The Home/Road Splits

As most of you are probably aware, player pages now feature splits. As such, we’re beginning a splits blitz which should educate our readers as to the many different usages of the newest toy. Home and road splits are probably the most commonly used and misused of all splits. Here I would like to show why just using career home/road splits to evaluate a batter isn’t a good idea.

A quick Google search of Matt Holliday + home/road splits brings back multiple results from this very chunk of the internet. For his career, Holliday has hit .351/.420/.632 at home and .284/.353/.455 on the road in 1,860 and 1,778 plate appearances apiece. It’s fair to say that he has performed better at home. Holliday has spent the majority of his career – read: every season but his last – playing home games inside of Coors Field.

That factoid helps explain some of the difference between his .442 home wOBA and .353 road wOBA, but not all. Far too often folks point out a player hitting worse on the road as an indictment on his talent, or as a doubt in his ability. The reality is that most players hit worse on the road. In 2009, the average major leaguer hit .267/.340/.430 at home and .258/.326/.406 on the road. The exact reasoning can be debated for eons; the point is the home field advantage does exist and Holliday was no exception to the rule:

Home: 8.9% BB, 17.1% SO, .281 ISO, .378 BABIP, 20.4% HR/FB, .442 wOBA
Away: 8.9% BB, 20.4% SO, .171 ISO, .329 BABIP, 12.3% HR/FB, .353 wOBA

Leading up to his trade, people referenced the career numbers – in part as an adjustment to the small sample sizes naturally associated by slicing and dicing an already small dataset. In theory, 1,000 plate appearances over five years is worth more than 600 over three, but when dealing with past data and attempting to find the true talent level of a player, we have to weigh the most recent data the heaviest, something lost in this method.

The other big issue was that people took the road numbers as gospel, applying no adjustments or considerations to the numbers and completely ignoring obvious factors. For instance, Holliday’s road numbers excluded Coors. Meanwhile every other National League hitter would have those numbers included in their road totals. That means Holliday’s road numbers were naturally deflated just based on the ballparks he batted in.

Just using his career numbers, nobody would’ve predicted that Holliday could succeed to the tune of a .412 home wOBA in another environment, or that he would post a .367 road wOBA. That’s not to say that either of those numbers are his true talent levels, either. It is to say that while understanding park factors and how particular parks can affect batters (and pitchers) is important, that simply looking at career home/road splits as the gospel is not the best way to evaluate whether a batter is a figment of the park’s construction or simply behaving like most major leaguers.

I would recommend simply allowing the built-in park adjustments within projection systems do the math for you while exercising common sense in extreme cases.

Print This Post

11 Responses to “The Home/Road Splits”

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

    Good post, RJ. One of my least favorite analysis of baseball is people who misuse splits. “This guy sucked on the road, therefore we shouldn’t want him on our team!” It’s lazy, and misguided.

    Vote -1 Vote +1

  2. Joe R says:

    I actually just wrote something similar to this about Adrian Beltre right here

    Safeco absolutely killed him.

    Vote -1 Vote +1

  3. Chris W says:

    Another fun split – last year Griffey was the opposite of Beltre. He hit like an all star at Safeco, and was absolutely putrid on the road. I know Safeco is somewhat favorable for lefties, but it doesn’t come close to accounting for that split. Smaller sample sizes do play a part, though.

    Vote -1 Vote +1

    • Joe R says:

      I’m just dwelling back on home/road splits for the league. A good 3.5% increase in AVG, 4.3% increase in OBP, and a 5.9% increase in slugging?

      So if I want to, I can argue Beltre goes .286/.340/.500? That would be beyond fantastic given his glove.

      Vote -1 Vote +1

      • Alex B. says:

        Joe R, don’t forget that he put up a 334/388/629 year in 2004 and was the MVP runner up in his last season before being subjected to death by Safeco.

        Vote -1 Vote +1

      • Joe R says:

        I think it’s safe to assume he’ll never do that again, though.

