twice – with 24 or more victories. Johnson should have gone 26-5 in 2002, Zito 25-5 the same year.
 
·         The most Added Wins in a season – at least, the most I’ve found – is five: Jim Perry in 1969, Joaquin Andujar in 1983, Orel Hershiser in 1998, and Tim Hudson in 2002. Perry, who actually was credited with 20 wins, should have gone 25-6 (a record that probably would have given him the Cy Young Award). Hudson’s record goes from 15-9 to 20-9.
 
·         Hudson was particularly unlucky in 2002 and 2003. He was robbed of five wins and a 20-win season in ’02, and the next year was jilted of four wins and another 20-win year. In the four years from 2000 to 2003, here’s what Hudson’s records look like with our adjustment:
 
Year
W-L
2000
20-6
2001
18-9
2002
20-9
2003
20-7
 
All told, Hudson should have 162 career wins, rather than his current 146. He’s going to have Tommy John surgery and miss most of 2009; apart from that, he’d be a great 300-win candidate.
 
Who tends to rack up the most Added Wins? Obviously, the more complete games a pitcher throws, the fewer opportunities his bullpen will have to blow leads for him. Juan Marichal completed 206 of his 238 career starter wins (86.6%, the highest in the study). Correspondingly, he had a mere five Added Wins in his career – just 2.1 Added Wins per 100 wins as a starter, the lowest rate of any 200-game winner. On the other end of the spectrum is someone like Woody Williams, who completed only 4.7% of his starter victories and had a very high Added Win rate of 10.9.
 
Altogether, the pitchers in this study completed 39.5% of their wins and had a rate of 6.1 Added Wins per 100 starter wins. But as you might expect, the rates have varied considerably over time.
 
Pitchers in the study born between 1935 and 1944 completed almost two-thirds of their starter wins (66.4%), and they had a rate of 4.6 Added Wins per 100 starter wins.
 
Pitchers born between 1945 and 1954 completed just over half of their starter wins (50.6%) and had an Added Win rate of 5.7%.
 
Pitchers born between 1955 and 1964 completed 29.2% of their starter wins and had an Added Win rate of 7.0%.
 
Pitchers born between 1965 and 1974 completed 17.5% of their starter wins and had an Added Win rate of 6.6%.
 
Pitchers born between 1975 and 1984 completed 11.9% of their starter wins and had an Added Win rate of 7.5%.
 
Apart from the 1965-1974 blip, the pattern is clear – fewer complete games, more Added Wins. And it makes intuitive sense. But this general rule, which holds pretty steady in large groups, doesn’t necessarily apply to individual pitchers. For instance:
 
·         Luis Tiant completed 67.1% of his starter wins, but had a rather high Added Win rate of 7.8.
 
·         Earl Wilson completed nearly half his starter wins – 48.7% – but had a very high Added Win rate of 9.2.
 
·         Mike Flanagan had about the same completion rate (46.1%), but his Added Win rate was even higher, 10.3%.
 
·         On the other hand, some of the lowest Added Win rates in the study come from pitchers who hardly ever pitch complete games. Justin Verlander and Kelvim Escobar have Added Win rates under 2.5, but neither has completed even 10% of their starter wins.
 
·         Want someone with a longer career? Andy Pettitte has completed a measly 8.5% of his starter wins, but his Added Win rate is a very low 3.8.
 
·         Other pitchers with low completion rates and low Added Win rates include Roy Oswalt, Carlos Zambrano, Pedro Martinez, Brandon Webb, Mike Hampton, and Mark Buehrle.
 
I haven’t done an exhaustive study, so I don’t know what the causes of these trends are. Completion rate – and innings per start, which I didn’t check – are probably the most important factors, but they aren’t the only ones. Bullpen quality plays a role, as does offensive support – you want a good offense if you’re going to take the lead, blow it, and then come back to win.
 
With this relatively minor rule adjustment, we’d have several more Hall of Famers (Blyleven, John, probably Kaat, and possibly Tiant). We’d have more 300-game winners and loads more 20-win seasons. The rules governing pitcher wins have been around since, what, the Middle Ages, right? They were written when pitchers completed almost all of their starts – back in the day when men were men, when catchers didn’t wear gloves and major league baseball was played on fields that would make today’s Little Leaguers weep. They’re about due for a makeover.
 
