By In Stuff

A bit more math: Park Factors

Some good response on my bit trying to explain why Trout leads Cabrera in WAR (it’s 7.8 to 7.0 according to Baseball Reference this morning; 8.8 to 7.6 on Fangraphs) but I have to say that quite a lot of what I’ve seen shows a fundamental misunderstanding of Park Factors. I think the Twitter exchange I had with Brilliant Reader Joe more or less captures it:

BR Joe: Don’t you find it a bit odd that Comerica “became” a hitter’s park at the exact time that Miggy went from great to otherworldly.

Me: Fair question. Now ask it in reverse.

Fair to say BR Joe didn’t buy my counter-question. He wasn’t alone. I got lots of people wondering if the sole reason that Detroit is NOW rated as a hitter’s park by Park Factors is because Miguel Cabrera is so awesome. I also got lots of people saying that the dimensions of Detroit’s park have not changed so it cannot just suddenly have become a hitters park. I also got lots of people who say they KNOW Detroit is a pitcher’s park because, well, they’ve seen it.

Like I say, all of these seem to fundamentally miss how Park Factors are calculated.

I can’t explain all of it because, honestly, I don’t understand all of it. There are a lot of adjustments made to equalize the innings pitched and to adjust for various quirks (such as the fact that hitters do not face the pitchers on their own team). If you want to get into all that, you can work your way through the millions of calculations here … and have fun.

But the BASICS of Park Factors are the easiest thing imaginable.

All you do is this:

Step one: You take the average runs scored in a ballpark (both teams).

Step two: You take the average runs scored in that team’s road games (both teams).

Step three: You divide the first total by the second.

And that’s all. Park Factors. There is so much contentiousness about Comerica Park but it’s all simple math. This year, the Tigers have scored 355 runs at Comerica and allowed 275 runs. That’s a total of 630 runs in 67 games — 9.4 runs per game.

This year, the Tiger have scored 320 runs on the road and allowed 242. That’s a total of 562 runs in 66 games — 8.5 runs per game.

You take 9.4 runs, divide it by 8.5 runs — that’s 1.10 — or a Raw Park Factor of 110. That’s raw so you still have to make all those adjustments I mentioned above. But it’s a good starting point. At 110, you have a Park Factor that shows Detroit to be a fairly extreme hitters park.

So, you see how that works? Cabrera’s awesomeness has nothing to do with it — we’re comparing Miggy (at home) to Miggy (on the road). Changing dimensions have nothing to do with it. How tough the ballpark LOOKS to hit in has nothing to do with it. All these various little things people keep bringing up have nothing to do with it. You are simply comparing how many runs are scored in the ballpark against how many runs are scored in other ballparks. That’s all.

Now, it should be said that a single season is a very small sample size — and we’re not even through this season yet. That Raw Park Factor for Detroit this year is not reliable at all. You need more than five months of data to feel good about what you have. This is why Baseball Reference tends to go with multi-year sample sizes.

Let’s look at Comerica since the start of the decade. In 2010, Comerica Park was more or less neutral — a slight lean to pitchers.

Average runs per game in 2010:

Comerica Park: 9.13

Road: 9.30

Raw Park Factor: 98

But it shifted in 2011:

Comerica Park: 9.51

Road: 8.98

RPF: 106

Still a hitters park in 2012:

Comerica Park: 8.91

Road: 8.32

RPF: 107

And you know it’s a fairly dramatic hitters park this year.

Now, look at a place like Texas. Every year, it’s a great hitters park — but it’s been weird this year.

Average runs per game 2010:

Rangers Ballpark: 9.49

Road: 8.70

Raw Park Factor: 109

And then the crazy 2011 season:

Rangers Ballpark: 11.06

Road: 7.85

Raw Park Factor: 140!

A little bit more settled in 2012:

Rangers Ballpark: 10.13

Road: 8.56

Raw Park Factor: 118

So how do you explain more runs on the road this year?

Rangers Ballpark: 8.46

Road: 8,.59

Raw Park Factor: 99

I don’t know how you explain it. Could be a fluke. Could be a weather issue. Could be a shift in the wind. But the point is Park Factors don’t TRY to explain. They just count up the numbers.

Everybody knows San Diego is a great pitchers park. But what do the numbers say? This time I’ll use the overall numbers.

Since 2010, at PetCo Park, there have 2,268 runs scored. Both teams, remember.

Since 2010, away from Petco, there have been 2,660 runs scored in Padres games. That’s about 400 more — about 100 more runs per year on the road.

The Padres offense has scored more on the road. Since 2010, the Padres offense has scored about 13% more runs on the road. The Padres pitching staff has allowed about 16% more runs on the road. Every single year, the Padres offense has scored fewer runs at home than on the road. Every single year, the Padres pitching staff has allowed fewer runs at home than on the road.

THAT is a pitcher’s park.

