Making Win-Loss Record a More Useful Stat

We've all heard it during broadcasts or recap shows. After the 2009 season, many broadcasters, especially the "old school" majority, dismissed Zack Greinke as an AL Cy Young candidate because he only won 16 games, compared to Felix Hernandez and C.C. Sabathia's 19 wins apiece. Was it Greinke's fault that he pitched for a 65 win Kansas City squad, while Sabathia pitched for a 103 win Yankees team? Of course not, but many people were convinced that Sabathia had a better season nonetheless. And don't even get me started on Bert Blyleven...

At this point, I don't really need to rehash the many arguments against using win-loss record as a primary indicator of a pitcher's performance. There are too many variables, offensive run support, having a decent bullpen to lock down leads, etc to say too much about a pitcher's performance compared to other metrics like earned run average, wins above replacement and win probability added. But what if we could examine a pitcher's win-loss record in a way that we normalize for run support and bullpen? Can we determine how many wins Sabathia and Greinke should have expected based on the number of runs they gave up, or how many they could expect based on the run support provided by their respective offenses? And most importantly, how many wins their teams should have had during their starts?

As you might suspect, I've run the numbers for both the 2008 and 2009 seasons. I'll provide much more detail and explanation after the jump, but here's how Greinke and Sabathia graded out in 2009, with the Twins starters as a comparison. As you can see, based on expected win-loss record, Greinke was over three wins better than Sabathia, which is in line with the difference in WAR between Greinke's +9.4 and Sabathia's +6.0 last season. And when one looks at the team's expected win-loss record, the gap is a bit smaller, but again in line with the difference in WPA/LI (+5.40 Greinke, +3.06 Sabathia).  On the Twins side, Scott Baker and Kevin Slowey didn't pitch as well as their 15-9 and 10-3 records indicate, and Nick Blackburn pitched better than his 11-11 record.

If you're interested, the entire 2008 and 2009 spreadsheet is posted to Google Docs here.

Player GS W-L Expected W-L Team W-L Exp Team W-L
Zack Greinke 33 16-8 18.5 - 7.1 17-16 21.4 - 11.6
C.C. Sabathia 34 19-8 15.3 - 10.5 22-12 19.3 - 14.7
Scott Baker 33 15-9 12.6 - 11.2 20-13 17.2 - 15.8
Nick Blackburn 33 11-11 13.8 - 11.3 16-17 17.5 - 15.5
Francisco Liriano 24 5-13 6.1 - 11.0 9-15 7.8 - 16.2
Glen Perkins 17 6-7 5.2 - 8.2 8-9 7.2 - 9.8
Kevin Slowey 16 10-3 4.6 - 6.0 10-6 7.3 - 8.7
Anthony Swarzak 12 3-7 2.9 - 5.7 4-8 4.7 - 7.3
Brian Duensing 9 5-1 3.9 - 2.2 6-3 5.1 - 3.9

Method

In order to determine "expected" win-loss records, I first needed to establish a historical baseline average for the 2008 and 2009 seasons. Using my "Total Run Accounting" software, I analyzed every start from the two seasons, capturing the number of wins and losses (both for the starter and for the team) and no decisions based on the number of innings pitched (I only used full innings completed in order to avoid small sample sizes) and runs allowed by the starter. As a result, I got data like this (2009 AL data):

INN RUNS PIT W PIT L PIT ND TM W TM L
9 1 24 1 1 25 1
8 2 23 3 3 24 5
6 3 74 45 49 94 74
5 4 22 49 50 46 75
4 5 0 28 27 15 40

 

These numbers shouldn't be surprising, since the only way a pitcher could lose pitching nine innings and giving up one run is to be on the wrong side of a 1-0 shutout. I found it interesting that despite all of Bert Blyleven's bluster about how the "quality start" is such a useless stat since 6 IP - 3 ER means a mediocre 4.50 ERA, the worst possible quality start still gives a team a 56% chance of winning the game. To put it another way, given an AL average offense, bullpen and luck, if a team were to get a 6 inning / 3 run start from their pitcher every time out, you would be looking at a 90-91 win team. But more on quality starts another time.

Because of the roughly half a run difference between run production between the American and National Leagues, I broke out the data between the leagues, using combined data for interleague games. For example, in the National League, a 6 IP / 3 R start simply doesn't get you as far, as pitchers went 55-57-58 (W-L-ND) and teams went 81-89 in these starts. This makes sense, since NL offenses would provide less run support over those six innings than in the AL.

Once I established the baseline, I went through each start again, assigning "expected" wins, losses and no decisions to each pitcher, accumulating over the entire season. So for an AL pitcher making a 6 IP-3R start, he would receive 0.44 expected wins (74 wins in 168 starts) and 0.27 expected losses (45 losses in 168 starts based on that start, with the remainder being "expected no decisions". This is how I calculated an "expected" win-loss record of 18.5 - 7.1 for Zack Greinke in 2009, or a major league best 20.6 - 7.8 expected record for C.C. Sabathia in 2008. 

