2013 Park Factors Updated

We’ve updated the site with 2013 park factors, which you can see on our FanGraphs Guts! page. These will be updated daily for the rest of the season and we will have the more granular park factors for 2013 updated shortly (on a weekly basis) as well.

To explain a little more about our park factors, we use up to 5 years of data to calculate park factors using Patriot’s method. When a park makes significant changes to their layout, or it is just a brand new park, we use only data from the changed or new park.

In the case of Safeco Field and PETCO Park, which had significant fence movement, we are only using 2013 data to calculate the park factors. With just this season’s worth of data, the park factors for these parks are being heavily regressed (90% towards 100, a neutral park).

With 5 year park factors, we use the surrounding two years of data when possible. And as a result, it is possible that park factors from up to 5 years ago will have some very slight changes, impacting wRC+, WAR, ERA-, and other park adjusted stats.

Here’s an example of which seasons of data are currently included for each season’s park factor of the new Yankees Stadium that entered use in 2009:

Years Included for the new Yankees Stadium in 2013
PF Year Years of Data Included
2009 2009-2013
2010 2009-2013
2011 2009-2013
2012 2009-2013
2013 2009-2013

In 2014, it will look like this:

Years Included for the new Yankees Stadium in 2014
PF Year Years of Data Included
2009 2009-2013
2010 2009-2013
2011 2009-2013
2012 2010-2014
2013 2010-2014
2014 2010-2014

With the rolling nature of the 5 year park factors at least the most current 3 years of park factors, assuming there hasn’t been a new stadium, will be identical.

Print This Post

David Appelman is the creator of FanGraphs.

14 Responses to “2013 Park Factors Updated”

You can follow any responses to this entry through the RSS 2.0 feed.
  1. bjoak says:

    Wouldn’t it make sense to regress parks where fences were moved to their original park factor rather than league average. Moving fences tends to effect HR park factors but not so much run park factors. The reason is obvious: if you have closer fences you have more home runs but less of everything else because it is easy for the outfield defense to get to everything in a smaller park. Good examples of this are the Mets last year and the Padres last time they moved.

    I’d like to also add that partial season park factors are prone to error because of the differing levels of competition teams may have faced at home and away.

    +10 Vote -1 Vote +1

    • MGL says:

      No. Within reason, if you move back the fences (or increase the height), you will decrease the run factor as well as the HR factor, although obviously it will impact the HR factor more than the run factor. And vice versa if you move the fences in.

      Now, you still definitely don’t want to regress toward average for any park, and certainly not parks that have only changed something, if you can help it.

      For example, let’s say that you have a park that has a previous regressed run factor of 1.03 based on several years of data and you move the fences in (granted, when you move fences in, a park was likely to have been a pitcher’s park). You expect the run factor to go up so you certainly don’t want to regress the new data towards 1.00. That would make no sense. You would want to regress it at least toward 1.03 if not 1.04 or 1.05.

      Even for a park that has been around a long time but has not changed. Surely you know some things about a park that would allow you to regress the park factors toward a number other than 1.00. For example, Chase Field in Arizona is small, hot, and at a altitude. You don’t need any data to know that it will likely be a hitter’s park.

      Similarly, in Oakland, the park is large, cold, at sea level, and with an enormous foul territory. Even without any data, you are pretty sure you have a pitcher’s park.

      So, no, regressing every park toward 1.00, especially when changes are made, but you can infer what the new park is going to be like from the data from the old park and the changes that were made, is a mistake. It will lead to very watered down (and likely incorrect) park factors.

      Personally, I establish regression values (the means toward which to regress) from the size, altitude, temperature and foul territories of each park. And if a park has been around for a long time but makes some changes, I use the old numbers, the new numbers, and some manual “tweaking” to estimate the new numbers.

      For example, SEA had a run factor prior to this year of around .90. They made some pretty severe changes in moving the fences in. So, let’s say that the new run factor for half a season is .88. Well, that tells me nothing. It slightly suggests that the new changes were not all that extreme. So I might make the new PF .92 or so. If the sample PF for 2013 were .98, again, that is a slight suggestion that the changes were extreme, so I might make the new PF around .94 or .95. Let’s say that the new sample PF for half a season is 1.09. I am certainly not going to regress that 90% toward 1.00 and say that the new “true”PF is 1.01. While the 1.09 suggests that the changes were extreme, I know that the park is still going to be a pitcher’s park and I know that in half a season, almost any number can occur regardless of a park’s true PF.

