August 21, 2006
Introducing SNLVAR

I know that you guys need another sabermetric statistic introduced to you about as much as you need more Jose Lima in your life, but here I am anyways. I want to explain the usefulness of a statistic that you don’t often see. This particular statistic is a Baseball Prospectus invention, “Support Neutral Lineup-adjusted Value Above Replacement”. Thanks to the wonder of acronyms, we can discard one of the bulkiest names in all of statistics in favor of SNLVAR.

The statistic itself is much simpler than its name suggests. In a nutshell, SNLVAR measures wins above replacement level for a pitcher; it avoids the noise of Wins Above Replacement Level (WARP, another BP invention) because it doesn’t take offense or defense into account. Considering some pitcher’s hit much better – like Dontrelle Willis – or worse – like Brett Myers— than others, it’s good to be able to separate those items out if need be.

Replacement level is measured for SNLVAR the same way it is measured for Value Over Replacement Player (VORP), as explained by Keith Woolner in Baseball Prospectus 2002 and in Baseball Between the Numbers. Essentially, replacement level pitchers make up 20 percent of the total starts in the league, which gives you a figure to compare to the league average. Here’s an excerpt from BBTN with the replacement-level Run Average formula:

“When we look at the remaining 20 percent of starts – those not made by any of the team’s top five starters – and compare their collective performance to the league average, we get the following formula:

Replacement-level RA = 1.37 x League RA – 0.66

…where RA is Run Average, like the familiar ERA concept except that it includes all runs allowed. This formula does a credible job of estimating replacement level throughout the modern era…”

This formula gives SNLVAR the jump point it needs to measure from replacement level in a given season. Not only is SNLVAR measuring from replacement level, but it adjusts for strength of lineups faced. This keeps pitchers who have only faced soft-hitting lineups out of the top of the rankings. The “support neutral” portion simply means that the run support the pitcher receives is not a factor in how many wins they are worth according to the statistic.

You want to think of SNLVAR in the same way as VORP, except in terms of wins. In this way, it is more useful than VORP is for pitchers. Now that the name hopefully seems less bulky – after all, it does explain everything the statistic is showing you, doesn’t it? – let’s move on to an example.

The Mets rotation has had a great deal of overhaul in just the past season, and it also makes a wonderful example for SNLVAR. Take a look at this table with Games Started, Innings Pitched and SNLVAR, starting with the 2005 season:

Pitcher           GS    IP    SNLVAR   
Tom Glavine       33   211.1    5.4
Pedro Martinez    31   217.0    7.6
Kris Benson       28   173.1    3.4
Victor Zambrano   27   166.1    3.0
Kaz Ishii         16    91.0    0.6
Jae Seo           14    90.1    3.5
Aaron Heilman      7   108.0    0.9
Steve Trachsel     6    37.0    0.7

To give you some perspective on how many wins is high, Roger Clemens led the majors in SNLVAR in 2005 with a 9.4 figure, Tom Glavine ranked twenty-second overall with his 5.4 number, and the last score above 0.0 is ranked 227. The worst starter in all of baseball last year was Joe Kennedy, at least in his sixteen Colorado starts. Old friend Jose Lima had the worst figure among starters who stayed in the rotation all year, at -1.1 below replacement level.

As you can see, Glavine and Pedro Martinez (ranked seventh) had rankings in line with their reputations. Kris Benson’s 3.4 SNLVAR is in the same range as starters such as Jason Johnson and Horacio Ramirez; capable back of the rotation guys – at least in 2005 – but nothing spectacular. Jae Seo appears to be the most underutilized portion of the starting pitching resources from the year before, as his prorated SNLVAR would be tops on the staff. His 2006 performance disagrees with that assessment though, so it might not be all bad that he wasn’t used more often. Here’s the same table with the 2006 starters:

Pitcher            GS    IP    SNLVAR   
Tom Glavine        26   160.2    4.0
Steve Trachsel     24   135.1    2.8
Pedro Martinez     20   122.0    3.6
Orlando Hernandez  14    79.1    1.6
Alay Soler          8    45.0    0.6
John Maine          8    50.1    1.9
Brian Bannister     5    28.0    1.1
Victor Zambrano     5    21.1    0.1
Jose Lima           4    17.1   -0.6
Mike Pelfrey        4    21.1    0.2
Jeremi Gonzalez     3    14.0    0.1

(note: figures do not reflect Hernandez’s start on Sunday)

Interestingly enough, the crop is not as strong looking as the 2005 version. This has more to do with the number of injuries on the staff, and the poor quality of replacements at certain points in the season. Alay Soler didn’t add much in his eight starts, and I don’t think I need to discuss how much Lima Time was unappreciated. Geremi Gonzalez and Mike Pelfrey added essentially replacement level performances in their short time filling on the staff, and Victor Zambrano was worth the same as a starter before his season-ending injury. John Maine and Brian Bannister have certainly made themselves valuable to the team in the short time they’ve been there, as has El Duque.

Even though the 2006 version initially looks like the weaker of two staffs according to SNLVAR – only 20.4 SNLVAR in a prorated 162 games versus the 33.4 produced by the 2005 pitchers — the final product of the top five is worth basically the same. The top 5 by SNLVAR in 2005 were worth a prorated 27 SNLVAR, while the 2006 rotation is worth 25. A healthy Pedro would certainly iron out that small difference, and would make October much easier to deal with. And now that Lima and Gonzalez are no longer around, and Brian Bannister and Mike Pelfrey are the most likely replacements, chances are good those figures would improve substantially. To put it bluntly, I wouldn’t worry about the rotation come crunch time.