Now, you should know that this is just a Park Factor for runs scored. Every year, Bill James in his Handbook does an involved Park Factors (he calls it Park Indices) where he calculates the best parks for home runs, for triples, for strikeouts, for all these other things. That’s very interesting. Kaufman Stadium in Kansas City, for instance, is usually the toughest home run park in the American League. But it is also a neutral to to slight hitters park overall because strikeouts tend to be low, walks tend to be up (both teams, so it’s not just Royals pitching), the outfield is so big that hits tend to drop. I think it’s a comfortable place to hit. The Park Factor there is 101.

Print Friendly, PDF & Email

124 Responses to A bit more math: Park Factors

  1. JMW says:

    If the dimensions aren’t changing, multi-year park factors seem to make a lot more sense as a component of WAR.

    • John says:

      Agreed. The dimensions aren’t changing, but the quality of the defenders does change (specifically, the home team, who account for half of the runs scored side of the equation).

    • Atom says:

      But, those defenders are playing at both home and the road. Why would that make a difference? This claim is still showing a fundamental misunderstanding of park factors. Which is frustrating because I had the exact same issue and can’t think of how to explain it to have it make sense!

      Think of it this way. You have two teams play each other 500 times. 250 times in park A, 250 times in park B. Overall, if those two teams combined scored 100 more runs in Park A…well, Park A is probably more favorable to hitting, right? The same guys are playing every day, so it doesn’t matter. Those offenses and defenses and pitchers are the same are both places, so a good offense wouldn’t make one park better.

      No, just expand that. Instead, we’re comparing the Tigers at home and on the road. Then, we’re comparing the teams that face the Tigers at home and the road. The Tigers are constant each way…same defense, same pitchers, same hitters. The opponents have the same team with the Tigers and their own ballparks. If more runs are consistently scored at the Tigers stadium than when both teams play away from Tigers stadium, a run has less value at Tigers stadium! Does this make sense?

      Side note: I agree with JMW about multi-year factors over single year, simply because a larger sample size can reduce anomalies.

    • But if park factors are calculated by measuring team’s performance at home and on the road, wouldn’t divisions that have bad defenders tend to be divisions with “hitter’s parks?” For example, because teams play more games against their own division, if a division was loaded with bad defenders, that division would be loaded with hitter’s parks. How does park factor account for this?

    • Isn’t the AL Central known for having some pretty terrible defenders? According to fangraphs, 4 of the 5 AL Central teams are in the bottom half of all team defenses by UZR. Might this account for the AL Central’s wealth of hitter’s parks?

    • This comment has been removed by the author.

    • Rob Smith says:

      What on earth does defense have to do with it? Presumably a bad defender plays bad defense at home and bad defense on the road. If their bad defense allows 10 extra runs at home, presumably he would also allow 10 extra runs on the road. That should cancel out. Unless… for some reason, the defender plays worse defense at home because the infield is chopped up. So then more runs would score at home and it would impact the park effect. So the presumption is that if he commits more errors at home then there is a reason (field, sun, weather, big dimensions to cover) that accounts for that & should be part of the park effect. Otherwise, over time, a player should play exactly the same defense home and away.

    • Yep, you’re right, Rob. I was thinking that Miggy might face more bad defenders because of his league, which would up his stats, but you’re right, it does divide him by himself so it doesn’t seem to be an issue. I guess my real issue is that there is no explanation for why Park Effects change, so to some extent it seems like it is a fluke. In that case isn’t it more like BABIP which often changes from year to year (obviously not with some speedy guys like Ichiro who have higher BABIPs)? Should we really take away from what Miggy’s done if we can’t explain WHY his park plays as a hitter’s park? I mean, it’s possible that next year it will be a pitcher’s park, just like it’s possible that Miggy’s BABIP next year will be .250. Thoughts?

    • invitro says:

      “Should we really take away from what Miggy’s done if we can’t explain WHY his park plays as a hitter’s park?”

      I don’t see why not. Surely trying to explain all the factors of park effects is a worthwhile thing to do. But just because we don’t know what all of them are doesn’t mean we should ignore the knowledge that we do have.

  2. Unknown says:

    The ask it in reverse answer was awesome….

  3. ericanadian says:

    Doesn’t this mean that the stadiums in which a team plays their away games have an impact on any given stadium’s park factor?

    How do you compare teams in different divisions effectively given that makeup of the away stadiums with which each team is compared to is completely different.

    For example, if Anaheim plays all their games in Texas, this would artificially turn Anaheim’s stadiums into a pitcher’s park. If they played all their games in Oakland, the reverse would happen.

  4. Tim Deveney says:

    I’ve always wondered about Kauffman Stadium. It feels like the only team that can’t hit home runs there is the Royals. I was looking at the Royals’ team home runs and they have almost the same home/road home run averages per game (.72/.69) in 2013. In 2012 the Royals pitching staff gave up eleven more home runs in away games than in home games (although the Royals hitters hit 7 more homers on the road that year). Of course it’s entirely possible that Luke Hochevar pitched more away games than home games last year.