"Expected" Win-Loss Record

Here are the top five pitchers from 2009 in "expected wins".

Pitcher Expected W-L
Zack Greinke 18.5 - 7.1
Roy Halladay 17.2 - 9.4
Felix Hernandez 16.9 - 9.8
Cliff Lee 16.4 - 9.8
Adam Wainwright 15.7 - 8.4

And for 2008:

Pitcher Expected W-L
C.C. Sabathia 20.6 - 7.8
Cliff Lee 16.7 - 7.4
Johan Santana 16.6 - 8.1
Roy Halladay 16.2 - 8.9
Tim Lincecum 15.6 - 8.0

 

Of course, the leaders are the best aces in the majors, and while there is a correlation with the WAR leader boards, it's not as strong as one would expect. In particular, Johan Santana 2008 (#18 in WAR) and Adam Wainwright 2009 (#11 in WAR) appear to be outliers on the list. Otherwise, everyone is top five on both the WAR and expected wins leader boards. A quick glance at the WAR leader boards suggests that total innings pitched isn't the culprit, so I suspect Wainwright and Santana may have outperformed their WAR by pitching more consistently than the other pitchers, but I can't be sure.

We can't just look at the best pitchers though. The bottom three in "expected losses" in 2009 were Braden Looper (9.0 - 14.4), Jeff Suppan (6.9 - 13.6) and Derek Lowe (9.9 - 13.6). In 2008, the bottom three were Brandon Backe (7.5 - 14.9), Zach Duke (8.1 - 14.4) and Livan Hernandez (9.1 - 14.4).

Luck Versus Fortune

Now that we've calculated an "expected" win-loss record for each player, we can compare the expected record to actual win-loss record. Who outperformed and underperformed their expected wins and losses by the most?

Top five "over" performers relative to expected wins, 2009:

Pitcher W-L Expected W-L W - Exp W
Jorge de la Rosa 16-9 10.2 - 11.5 +5.8
Joe Saunders 16-7 10.3 - 12.6 +5.7
Kevin Slowey 10-3 4.6 - 6.0 +5.4
Derek Lowe 15-10 9.9 - 13.6 +5.1
Braden Looper 14-7 9.0 - 14.4 +5.0

 

Top five "over" performers relative to expected wins, 2008:

Pitcher W-L Expected W-L W - Exp W
Brandon Webb 22-7 14.5 - 10.4 +7.5
Daisuke Matsuzaka 18-3 11.4 - 8.7 +6.6
A. J. Burnett 18-9 11.7 - 12.6 +6.3
Mike Mussina 20-9 13.8 - 10.4 +6.2
Ted Lilly 17-9 10.9 - 12.2 +6.1

 

As expected, the leader boards for the most part includes pitchers on good offensive teams (de la Rosa, Matsuzaka, Mussina) or with excellent bullpens and closers (Saunders, Slowey). The one exception appears to be Derek Lowe, whose Atlanta offense was not very good last year.

Top five "under" performers relative to expected wins, 2009:

Pitcher W-L Expected W-L W - Exp W
Matt Garza 8-12 12.3 - 10.8 -4.3
James Shields 11-12 14.4 - 9.9 -3.4
Josh Geer 1-7 3.9 - 7.6 -2.9
Aaron Harang 6-14 8.9 - 9.2 -2.9
Nick Blackburn 11-11 13.8 - 11.3 -2.8

 

Top five "under" performers relative to expected wins, 2008:

Pitcher W-L Expected W-L W - Exp W
Greg Smith 7-16 12.1 - 10.0 -5.1
Jason Bergmann 2-11 6.4 - 9.6 -4.4
Matt Cain 8-14 12.1 - 12.1 -4.1
C.C. Sabathia 17-10 20.6 - 7.8 -3.6
Jarrod Washburn 5-14 8.3 - 9.4 -3.3

 

And the leading "under" performers mostly came from teams with poor offenses, Oakland, Washington, San Francisco and Seattle.

 

Normalizing for Run Support

We would expect "fortune" (playing for a good offensive team or a solid bullpen to protect leads) and luck to play a part in a pitcher's win-loss record. But how can we determine how much is due to luck and how much is due to fortune? As I've noted above, two obvious components of a pitcher's win-loss record that cannot be attributed to random luck are run support and the ability of a bullpen to hold on to leads. Of the two, I suspect the more significant component is run support. Can we isolate the effect of a pitcher's run support on his win-loss record? Yes, we can. Using the same method as I described above for generating an expected wins baseline, I also analyzed every start from 2008 and 2009, examining how run support independent from a pitcher's runs allowed, contributes to a pitcher's expected wins. So if a pitcher goes seven innings, I determine run support from that pitcher's offense (seven innings if he's the home starter, eight innings if he's the away starter). And then a accumulate over the entire season to determine how likely a pitcher is to win a game if he pitches seven innings and gets three runs of support (41% AL), five innings with one run of support, etc.