 

COMMENTS (15 Comments, most recent shown first)

Trailbzr
J McCann, I agree about losses, and have been toying with another idea about wins. Theory: Starters deserve their losses and one decision per 9 IP; implying pitchers with "too many" decisions will have too many wins, and those with too few decisions will have too few wins.
To test this, I pulled Sean Lahman's database for this calendar decade to look for the starter/seasons with most excess and deficient decisions vs IP/9.

Excess decisions
Player TM/YR I/9 W- L R/Gm TOff EPyth
ROrtiz ANA03 20 16-13 6.05 4.54 07-13
MMarot DET03 21 09-21 6.10 3.65 06-15
JMarqu SLN06 22 14-16 6.30 4.85 08-14
Short Decisions
OPerez LAN04 22 07-06 3.48 4.70 14-08
JGuthr BAL07 19 07-05 4.00 4.67 11-08
KBrown LAN00 26 13-06 2.97 4.93 19-07

I/9 is expected decisions = IP/9
R/Gm includes unearned
TOff is team's offense for whole season
EPyth is Pythagorean W/L based on expected decisions

The excess decision pitchers had 18 excess wins and 8 excess losses.
The deficient decision pitchers were short 17 wins and 6 losses.
So, I agree, losses are assigned better than wins (by about 3:1).

5:49 PM Sep 5th
 
elricsi
I actually love the idea of wins and losses, and I like losses just as they are.

For wins I would actually like to see them divided up among the effective pitchers in the game. I'm not sure exactly how, but maybe something like (outs - runs allowed) / (team outs - runs allowed) and round off to the nearest tenth (when your team wins). So the starter would get like .5, and an effective relieve may get like .1 or .2

Then the starters and relievers can be compared at season's end. I think the totals would be close to what you have now, but at least the starter would always get some credit when his team wins the game.
2:33 AM Sep 4th
 
chuck
Matthew,
Besides using Added Wins (+W) to re-look at borderline HOF=ers, another good use for it would be as an additional stat to consider for the Cy Young debate each year. In this context, I would put Added Losses (+L) into the mix as well: that is, no-decisions where the pitcher leaves in a position to lose and the team later gets him off the hook in either a win or a loss.
As an example, last year's AL Cy Young: Sabathia vs. Beckett.
Sabathia's 19-7 record would have been augmented by two +W's and one +L: 21-8.
Beckett had three no-decisions in which his team trailed when he was taken out. Three +L's would put his record at: 20-10.
3:27 PM Sep 3rd
 
Richie
Ummm, you're the one suggesting MLB ought to change the rule. So if we're going to start throwing epithets like 'serious' around here, I'm rubber you're glue, it bounces off me and sticks to you. Hah.

Having established you're seriouser than I am, excuse me while I trounce you on the '20-game win' thingee. As it becomes much harder to reach, the figure will gain more of an aura. While it simultaneously becomes less of a Hall of Fame marker. It'll be akin to 50 home runs. Folks go "Wow! Now that was a Hall-of-Fame type season!", but nobody suggests "Well, Aaron/Killebrew/Murray never hit 50 homers in a year, so how good could they have been?" 20-game seasons will evolve towards that exact cachet. Except that gradually-increasing numbers of 'H of F' voters will see pitcher wins as kind of a bogus stat anyway.
6:37 PM Aug 31st
 
enamee
Richie, I'm sorry, I just think you're taking this a little too seriously. I did this study for fun, and I wrote the article for fun. Even if MLB changed the win statistic (which is unlikely), they certainly wouldn't apply that retroactively. What I thought was interesting was how much this minor change would affect the way people would perceive certain players. Tommy John and Bert Blyleven would both have 300 wins if the rules were just ever-so-slightly different. Mike Mussina would be a three-time 20-game winner. I just found that interesting.