  5. Eric says:

    I was looking back at the Barry Bonds era Giants and when he was hitting 60+ home runs the park factor was in the low 90’s. In the years after, when he was in the 40 home run range, it jumped up to right around neutral to slight hitters park. Park factors seem like the kind of thing that seem intuitive now without necessarily putting a number to it. For instance, we know if you take Cabrera for example and move him to San Diego he will probably see his home run totals decline. We can understand the effect weather/climate can have on hitting (i.e. Coors Field would give up more home runs than San Diego if the dimensions were exactly the same) and we understand that weather patterns can vary from year to year. It just FEELS wrong that a weather pattern could so strongly effect scoring. Especially when teams have such turnover from year to year, it’s easier to point to a player or players and say “that’s the reason scoring is up/down.” For example, from about ’05-’12 it looks like Philadelphia has slowly been moving towards 99-100 from a 104-105 range. It’s seems to make sense to point to the acquisitions of Halladay and Lee and the emergence of Hamels, as well as the gradual aging of Howard/Utley/Rollins and say that’s the reason for the change. A lot of people a lot smarter than me have come up with the calculations and I’m not intending to cast doubt on them, I just think those are some reasons park adjustments, at least as far as putting a number on it and including it in WAR, are hard for a lot of people to grasp, even those more sabermetrically inclined.

    • BobDD says:

      That cannot make a difference because the same players you note are also a part of the away stats for their team. The only way a player (pitcher or hitter) could affect a park factor is if he played only at home or away.

    • Which Hunt says:

      Or was injured for a major road trip/home stand…
      One year samples could get slippery.

  6. The assumption that all the same guys are playing on the road and home is really goofy. Park factors are interesting in and of themselves, they shouldn’t be in the mix for something like WAR.

    Let’s say Justin Smoak twists his ankle right before a long home stand. He’s out for four or five games. That affects Kyle Seager’s WAR because now an AAA guy is taking Smoak’s plate appearances at SafeCo. The same guys are NOT playing on the road and home. That a player’s WAR can be affected in any way, shape or form by such a scenario is just goofy.

    Fangraphs and BR also need to get their $%*@ together if they are both going to stubbornly use the same acronym.

    • Really goofy?
      I can’t think of any players that play only on the road or only at home.
      You’re talking about a completely random occurrence. There are many random occurrences beyond that. That’s why they use such a large sample size, to balance out randomness and reduce the margin of error.

      I doubt you know much about statistics, but margin of error can be calculated. I bet there’s a reason 5 years was chosen as the sample.
      You should check it out, rather than blast rigorous statistical calculations based on an anecdote.

    • WAR is not *fact*. It is an *estimate* of value. The number of RBIs, hits, home runs, etc., those are facts. But to determine a player’s total value, given all of the variables involved, the best we can do is an estimate.

      So, yes, some players’ estimates will be slightly off because they played considerably more games at home than they did away, because they faced harder pitching on average, because … well, because any number of factors.

      But all of these variables can be accounted for to some degree, and then the value estimate improves.

      That a player’s WAR can be affected in any way, shape or form by such a scenario is just goofy.

      It’s not goofy; it’s the nature of statistics, which still work well.

    • Dinky says:

      Ice Cream, I have to strongly disagree with you. A great hitter in Coors pre-humidor would hit even better. A lousy hitter in Coors pre-humidor would also hit better. You cannot adjust for games missed; that’s just silly. It also assumes you know what Smoak would have done in the games he missed. In the long run, except for a few rare circumstances with very old players, players will play about the same number of games at home as on the road. In my 50 years of following baseball, I can only think of a couple of exceptions. Really popular starting pitchers (Koufax, Fernando, etc.) might get held back a day to fill up the seats at home instead of on the road. And Barry Bonds, his last couple of seasons, would tend to sit out on the road so home fans would get to see him more often. More Barry at AT&T never seemed to make it a hitter’s park, either. But he still got lots of games on the road and at home. Barry’s career splits were 1576 games at home, 1410 away, and he was the only batter I can recall who played that long or sat that often. And now that I’ve looked them up, my memory was wrong; Fernando had one more game away than in Dodger Stadium, and Koufax pitched in many more games on the road than at home. It just doesn’t happen.

    • “You cannot adjust for games missed; that’s just silly. It also assumes you know what Smoak would have done in the games he missed.”

      I… what? I am arguing the exact opposite.

  7. Haha, this is great. You guys completely take Sabremetrics on faith, deigning to condescend to anyone who can poke at it, and then something like Brett Lawrie shifting happens and it completely breaks WAR. The same system you can’t imagine having any holes had Lawrie as the AL MVP.

    And yes, there are managers who are loathe to play their closers at home unless they have a lead. You’ve got a factor involving people other than the individual player and who he faces affecting an individual player stat, which is what WAR is.

    By the way, how many years of statistics knowledge is going to be required for Baseball Reference’s WAR and Fangraphs’ to agree? Keep in mind the universe is only 14.5 billion years old. Until that’s resolved, you people look like clowns. I’m laughing at the superior intellect.

    • Rob Smith says:

      Everyone is pixilated. Except you, of course. (Paraphrasing Jane and Amy Faulkner)

    • Haha, this is great. You guys completely take Sabremetrics on faith

      I’m not certain what “on faith” means. And I, for one, do not completely believe in any stat. Yes, there is likely a better way to evaluate players than WAR does, but at the moment, it seems to be the best estimate that is publicly available.

