How Not To Use Park Factors

Dave · September 16, 2010 at 11:03 pm · Filed Under Mariners 

A few days ago, Baseball Reference updated the park factors they use on their site to include data from 2010. This caused some pretty decent sized changes in some of the stats that they host that are park adjusted, including their version of Wins Above Replacement. Most notably, the addition of the park factor data caused CC Sabathia to overtake Felix Hernandez for the AL lead in their version of pitcher WAR, as they’ve significantly upped the offensive level for New Yankee Stadium and significantly lowered it for Safeco Field. As Sean Forman noted in his post announcing the move, the data indicates that shas “become more of a pitcher’s park”.
en
However, when you look at the data for Safeco this year, you’ll see some interesting things.

At home, the Mariners have hit .236/.302/.325, and on the road they’ve hit .232/.294/.352. The plate appearances are almost exactly equal, so we can actually just compare straight up numbers from the M’s offense, home and road.

Singles: +33
Doubles: -6
Triples: +1
Home Runs: -25

Non-HR extra base hits have basically been the same. The team has more singles at home, but fewer home runs. That translates to slightly less offense, but not that much. The offense’s overall production suggests that Safeco has played as a slight pitcher’s park, but not an extreme one.

On the pitching side of things, it’s another story though. Mariner pitchers have held opposing hitters to a .235/.294/.361 line at Safeco, while giving up a .274/.333/.440 mark on the road. Despite facing 140 more batters in Safeco, they’ve given up 32 fewer singles and 21 fewer home runs. The pitching staff has been helped quite a bit by Safeco, which is the main reason that the park is appearing to be an extreme run suppressor. But, remember, Safeco plays very differently for left-handed and right-handed pitchers, and the Mariners built their pitching staff to take advantage of that fact. And, not surprisingly, the vast majority of the difference in performance at home and on the road can be attributed to those lefties.

At Safeco:

Ryan Rowland-Smith: 222 PA, .263 BABIP, 2.25% HR/PA
Jason Vargas: 401 PA, .254 BABIP, 2.00% HR/PA
Luke French: 148 PA, .222 BABIP, 2.70% HR/PA

On Road:

Ryan Rowland-Smith: 261 PA, .344 BABIP, 7.28% HR/PA
Jason Vargas: 333 PA, .287 BABIP, 3.00% HR/PA
Luke French: 146 PA, .272 BABIP, 2.74% HR/PA

Those three are driving almost all of the difference in performance of the M’s pitching staff at home versus on the road. Remember how we said that the team had given up 21 more home runs on the road than at home? Ryan Rowland-Smith has given up 14 more away from Safeco by himself – he is 67 percent of the difference in team home run rate.

The Mariners have intentionally loaded up on soft-tossing flyball lefties because Safeco Field is death to right-handed power hitters, and those pitchers can exploit that difference by letting long fly ball outs get tracked down in the left center field gaps.

Know who can’t do that? Right-handed pitchers. It’s actually pretty easy to hit a ball out to right in Seattle, so left-handed hitters have few problems pulling balls over the wall. That’s why RHPs rarely exhibit home/road splits at Safeco that are anything close to what LHPs offer.

Felix’s home/road splits, by the way? His BABIP is four points LOWER on the road, and his HR/PA is 1.39% at Safeco compared to 1.87% on the road. To translate that into actual numbers based on his PAs, if his home and road HR rates were equal, Felix would have given up an additional two home runs in Safeco this year, going all the way from six to eight.

However, because blanket park factors make no attempt to correct for how differently parks play based on the handedness of the player, we’ll now get to see people making claims about Felix benefiting dramatically from the extreme pitcher’s park that he calls home, ignoring the fact that it is not an extreme pitcher’s park on the days that he takes the hill because he is not left-handed.

The sabermetric community currently uses one-size-fits-all park factors right now only because, to this point, we’ve been too preoccupied with other things to actually put in the time and effort it takes to do them right. In a few years, once we’ve applied component park factors based on handedness to each player’s performance, we’ll look back and laugh at some of the conclusions that using one park factor for every player forced upon us. One of the things we’ll laugh at is the notion that Felix Hernandez got some dramatic benefit from Safeco Field in 2010. He didn’t, and we shouldn’t act like he did.

Comments

18 Responses to “How Not To Use Park Factors”

  1. dnc on September 16th, 2010 11:41 pm

    If we’re trying to measure the value of his contribution to his team then it doesn’t matter that the park is a poor fit for him, correct?