        Given that there’s a good chance he’ll gain a marginal win for Boston above average w/ nothing but his glove, being an average hitter would do the trick for me. An .840 OPS is like 15 runs above average with the bat, and Beltre would be a legit all star.

        And the beauty about Beltre is, his value is outside of walking, so I can actually argue his merit without WEEI/Sports Hub Mouthbreathers accusing me of Theothink.

        Vote -1 Vote +1

  4. Alex B. says:

    That would make a 286/340/500 season seem to be perfectly reasonable if not downright conservative.

    Vote -1 Vote +1

  5. While I agree that using home/road splits is not a perfect solution, I think absent more advanced levels of analysis (for example, adjusting each of the hitter’s performance in a park plus equalizing the weights), this is a perfectly legitimate way of pointing out the negatives of a hitter.

    Else, why use any data at all for comparisons, given your objection to the flaws of using the road split? Any offensive data you use today are flawed in some way, the same as you note for home/road splits. L/R frequency. Type of pitchers faced. Caliber of pitchers faced. Day/Night. Grass/Artificial. Even the type of road faced by a hitter (for example, Holliday is penalized by playing more games in LA and SD than most other players, admittedly, same for AZ and SF). If this is such a big problem with using road splits, then why not make the case that we should not use any statistical data at all? All are flawed to some extent. (Say, this would make a good study for those with massive databases to play with, hint hint, correlating future hitting on the road with current hitting on the road and with current hitting overall and seeing which correlates better).

    One relatively easy way to incorporate home stats along with road stats, per your objection, for a better comparative stat is to prorate his home stats to the average, say, PA he has had at all the other parks and then add to his road stats. Any on-line service can do this easily and include it with their statlines for all their splits. Then there goes the argument that his home stats are not included and thus inflated. I would do this myself except it takes a lot of work to do, just to stop this line of objection.

    And people think that this inflation for other hitters are significant, but take, for example, Chase Utley.

    For his career, road batting line is .285/.374/.495/.869
    For his career, Colorada batting line is ..333/.365/.667/1.031
    For his career, w/o Colorado his line is .283/..375/.486/.861
    (note, road equals all his bb-ref parks data except for Citizen and Veterans, which does not equal the road stats there; then I removed Coors for the second line. For some reason their road does not add up to my summation of parks other than home.).

    Not much of a difference to Utley’s line.

    As you can see, Colorado missing from the majority of players will not make that big a difference because the vast majority of them do not play much in Colorado, plus, more importantly, Holliday will not be playing that many games in Colorado while with the Cardinals. (I would have used Pujols for a better comparator, but he actually has hit worse in Coors than his career overall and career road numbers).

    I use this method to explain why the Giants (or even any other team) should not want to get Hank Blalock, and many used the argument that his stats in Arlington counts, so it should be included for comparative purposes, but nobody has had an answer to my most pertinent point: he won’t be playing much in Arlington if he’s with the Giants.(and they aren’t at all this year). And Holliday won’t be playing much in Coors, I don’t know the Card’s schedule, but Pujols in his 10 year career has amassed 6,082 PA and has had only 132 PA in Coors, or roughly 2% of his PA.

    Or perhaps FanGraphs can have, say, CHONE run his projection system on actual data, since you like that (I like it too), and provide those for comparison purposes as an option in your data tables.

    Vote -1 Vote +1

  6. for the vast canvass, but I’m rattling overfond the new Zune, and prospect this, as rise as the excellent reviews several another people person printed, gift better you decide if it’s the redress deciding for you.

    Vote -1 Vote +1

  7. If you are like me, you often find that you create duplicate files as you rename or … The following is an example of why you may require duplicate file software: …

    Vote -1 Vote +1

  8. jerseys says:

    Great goods from you, man. I have understand your stuff previous to and you’re just extremely wonderful. I really like what you’ve acquired here, really like what you’re stating and the way in which you say it. You make it enjoyable and you still take care of to keep it smart. I cant wait to read far more from you. This is actually a great web site.

    Vote -1 Vote +1