      But Sabermetrics, like other fields of research, relies little on faith and “completely” and instead focuses on always asking questions, always re-evaluating, always improving, always testing, etc.

      Screw faith.

    • Dinky says:

      I think you have a different definition of the word faith than I do. Religion is faith; belief in something that cannot be measured. Sabermetrics is statistics applied to baseball for predictive, evaluative, and entertainment. For example, the shift is a GREAT example of sabermetrics. Look at the teams that put on a shift against a certain hitter, log the results, and compare those results to the teams that don’t put on a shift. Latest studies I’ve seen indicate that the teams that shift the most get better results against the shifted hitters than the teams that don’t shift at all.

      Just because B/R and Fangraphs disagree one some of their proprietary calculations does not invalidate the notion of WAR. We’ve had lots of intellectual struggles, such as Blu-Ray versus HD recently. Eventually one will be chosen as superior and the other will adapt. The decision for both web sites to call their statistic WAR is marketing, not sabermetrics.

    • Vidor says:

      Brett Lawrie? Really? God.

    • BobDD says:

      Don’t conflate WAR and park effects; WAR is indeed an “estimate”, but park factors are just math. Over a short period of time randomness can “break” even park factors, but the longer the time, the more exact the numbers become.

    • Yes, Vidor. I assume you are unaware, because I can’t think of any other reason why you would make a sniveling comment, but when Brett Lawrie was shifted last year, he got credit for making plays out of his zone. This completely broke WAR. Please do some reading before posting again.

      Although I am really, really hoping that you know about that and are agitated that someone had the GALL to bring a failure of WAR. Because that would be absolutely perfect.

    • Vidor says:

      You might consider not being so snotty in your comments, Ice Cream Jonsey, because I was agreeing with you.

    • I genuinely apologize, Vidor. I’m sorry.

  8. Brent says:

    FYI, The Ballpark in Arlington had construction behind home plate before this season to try and cut down on the “jet stream” and shift the park away from being such a hitter’s park and to help push the advantage of their strong young pitching.

  9. “we’re comparing Miggy (at home) to Miggy (on the road)”

    That’s where park factor loses me. From what I see, Cabrera is:

    -.365/.454/.643, with 1.097 OPS and 148 total bases at home

    -.349/.443/.719, with 1.162 OPS and 179 total bases on the road

    So it seems like Cabrera is always on a lose-lose situation.

    “Miggy has great numbers!” “Well, that’s because he plays on a hitters park” “But his road numbers are also great!” “Well, that’s because his division is full of hitter parks”

    • J Hench says:

      Miggy’s great numbers are not a result of Comerica Park being a great hitter’s park. Miggy’s great numbers are not worth as many wins as Trout’s slightly less great numbers because many more runs are scored in the parks that Miggy plays most of his games in than in the parks that Trout plays most of his games in.

    • I see the point, but I don’t see why this goes against Cabrera.

      Most of the advanced stats are meant to be based upon only what the player can control (i.e. he can’t control his teammates, so RBIs are flawed), but park factors are not controlled b y the player. If anything, Cabrera rises Comerica Park’s effect, and that goes against him? It’s doesn’t make sense.

    • Dinky says:

      Miggy has great numbers because he’s a great hitter. But the Tigers as a whole have an .817 OPS at home, and a .761 OPS on the road. That’s an ENORMOUS difference for the team. They’ve scored 355 runs at home, and 320 on the road (in one fewer game right now). Take away today’s game (where they scored 7 runs, above average at home) and it’s 348 to 320 in the same number of games, a difference of about 0.4 runs a game. For whatever reason, the Tigers hit better at home. Take away today’s game (to make home and away even; it was 7-6 so a bad day for the pitchers, thus a game AGAINST my thesis) and they have allowed 27 more runs in Detroit than on the road, almost exactly the same average.

      Now you could argue that hitters do better at home (because league wide, they do). But so do pitchers. Yet the Tigers have 1 run difference between their home and road splits, with 28 more runs scored at home, 27 more runs allowed at home. And that’s with the Tigers hitting more homers on the road. So Detroit, for an assortment of reasons, is a good hitting park this year. Maybe the backdrop makes it easier to see pitches, maybe the wind has been blowing out more (curse you, climate change!), maybe Detroit’s mound is lower than average. More likely, it’s a bit of several factors. But the same fielders are playing in Detroit as on the road behind the Tigers, and the same hitters are hitting. Thus, park adjustments are necessary.

      Remember, and I’m a huge fan, but there’s a big reason why the last guy to bat .400 was playing half his games in Fenway. There’s also a big reason why the Yankees had so many left handed hitting stars; Yankee Stadium favored left handed hitters, so it made sense to get them.

    • (I’m not trying to troll, just understand, sorry if I’m so insistent)

      The point is that players hit better at home, but Cabrera is actually hitting better (per OPS, at least; with one more run scored on away games) away from home. I don’t get tthis at all. Shouldn’t he be credited for it?

      Players can’t control where they play, or the wind, or nearby buildings. Just like they can’t control how many runners are on base. Why is only one of these discarded?