    To clarify, I’m not at all advocating for CC over Felix (any metric that values the park factor that significantly is clearly overdoing it). However, in general, shouldn’t we ignore the component park effect when assessing the value a player contributed to his club, while using component park effects to assess a player’s true talent? And since WAR is supposed to be a value based counting stat, it should not incorporate the component effect.

    I seem to remember this type of issue coming up when Larry Walker didn’t post much of a home/road split and everyone said he deserved the MVP because Coors wasn’t helping him out. The response was that, just because Walker’s skill set didn’t exploit the advantage of Coors (at least that season) didn’t mean that the advantage doesn’t exist. I would assume the same applies to Felix El Rey in regards to the Safe.

    This is how I’ve always understood it anyway, but I’m certainly open to being wrong.

  2. johnfree63 on September 17th, 2010 2:24 am

    So in some weird way the M’s completely sucking at home (not that they can hit on the road) is actually an advantage to Felix?

  3. justinh on September 17th, 2010 2:39 am

    In terms of pitching and defense, WAR is still far from perfect. When Jose Lopez has the third best WAR for all defensive players, and it is mid-September, you know not to trust UZR completely. Felix is just in a different class when it comes to statistics, because a pitcher with his ability can bear-down anytime and get someone out. There are so many intangibles, and it makes it very difficult to correctly analyze Felix. When a pitcher is so dominant, it doesn’t matter if he pitches on Mt. Rainier. I liken it to Bonds in early 00s, it didn’t matter where he played because he would walk or hit the ball 450 feet. For the average player, park factors are legit, but there are 4-5 guys that it doesn’t make a lick of difference because they are so dominant.

    Hope that makes sense, it is 230AM and I have been up for 22 hours, but had to check USSM before I hit the sack.

  4. lesch2k on September 17th, 2010 5:33 am

    i wish the baseball writers who need to see this information would be able to consider it in their analysis. something tells me they won’t see this though, and if they do see it they wont give it a fair/critical analysis. after all, the easiest way to counter logical arguments is with outrageous claims :)

  5. Dave on September 17th, 2010 6:47 am

    If we’re trying to measure the value of his contribution to his team then it doesn’t matter that the park is a poor fit for him, correct?

    This is mostly true for hitters, but not for pitchers, because while hitters face a blend of different pitchers at their home park, a pitcher is always right or left-handed.

    Safeco does not have the same run environment every day. It changes, depending on what kind of pitchers are on the mound, due to the asymmetric dimensions and the wind patterns. Over a five day stretch, it’s park factors may look like this:

    100
    85
    85
    100
    85

    On any given day, the park factor is never 91, but that’s what we use because it’s the average of the five. That’s a problem. We’re misrepresenting the park’s effects on everyone by trying to use a one-size-fits-all model.

  6. jonahkeri on September 17th, 2010 7:54 am

    Yet another example of Strat-O-Matic players being years ahead of the curve. By my count that’s defense, baserunning, platoons, bullpen usage and adjusted park factors that Strat had right years before most sabermetric analysts had it (not impugning you, Dave).

    Seriously, if you want to write about baseball, manage a baseball team, or just be a more informed fan, follow the old Bill James advice: Bang out a few hundred games of Strat.

  7. dcmarinerfan on September 17th, 2010 8:29 am

    Non-HR extra base hits have basically been the same. The team has more singles at home, but fewer home runs. That translates to slightly less offense, but not that much. The offense’s overall production shows that Safeco has played as a slight hitters park, but not an extreme one.

    Did you mean pitchers?

  8. todmod on September 17th, 2010 9:17 am

    I’m also curious why park factors for WAR should be adjusted for pitcher handedness. While that may help you better determine Felix’s skill level, does the adjustment really indicate his value?

    Wouldn’t a replacement level pitcher at Safeco include left handers that could perform better with the park dimensions? So wouldn’t this raise the value of the replacement player he’s measured against in theory?

    Since any team can put a left handed pitcher to oppose Felix on any given day, it seems that using overall park factors is the best measurement for true value. It comes down to whether you want to measure a player’s skill level or value level. Which one is the Cy Young award honoring? And which one is WAR supposed to be measuring?

  9. JMHawkins on September 17th, 2010 9:26 am

    The sabermetric community currently uses one-size-fits-all park factors right now only because, to this point, we’ve been too preoccupied with other things to actually put in the time and effort it takes to do them right.