I also disagree that 20 wins will cease to be a significant figure. People like round numbers; always have, always will. I don't think we'll suddenly start reading articles that say "three-time seventeen-game winner..." Twenty may become rarer, but its aura isn't going anywhere.
3:47 PM Aug 31st
 
chuck
Scratch what I just said. I realized Added Losses wouldn't negate the Added Wins you found. I like the added win idea and wish that it would like to see it get spread to the HOF voters somehow.
1:53 PM Aug 31st
 
chuck
Matthew,
What about the idea of looking at added losses for these pitchers, as well? I'm scratching my head over it a bit, but what about games where the starter is taken out in position to lose the game, and his team takes him off the hook by at least tying the game, if not winning it? It would be interesting to look at those games combined with the added wins, even if you were to do it just for the borderline HOF guys: Blyleven, Kaat, Mussina, John, and Tiant.
1:50 PM Aug 31st
 
Richie
On what basis do you think there's a chance MLB will change the rules along the lines you're proposing?

I'd also suggest you're overstating what '20 vs. 19' will mean going forward. Since pitchers so seldom win 20 anymore, it will lose its importance with respect to future Hall of Fame votes. Those voters are not going to refuse to vote in pitchers from their youth for not meeting a standard they know became incredibly difficult to attain.
8:35 AM Aug 31st
 
enamee
Just to respond to some of the criticism... Whether you like it or not, wins are not going anywhere. People still take them seriously. It's ridiculous that Mike Mussina will go from borderline to shoo-in Hall of Famer if he can manage to get 20 wins this year (but not 19!), but that's just how it is. There's a better chance that MLB will change the rules along the lines I've suggested than that MLB and the BBWAA will all of a sudden be converted to sabermetrism. (Sabermetrism? Why not?) I appreciate that you care about reality, but I'm not really focusing on that. I'm focusing on perception, which can often be a lot more powerful than reality.
12:21 AM Aug 31st
 
monahan
I think this rule change makes sense and is in keeping with the current intent of the rule. Rule 10.17(b) allows the official scorer to credit the win to the relief pitcher that is "most effective" in the event that the reliever on the mound when his team takes the lead has not pitched effectively. It seems a very minor leap to give the Official Scorer the power to credit the starter with the win should his performance be the "most effective."

Of course, the official scorers won't touch 10.17(b) with a ten foot pole right now, so giving them the power to credit the starter may not result in any change.
10:13 PM Aug 30th
 
Trailbzr
Pitchers average about one decision per 9 innings (surprise!), so someone who pitches six inning should get a decision about 2/3 of the time. I would guess that guys who pitch six innings and exit with a five-run lead get a win at least 90% of the time, so the ones who leave with a one-run lead must get one a lot less often.
If the Red Sox had scored their run BEFORE the A's scored theirs, Wakefield would have been credited with the win. So it's 50/50 that a pitcher who exits with a 1-0 lead will win 2-1 game, and 50/50 he'll get no decision. That sounds about right to me.
7:47 PM Aug 30th
 
Richie
'We' don't get to change such measurements. There is negative-zero chance that 'they' will listen to us and change something that's been measured its current way for 100+ years. Only reason to change it would be to make it useful for our own analytical purposes. And to what use would we put such a still-badly-flawed statistic?
3:04 PM Aug 30th
 
evanecurb
Wins and losses are here to stay, flawed or not. It helps if we make the measurement better than it is, while still acknowledging its flaws. I agree with Richie's observations, disagree with his conclusion.
12:20 PM Aug 30th
 
Richie
1. Why bother tweaking a massively-flawed statistic?

2. Any reason, then, for not saddling a starter who leaves the game with his team trailing with an 'added loss' for times when his team wipes out the deficit but still goes on to lose?

3. What does an accounting adjustment have to do with who really deserves or just doesn't quite deserve to go into the Hall of Fame? Tells us nothing about whether Yogi actually was 2.8 times better than the average bear, or not.
12:13 PM Aug 30th
 
evanecurb
Works for me. Let's do it.
1:32 AM Aug 30th
 
 
©2022 Be Jolly, Inc. All Rights Reserved.|Powered by Sports Info Solutions|Terms & Conditions|Privacy Policy