    • Robert says:

      His hitting and run generation are not discarded. But if you want to translate that run generation to wins (to calculate an estimate like WAR), you have to know how much those runs are worth. They are not worth the same for a San Diego Padre as they are for a Colorado Rockie, which most people intuitively get. So, without even quibbling over the specifics of Park Factor, it’s pretty obvious that excluding them and pretending San Diego and Colorado runs are worth the same is even more incorrect.

    • Jeremy T says:

      The idea that we only want to measure what he can control actually supports using park factors. It’s not so much that Cabrera is being penalized as it is Cabrera doesn’t have complete control over his own OPS, Batting Average, etc. Park factors are merely an attempt to account for one more variable and put the players on a level playing field.

    • BobDD says:

      GC: Exaggerate park dimensions in a theoretical case.

      Home Park is 400 ft down both foul lines and 600 to center.

      Away Parks average 330 ft down foul lines and 410 to center.

      Or exaggerate even more – until everyone has to admit that in this home park example it is harder to hit HRs. It would definitely be a disadvantage to a HR hitter for the home team in his personal stats. Obviously. And a different point is that there would be a park factor that gave more “credit” to HRs at home based simply on the scarcity principle. So if a player from that home park had the same number of HRs as a player that played in a neutral park, everyone would know that this particular player from the home team is a better HR hitter than the guy he was tied with in raw numbers.

      Leaving behind my good or bad/whatever example and use real world park factors, the differences are just smaller and in some cases do not even show well until multiple years. But it is still the same mathematical principles involved.

  10. Lol at the people suggesting park factors shouldn’t be included in WAR.
    Example: let’s pretend San Diego is in the AL Central and Miggy plays for them instead of Detroit. As long as you’re willing to admit that San Diego is a pitchers park (if not, I’d love to hear your reasoning), isn’t it safe to assume that Miggy would have MUCH worse numbers than he does now, even as the exact same player? Without park factors, he’d have a lower WAR, simply because of where he plays. How does that make sense?

    Seems like a pretty simple concept, and there are enough adjustments to tease out the biases. These manufactured scenarios of why park factors are wrong are just ludicrous.

    There’s substantial randomness in several WAR components. It’s not supposed to be a 100% precise number. The difference right now between Trout and Cabrera is right on the borderline, IMO, of definitely saying who has been more valuable.

    • And who’s to say he wouldn’t have more non-HR hits in San Diego? Or change his approach so he hits longer balls (I don’t know, maybe he’s holding back since Comerica already favors him)? Or if his new approach cause an injury and he never plays another game?

      Too much “maybes” screw what’s actually happening. And real-life Cabrera is the best hitter in the AL, which such a great margin that he’s probably the best player in the AL.

    • Too much “maybes” screw what’s actually happening. And real-life Cabrera is the best hitter in the AL, which such a great margin that he’s probably the best player in the AL.

      But your “probably” has a lower confidence level than WAR’s confidence level.

      And as I’ve stated previously either in this thread or the other related thread, WAR, as an estimate, suffers from not having error bars/values associated with it. A player’s WAR (which is an *estimate*) is not exactly the number being displayed. It’s that number +/- another number. Perhaps 5.0 +/- 0.5.

      Given the error/variance/etc. involved in calculating WAR, it’s more reasonable to maintain that Trout and Cabrera are the best players in the AL, and it is too close to declare one *the* best for this season, according to WAR (or any other estimate). If I had an MVP vote, and I could split it in half, I would.

    • Dinky says:

      I think Trout is the best player in the AL, and I’m willing to extend it to the best player in baseball. But I’d still vote for Cabrera for MVP this season for several reasons, not least of which is that he’s in a pennant race and Trout is waiting for Pujols to get healthy, Hamilton to find his swing, and the Angels to find some arms.

    • Wilbur says:

      If the post-season award was titled “Player of the Year” or “Best Player of 2013”, would you change your vote to Trout?

      Which leads to the question: Does anyone think when the award was established in (I believe) 1931 that the BBWA intended it to be anything but the award for the best player of the year? And should it be construed today as anything but the award for the best player of the year? If not, then shouldn’t there be an award for the best player of the year?

    • invitro says:

      What did/does the BBWAA say the MVP is intended for? Do they give any instructions to the voters? Is this information easily available online? 🙂

    • “Example: let’s pretend San Diego is in the AL Central and Miggy plays for them instead of Detroit. As long as you’re willing to admit that San Diego is a pitchers park (if not, I’d love to hear your reasoning), isn’t it safe to assume that Miggy would have MUCH worse numbers than he does now, even as the exact same player?”

      Good point! This is why Adrian Gonzalez set personal highs in all offensive categories when he went from San Diego directly to Fenway.

  11. Mark Daniel says:

    It’s been shown that home teams win about 54% of the time, and that home hitters and pitchers perform better than road hitters and pitchers in the same park. It seems this basic fact should be taken into consideration. Maybe it is, but I don’t think so.

    • Dinky says:

      This year’s Tigers are one counterexample. The hitters hit better at home. The pitchers pitch worse at home. There have been many others. Teams win more games at home for lots of reasons, not least of which is that they always know how many runs they need to score in the bottom of the 9th (or extra innings).