    I’m actually kind of surprised at this. L/R splits is one of the oldest “innovations” in baseball statistics, and it’s not like SafeCo is the first park to ever play asymmetrically. (“The Green Monster, altering results for RHB since 1912″).

  10. JEH on September 17th, 2010 10:07 am

    Curiously, I believe, the opposite situation has occurred in New York.

    Sabathia, as a lefty, can take advantage of Yankee Stadium in a way the rest of the rotation cannot. Sabathia has, far and away, the best home/road split among Yankee pitchers and so the 2010 Park Factor for Yankee Stadium indicates a degree of difficulty in pitching there that a lefty does not have to contend with.

  11. Matthew Carruth on September 17th, 2010 10:37 am

    I don’t know about perfectly right, but we haven’t all been too preoccupied

    Handed park factors

    Safeco

  12. Rick Banjo on September 17th, 2010 11:10 am

    So what does this mean for the future of right-handed power hitting in Seattle? I guess the argument could be made that Boone has been our best RH power hitter since the move from the dome, but his splits in his monster year weren’t extreme:

    Splits, 2001

    Great exegesis on how it relates to our pitching staff, though. I love this blog.

  13. Shanfan on September 17th, 2010 11:21 am

    I’m in over my head here, but why are only three years of data used? If the dimensions haven’t changed and the surrounding structure hasn’t been altered significantly, wouldn’t more years/data make for a better model of a park? Safeco was less pitcher friendly when it was Edgar and Boonie instead of Lopez and Guti in the middle of the order. I’m asking honestly out of ignorance and am seeking illumination, not an argument.

  14. Westside guy on September 17th, 2010 12:29 pm

    I’d like to offer an executive summary for the casual fan or the vote-wielding baseball writer.

    Felix Hernandez is a stud!

  15. Matt Staples on September 17th, 2010 12:29 pm

    Fantasy baseball players have been using park factors adjusted for handedness, and isolated for particular events (e.g., HR, K, BB, XBH, etc.) for years. I have, at least. A number of services publish them. Not to put words in his mouth, but I believe Dave isn’t saying that component (or otherwise advanced) park factors are available but, rather, that advanced metrics such as WAR do not presently utilize them.

  16. Breadbaker on September 17th, 2010 1:40 pm

    This is exactly why anyone voting on a baseball matter, be it MVP, Cy Young or Hall of Fame, should exercise some independent judgment and not just rely blindly on some statistic, any statistic. Look at the difference between Bill James’ top ten players at each position in his two historical abstracts, you’ll see tremendous differences. Statistical analysis of baseball is constantly evolving. That is no excuse for ignoring what had been learned so far, but awards are forever while this year’s statistical analysis is not as good as next year’s.

  17. dnc on September 17th, 2010 2:04 pm

    I’m in over my head here, but why are only three years of data used? If the dimensions haven’t changed and the surrounding structure hasn’t been altered significantly, wouldn’t more years/data make for a better model of a park?

    My guess is that factors such as climate, prevailing winds, hitter’s eye, groundskeeping, etc do not actually remain constant over time but are constantly evolving.

    Also, park factors are supposed to measure the effect of the ballpark on run scoring relative to the other parks in the league. Just because those factors all remain static at Safeco for a period of time doesn’t mean they do everywhere else. When Target Field is opened and suppresses runs compared to its predecessor, Safeco looks a little less pitcher friendly relative to the league average without even changing itself, whereas when New Yankee opens and increases runs relative to its predecessor, Safeco looks a little more friendly to hurlers relative to the league.

    This is apparent on a macro level (new ballparks) but is constantly occurring on a micro level as tiny changes in wind, weather, batters eye, dimensions and groundskeeping at all the parks affect the overall run scoring environment of the league.

  18. NorthofWrigleyField on September 17th, 2010 4:53 pm

    Felix Hernandez has pitched only 15 times at home this year, so far. It really wasn’t yeoman’s work to look at the games and look over his PAs against LH batters. He’s given up a grand total of 6 XBHs to LH batters in 15 home starts. 3 HRs, 3 2Bs… that’s it. He’s given up 18 XBHs to LH batters on the road.

    Opponents have loaded lefties into their lineup when Felix takes the Safeco Field mound, and they’ve mostly failed – Felix was wild like he was against Texas and LA early in the year (the righties killed him pretty bad in those games too).

    Sure, lefties hit Felix better in Safeco than righties. But bad is still bad.

Leave a Reply

You must be logged in to post a comment.