Examining a pitcher's expected wins and losses based on run support helps us to determine how much of a pitcher's record is due to giving up a low number of runs, and how much is due to the run support he receives. A few examples help illustrate how we can use this data.

Pitcher W-L Exp W-L (runs allowed) Exp W-L (run support)
Cliff Lee (2008) 22-3 16.7 - 7.4 21.6 - 3.5
Livan Hernandez (2008) 13-11 9.1 - 14.4 14.3 - 9.4
Cliff Lee (2009) 14-13 16.4 - 9.8 17.2 - 9.4
Nick Blackburn (2009) 11-11 13.8 - 11.3 15.0 - 9.3


In 2008, Cliff Lee had a once in a decade type of career year, going 22-3 and winning the AL Cy Young. But based on the runs he allowed, his expected record was 4-5 wins and losses lower. So why did he go 22-3? Look at the run support Lee received. If Lee was simply a league average pitcher who managed to pitch as deep into games as Lee did in 2008 (admittedly, deep into games), we see an expected record very close to his actual record. For Lee, we can say his record was probably more a result of his run support than his performance (although 17-7 is still a very good record). We also see that Livan Hernandez benefited from a great deal of run support in 2008. In 2009, we saw Cliff Lee, and to a lesser extent, Nick Blackburn under perform both expected win-loss records, suggesting that their .500 records may have been due to bad luck (Blackburn) or a poor bullpen (Lee - Cleveland's pen was a mess last year).

Putting It All Together: Expected Team Win-Loss Record

Looking at a pitcher's win-loss total is fine, but in the end we want to know how the numbers relate to the pitcher's team winning the game. Using the same method as for the expected pitcher win-loss records based on runs allowed or run support, I accumulated every pitcher's expected team win-loss record in his starts based on runs allowed/support. This gives us an idea of how many wins a team should expect with that pitcher on the mound. Over the last two seasons, the AL and Minnesota leaders were:

Pitcher Team W-L Exp Tm W-L (allowed) Exp Tm W-L (support)
C.C. Sabathia (2008) 22-13 23.7 - 11.3 23.9 - 11.1
Nick Blackburn (2008) 15-19 15.2 - 18.8 16.7 - 17.3
Zack Greinke (2009) 17-16 21.4 - 11.6 19.6 - 13.4
Nick Blackburn (2009) 16-17 17.5 - 15.5 19.2 - 13.8


At first glance, I was surprised that the very best pitchers by themselves provide around a .650 - .700 expected team winning percentage. It seemed a bit low, but 10-11 games over .500 is really a five game swing from average, in line with WAR calculations for the league leaders. For the Twins, even though Nick Blackburn's overall ERA was pretty much the same in 2008 and 2009, his expected team win-loss record improved by 2-3 wins last year, suggesting an improvement beyond what his overall numbers tell us.

Conclusions

To sum up, by looking at a pitcher's expected wins and isolating the effect that his runs allowed and run support has on his record, we can better understand his performance and how much he really helped put his team in position to win games. And at the end of the day, that's what it is all about. Winning ballgames.

X
Log In Sign Up

forgot?
Log In Sign Up

Please choose a new SB Nation username and password

As part of the new SB Nation launch, prior users will need to choose a permanent username, along with a new password.

Your username will be used to login to SB Nation going forward.

I already have a Vox Media account!

Verify Vox Media account

Please login to your Vox Media account. This account will be linked to your previously existing Eater account.

Please choose a new SB Nation username and password

As part of the new SB Nation launch, prior MT authors will need to choose a new username and password.

Your username will be used to login to SB Nation going forward.

Forgot password?

We'll email you a reset link.

If you signed up using a 3rd party account like Facebook or Twitter, please login with it instead.

Forgot password?

Try another email?

Almost done,

By becoming a registered user, you are also agreeing to our Terms and confirming that you have read our Privacy Policy.

Join Twinkie Town

You must be a member of Twinkie Town to participate.

We have our own Community Guidelines at Twinkie Town. You should read them.

Join Twinkie Town

You must be a member of Twinkie Town to participate.

We have our own Community Guidelines at Twinkie Town. You should read them.

Spinner.vc97ec6e

Authenticating

Great!

Choose an available username to complete sign up.

In order to provide our users with a better overall experience, we ask for more information from Facebook when using it to login so that we can learn more about our audience and provide you with the best possible experience. We do not store specific user data and the sharing of it is not required to login with Facebook.

tracking_pixel_9351_tracker