    • Mark Daniel says:

      This comment has been removed by the author.

    • invitro says:

      You do understand that the home team and the visiting team are counted equally when totalling up runs, right? I thought Joe explained this. If players hit 5% better at home, they are counterbalanced by the visitors that hit 5% worse on the road.

      A recent SI article claimed that home advantage is entirely or almost entirely due to umpire/referee bias in all sports.

    • Mark Daniel says:

      Dinky, I mean that the Tigers pitch better at Comerica than Tigers opponents’ pitch at Comerica.
      Since home teams win 54% of the time on average, that means home teams perform better (pitching, hitting, other things) at home than their opponents do in that park. I don’t know why. Comfortable surroundings? Close to family? Short drive to work? Whatever it is, it exists. This year, in the majors, according to B-R, the average home record is 35-30 and the average road record is 30-35. This is independent of park factor.

      Looking back at the AL, we see the league average OPS is as follows for the last 5 seasons:
      Year: Home OPS/Away OPS (home advantage)
      2013: .737/.718 (+2.6%)
      2012: .750/.712 (+5.3%)
      2011: .745/.716 (+4.1%)
      2010: .752/.717 (+4.9%)
      2009: .784/.744 (+5.4%)

      Average difference +4.5%

      Note that this advantage is sometimes difficult to find. For example, Mike Trout plays in a pitcher’s park, so any advantage he gains at Anaheim is masked. But it seems that he does gain a 4.5% advantage over his opponents when they playing in Anaheim. I just wonder whether this difference is accounted for. Maybe it is in some indirect way.

    • Mark Daniel says:

      invitro, I’m not talking about the definition of park factor. And I don’t think it matters how home players gain an advantage. Mike Trout, even in Anaheim, has an advantage over his opponents’ playing in Anaheim. When in a hitter’s park, a hitter’s stats are adjusted downward. When in a pitcher’s park, as far as I can tell, a hitter’s stats are adjusted upward. But there is still a home field advantage there, at least relative to opponents playing on that field.

    • invitro says:

      What are you talking about then?

      You say “It seems this basic fact should be taken into consideration.” Given that post’s context, and the title of Joe’s post, and what Joe’s post is about, and other comments, I think it’s reasonable to conclude that you mean “into consideration in the definition of park factor.”

      You say “I just wonder whether this difference is accounted for.” Given that post’s context, and the title of Joe’s post, and what Joe’s post is about, and other comments, I think it’s reasonable to conclude that you mean “accounted for in the definition of park factor.”

      Please show me where I am wrong… into consideration when calculating what, if not park factor? Accounted for where, if not park factor?

    • FranT says:

      But any advantage that allows the home team to score more runs is canceled out by the disadvantage that causes the road team to score fewer runs.

    • invitro says:

      Also, you say “And I don’t think it matters how home players gain an advantage.”

      Well, you said

      “that means home teams perform better (pitching, hitting, other things) at home than their opponents do in that park. I don’t know why. Comfortable surroundings? Close to family? Short drive to work?”

      so I just thought you might be interested in the research done to try to answer the question.

    • Mark Daniel says:

      invitro, thanks for your comments. Sorry for being unclear. I must be confused.
      I suppose the take home message is that players may gain a home field advantage (regardless of park) compared to road players at that park, but they have a road disadvantage in the other 81 games that is similar, so it all comes out in the wash. Thanks to you and FranT for clarifying.

    • invitro says:

      I apologize for being obnoxious.

      I suppose it is possible that a team has a home-field advantage that would exist even in a neutral park and that a correction is needed for that… but you would need to demonstrate such a thing exists, and we’re getting into possibilities and vagaries that I’d rather not comment on. I believe I completely understand the first, raw estimate of Park Factor, but not the various corrections. I trust the result, though.

  12. wee 162 says:

    First, I’m glad Joe done this post. It’s a highly interesting thing to me as someone who’s slightly sceptical about park factors. However, something which I do think is that parks are not a vacuum, and teams will build their teams to suit their own park first. Bill Mueller signed for the Red Sox in 2003, and his OPS splits in the 3 years he was with them were consistently higher at Fenway than away from it. He was signed for that very reason. That is going to come into park factors. Texas & Colorado are always going to be looking for players who hit the ball in the air so as to take advantage of their ballparks, Seattle & San Diego not so much. Also, this is just anecdotal, but I’ve always thought that sides who are out of it in the last month of the season are more likely to play those brought up after 1st September away from home than at home (to avoid annoying their fanbase).

    I definitely think there’s something there, I’m just not convinced that Park Factors are as definitively correct as it seems to be assumed to be. For example I’ve just had a quick look online for what BABIP for stadiums are and came up with this article on Fangraphs.
    Now that to me looks like Comerica shouldn’t be getting considered as that much of a hitters park as it has fractionally below average BABIP, and lower HR per flyball. It is however marginally higher than the median BABIP (18th of 29 stadia).

    Anyway, interesting discussion.

  13. wee 162 says:

    This comment has been removed by the author.