13 Responses to “Introducing SNLVAR”

  1. Comment posted by Grand Inquisitor Ramon on August 21, 2006 at 7:15 am (#81368)

    I am not sure what this new statistical computation quanitifies or proves that I didn’t know already, but if a comparison is going to be made betwee the 2005 and 2006 Mets pitching staffs then other factors will have to also be considered….

    Most importantly, the upsurge in runs scored and homeruns in this season compared to last needs to be seriously taken into account. For this reason, I would find it more fruitful to measure how the Mets SNLVAR for 2006 compares to the rest of the NL as opposed to how the 2005 team’s did in comparsion to the other NL rotations that season.

  2. Gravatar
  3. Comment posted by Walt on August 21, 2006 at 8:36 am (#81369)

    Interesting way to evaluate pitching. I like the VORP formulae in that their effect is cumulative, as compared to something like ERA which is independent. The more starts you make the more value you add above replacment levels. Example – John Maine seems very low at 1.9. But if he were to sustain his level of performance for 16 starts instead of eight, he would leap to the top of the staff. Or in other words, after just 8 starts, he makes us two wins better than the average AAAA starter, which sounds about right.

  4. Gravatar
  5. Comment posted by Walt on August 21, 2006 at 8:37 am (#81370)

    * Meant to say VORP-style… obviously SNLYVAR isn’t the same thing, but it’s the same kind of thing…

  6. Gravatar
  7. Comment posted by Confused on August 21, 2006 at 8:42 am (#81372)

    how about working on a formula that will guestimate what game the mets will clinch? i want to be there.

    only half kidding.

  8. Comment posted by peeder on August 21, 2006 at 10:43 am (#81385)

    What would be really useful would be a histogram of SNLVAR for each entire league. Ideally, it could be marked up with a mean as well. It would give an idea of how it is skewed from a normal distribution…i.e. the lesser pitchers are carted off to the minors.

    Rankings are good, and ideally, you could have the rank, total and mean when you quote these stats. So Pedro is ranked 7th out of 259, 7.6 with a mean of 2.40. Or whatever it is. See how useful some context is?

  9. Comment posted by cruz on August 21, 2006 at 2:14 pm (#81403)

    this is ugly. looks like we picked a headcase.

    HOUSTON — Former All-Star outfielder Jesse Barfield was taken to a hospital Sunday after he suffered a head injury when he was shoved down a flight of stairs by his 18-year-old son, the Houston Chronicle reported.

    Barfield, 46, was treated and released, but Jeremy Barfield was arrested shortly after 8 a.m., the newspaper reported.
    Jesse Barfield, whose older son Josh plays for the Padres, was known for his strong arm and big bat during his playing days with the Blue Jays and Yankees.

    The younger Barfield, a high school star outfielder, was drafted in the ninth round by the New York Mets in June. According to a spokesperson for the sheriff’s office, Barfield will face a Class A misdemeanor charge of family assault, KTRK of Houston reported. The Harris County District Attorney’s office has accepted that charge, according to the ABC affiliate.

    “[Jeremy] and his father had some kind of argument that apparently took place on the stairs in his house,” Lt. John Martin, a spokesman for the Harris County Sheriff’s Department, told the Chronicle. “The father was lower on [the] stairs than Jeremy was and … at some point during argument, Jeremy shoved him and he ended up falling down the stairs and struck his head pretty hard on the floor.”

    The arresting deputy said the Barfields were getting ready for church, and went to check on what was delaying Jeremy, “and they got into an argument,” Martin said.

    Jeremy Barfield, whose older brother, Josh, plays second base for the San Diego Padres, hasn’t yet signed with the Mets. He signed a letter of intent with San Jacinto College in Texas.

    Jesse Barfield, who played 12 major league seasons with Toronto and the New York Yankees, hit .256 with 241 career homers and 716 RBI. He was an All-Star in 1986, when he won the first of two consecutive Gold Gloves.

  10. Gravatar
  11. Comment posted by Marc Normandin on August 21, 2006 at 3:20 pm (#81406)

    Most importantly, the upsurge in runs scored and homeruns in this season compared to last needs to be seriously taken into account.

    It is taken into account. SNLVAR is lineup-adjusted, so it’s all relative and shifts along with the replacement level for each year.

    What would be really useful would be a histogram of SNLVAR for each entire league. Ideally, it could be marked up with a mean as well. It would give an idea of how it is skewed from a normal distribution…i.e. the lesser pitchers are carted off to the minors

    I was thinking the same type of thing as I was writing this. I’m sure that some evening when I’m bored I’ll put it together. I’d probably do it after the season so I could work the kinks out and figure out how to make it update automatically.

    #3-#4: I agree that it’s useful in that you can see just how much value has been added, and you can always prorate it to see how much could have been added, as I did in the article and you just did with Maine.

  12. Comment posted by Wally Dykstra on August 21, 2006 at 10:54 pm (#81462)

    No post since 3:20…can’t recall it being so quiet.

  13. Comment posted by Wally Dykstra on August 21, 2006 at 10:57 pm (#81466)

    Now Ramon Castro has blown out his knee stretching. Stretching? Somebody’s put a voodoo hex on this team.

  14. Comment posted by Grand Inquisitor Ramon on August 22, 2006 at 6:52 am (#81495)

    It looks like Tommy is Terrific! Yes! Don’t pack it in yet guys, the rotation will be alright.


    I am happy for Glavine and his family!

  15. Comment posted by Grand Inquisitor Ramon on August 22, 2006 at 6:56 am (#81496)

    Here it is. Sorry.

  16. Comment posted by Nick in Westchestah on August 22, 2006 at 9:22 am (#81503)

    Awesome–thanks for the great news Ramon!!!!

  17. Comment posted by bmc on August 31, 2006 at 11:37 am (#88416)

    SNLVAR ranks the Mets as not the best in the league.

    Therefore, SNLVAR is bad.