      So, no, I don’t really like the way FG is doing their PF’s.

      I am also not sure how they calculate the sample numbers. I don’t know how Patriot does it off the top of my head. I would hope that they/he take into consideration the unbalanced schedule in doing the “away” numbers.

      Vote -1 Vote +1

      • bjoak says:

        You’ve thought a lot about this but in my opinion you’re relying a little too much on conjecture. You seem to be saying, the park factor should be this so I’m going to regress it to what I think it should be. I think parks are sometimes counter-intuitive. A good example is the Marlins park for which you might have said a couple years ago, “You don’t need any data to know it’ll be a pitcher’s park. It has almost identical weather, dimensions, and altitude as PETCO.” Going on two years it’s almost exactly neutral.

        It’s all well and good to say, “you will decrease the run factor,” but the evidence does not bear that out. When Detroit moved the fences the factor did not change. Citi Field? The factor did not change. PETCO? Twice now and if you’re to believe the unregressed PF for 2013, nothing has changed. I’ll grant you that I don’t have a specific study to point out, but I think the anecdotal evidence suggests that we at least need one before we can just assume there will be a difference.

        Vote -1 Vote +1

  2. Josh M says:

    tripod still exists???!?

    Vote -1 Vote +1

  3. Baltar says:


    Vote -1 Vote +1

  4. MGL says:

    Interestingly, in looking over their PF’s for 2013 and comparing them to mine, the ones that stand out the most are the Mariners and Padres. They have the Mariners at 1.03 after being .94 in 2012. It is unlikely that the changes made it go from an extreme pitcher’s park to a hitter’s park. That numbers for 2013 are likely a fluke.

    Same thing with Petco.

    The other parks that I have very different numbers for are Cubs, D-Backs, A’s, and Rockies.

    Vote -1 Vote +1

  5. JKB says:

    Has anybody ever asked what specific outcomes park factors are supposed to predict, and then done a predictive validation study to compare the different park factor formulas and see which is the most predictive in terms of those specific outcomes? I would love to read that article…

    Vote -1 Vote +1

  6. oldschoolways says:

    Will they be updated by handedness soon too?

    Vote -1 Vote +1

  7. JKB says:

    How do the park factor formulas statistically adjust the home team sample size down, or adjust the visiting team sample sizes up, when creating the park factor indices (so that the home team does not overly influence the park factor for their stadium)? Do they scale each team to a 10 game season in each park they play in or something like that?

    Vote -1 Vote +1

    • bjoak says:

      I think you’re just asking how they’re calculated. Let’s say you want the Wrigley Field park factor. You take all Cubs and opponent numbers at Wrigley vs. all Cubs and opponent numbers away from Wrigley. You’re only using the Cubs games so away teams will have the same level of competition.

      Maybe more to your point, the Cubs might face the Phillies three times at home and six times away, in which case as far as I know no one adjusts for that. As I mentioned above, partial season PF’s aren’t that good because those numbers are likely even more skewed, but with the amount of league-wide competition over 162 games most of that is going to come out in the wash.

      Vote -1 Vote +1

  8. johnnycuff says:

    while we’re on the subject, can anybody explain this disclaimer on the page means:

    “All Park Factors have already been halved for use on full season stats.”

    Vote -1 Vote +1

    • AC_Butcha_AC says:

      That just means that these numbers can be applied to the full season stats of a player X on team Y.

      Assuming that half of the games are played at home and half of the games away. For home games, the home park factor applies, obviously. Away park factors tend to even out to an absolute “neutral” park.

      Yankee Stadium has a HR factor for lefties of 136. Player X plays for the Yankees, bats L and hits 30HR for the season. Half of his games he played at Yankee Stadium, half of his games he played away. Away park factor for leftie HR is 100.
      So he should see an increase of 36% of his HR total for all home games, but no increase at all on away games.
      So he should expect to see his HR total increase by 18% or exactly half of the increase for his home games over the full season.

      hope this helps

      Vote -1 Vote +1

Leave a Reply

Your email address will not be published. Required fields are marked *

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <strike> <strong>