  14. wee 162 says:

    Okay, I’ve just read The Hardball Times article linked to in the Fangraphs article. I think that’s got a lot of valid criticisms of Park Factors as they’re currently constructed. Most pertinent points which hadn’t occurred to me were;
    “There are two other factors that are controllable (by the league), but are currently configured to introduce significant differences between home and away home run totals:
    Interleague play: because the designated hitter is allowed only in AL parks, AL teams play with a weaker lineup in about 1/8 of their away games than they do at home, while NL teams play with a stronger lineup in one-eighth of their away games than they use at home.
    The unbalanced schedule: because teams play more games against their division rivals than against other clubs in their league, their home/away home run numbers may be distorted. This effect is most pronounced when all of a team’s division rivals play in parks that are on the opposite side of 100 on the HRPF spectrum (as was the case in 2006 for Boston, Chicago (AL), Texas and Philadelphia). That team’s HRPF trend (calculated the traditional way) will be magnified, e.g. Boston’s low HRPF is made even lower because their AL East rivals all play in high HRPF parks.”
    Article is here

    There’s lots of other things in there about potential problems, and frankly I like things that recognise issues and try to take a different approach.

  15. I was at Fenway on Tuesday and was sitting in the bleachers with my girlfriend. We’ve been to many games together over the years and I try to teach her new things about the game each time. This was her first game at Fenway, so I was telling her all about the history of the park and talked all about the weird dimensions, green monster, Pesky pole, Williamsburg, etc and how Fenway had much shorter fences in some places and much longer fences in others. After that all sunk in, a couple minutes later, completely unprompted she asked:

    “So do they take the different dimensions into account in the players’ statistics?”

    Like I said, this wasn’t her first baseball game, but I guarantee you this is the first time she’d ever thought about this. And lo and behold, her first question when considering differences in ballparks is basically asking if there are park effects!

    She completely validated my opinion that the idea of park effects should be natural, intuitive and obvious to anyone who actually thinks about it and doesn’t place inflated value on their pre-existing notions of how baseball works. The folks who denigrate efforts to place baseball stats into context either don’t enjoy challenging their own opinions and learning, or they just like being contrary.

    The fact is, every single baseball front office uses park-adjusted numbers when evaluating players. To think park effects and linear weighted stats are the realm of “bloggers” fundamentally misunderstand how teams evaluate players today. It’s hard to blame them given the level of discourse on MLB Network and ESPN, but the revolution is already over.

    • Phil says:

      You’ve got a keeper there, Tyler! 🙂

    • “She completely validated my opinion that the idea of park effects should be natural, intuitive and obvious to anyone who actually thinks about it and doesn’t place inflated value on their pre-existing notions of how baseball works.”

      There’s two things going on here. The first is what you bring up – how do we think a guy will perform if we sign him as a free agent or trade for him. Park Factors absolutely come into play. You’d have to be willfully obtuse to not consider this. I don’t think anyone is arguing against that.

      All I am saying is that for an end-of-year stat that tries to boil down a player’s performance into one number, it is silly to credit or detract based on what MIGHT have happened. Or what DIDN’T happen. Lowering a guy’s WAR because he hits in a pitcher’s park is as dumb as lowering a guy’s WAR because he’s in a contract year.

    • invitro says:

      “Lowering a guy’s WAR because he hits in a pitcher’s park is as dumb as lowering a guy’s WAR because he’s in a contract year.”


  16. Joe says:

    This was an interesting post that made me question the logic behind WAR even more. Essentially what it says is that a team scores more runs at home than on the road so the park becomes a better hitting park. Couldn’t you also make the case that it’s a veteran team and veteran teams do better in the comforts of home, where as San Diego is and has been a fairly young team over the course of this time and may not be as affected by external factors? This stat also appears to make team changes coincidencidental. In or around 2011 the Tigers added Prince Fielder and this year the Rangers lost Josh Hamilton – it could be a real factor that both parks changed as the team changed. I will admit that the park could be why Pujols and Hamilton have struggled since joining the Angels – but then why didn’t Torri Hunter when he left the Twins (and the Homerdome) and joined the Angels? with all stats and large data samples, it could be something, or it could be nothing – one things for sure, it’s a fun debate.

  17. This comment has been removed by the author.

  18. Louis Poulas says:

    WAR is an estimate, that’s it. It’s a darn good estimate, but just as with any other stat, it’s not perfect for every player and every situation. No one has ever said it was.

    Every great stat, like every great cut of meat, still needs seasoning.

  19. Frank says:

    I wonder what Park Factor adjustments were made to the old Metrodome when they did / did not have the fan on.

  20. The biggest problem with WAR and especially the Park Factors and defensive metrics is that they require faith to believe in them, and then the people who believe in them often go off about how sophisticated their new numbers are. Its a very smug attitude that I can’t understand and don’t appreciate.

    Joe doesn’t generally have this attitude, though occasionally it comes out. His readers (or rather, many of is commenters) on the other hand… Well let me just end that sentence right there.

    For example, Joe starts off immediately in his blog with this statement: “I can’t explain all of it because, honestly, I don’t understand all of it. There are a lot of adjustments made to equalize the innings pitched and to adjust for various quirks (such as the fact that hitters do not face the pitchers on their own team). If you want to get into all that, you can work your way through the millions of calculations here … and have fun.”

    From that statement I gather that Joe BELIEVES in the way park factors are calculated, but really has no idea how its done, and doesn’t even know how to check the work that others do. So far none of the commenters on this board have explained it either, but they are adamantly certain that their numbers are better than anyone else’s, (even though the two popular WAR calculators disagree with each other, often significantly)

    Can we all agree that park factors are a real thing at least in some parks, but that the effects of park factors are not fully quantifiable? I think that is blindingly obvious, just like we can agree that steroids are a real thing, but their effects on homerun rates are not fully quantifiable either? The math is probably easier in some extreme parks like Colorado, but even there it isn’t fully quantifiable, just more obvious. As far as the vast majority of other parks, where the park factors change annually, and stay close to neutral, you just look like fools trying to say that you have it down to an exact number and an exact formula. (even though the numbers are differeent depending on who you ask and nobody knows the formula and nobody can calculate it without relying on TONS of crazy things that they can’t explain, but believe in anyway) Not a chance.

    You will never convince me and a great majority of others until you can actually explain your formula, and your explanation better be complete, and better make total sense and resolve every variable, and for that matter, you better be able to think up every variable, then resolve it, then explain how you resolved it, then explain why your resolution to the variable is the correct one. Right now there is nobody out there doing this adequately, not even Joe Posnanski and Tom Tango working together. My guess is it can’t be done just like the effects of steroids can’t be quantified either.

    Can we just say that park factors are real and obvious, but also impossible to quantify? I would be happy with that and I think it is obvious to most people.

    • Well Stephanie I think you make some valid points but I take issue with the steroids analogy because it hasn’t actually been proven that steroids cause homeruns. There are lots of other considerations to take into account too. It seems pretty clear that the baseballs were bound tighter, scouting and video analysis probably payed a role as well. Basically a lot of smart people (including Joe, but that article seems to be lost) out there don’t really think steroids had a major effect, at least not as big as Rick Reilly seems to think.

      Read these articles and judge for yourself

    • invitro says:

      “Can we just say that park factors are real and obvious, but also impossible to quantify?”

      If by “impossible to quantify” you mean “impossible to quantify with a zero margin of error”, then I agree.

      If you mean “impossible to estimate”, well, that’s just plain wrong.

      So which is it?

    • This comment has been removed by the author.

    • Well, first, Invitro, I apologize because I didn’t realize you were an arbiter of truth on this blog with the right to say that someone’s opinion is just plain wrong. I should have simply asked you and then kept quiet I suppose. After all, whether my opinions are something you agree with or not is of utmost importance to me.

      But since you’ve deigned to respond to me, how about this for an answer. By “Impossible to quantify” I mean “Impossible to attach a number to that a majority of paying baseball fans will agree with.

      For that matter, try to get an estimated number that you can get a majority of MVP voters to agree with. These are the people who talk about baseball for a living and have a vested interest in the game. When a majority of those guys agree with the exact estimates of park factors that change every year, then you can start in on the majority of baseball fans.

      Tyler Mccormick – You are trolling. But even so, you admit that steroids cannot be accurately quantified in a way that thinking people can agree to, so your point reinforces mine.

    • invitro says:

      “that a majority of paying baseball fans will agree with.”

      So you think WAR is wrong because… paying baseball fans don’t agree with it. That’s your reason. OK.

    • Nope, I ASSUME it is wrong. Its an estimate after all. Estimates aren’t expected to be right, just estimates. So I assume it is wrong. I thought we agreed on that?

      I also think it is impossible to even estimate these numbers in a way that the majority of paying fans will agree with. I think park effects are intuitively obvious, but I also think that calculating park effects is obviously impossible to do in a way that satisfies a majority.

    • Phil says:

      Wow: just wow. Nothing like some indignant self-righeousness to rub everyone the wrong way:

      “You will never convince me and a great majority of others until you can actually explain your formula, and your explanation better be complete, and better make total sense and resolve every variable, and for that matter, you better be able to think up every variable, then resolve it, then explain how you resolved it, then explain why your resolution to the variable is the correct one.”

      I’d love to wait for the unicorns and rainbows, too, but sometimes we just have to do the best with what we have — once you get that perfect knowledge, please do pass it along to the rest of us. I err on the side that more knowledge is better than less, even if it’s not quite perfect yet. But that’s just me.

    • invitro says:

      “Estimates aren’t expected to be right, just estimates. So I assume it is wrong.”

      Estimates are not right or wrong. They have an accuracy. One estimate can be more accurate than another. In that case, you don’t say the first is right and the second is wrong.

      “I also think it is impossible to even estimate these numbers in a way that the majority of paying fans will agree with.”

      I’m not sure that I am concerned with whether the majority paying fans agree with WAR or Park Factor. I am concerned with how accurate they are toward what they are supposed to measure. I suppose I am also concerned with what sabermetricians think about them.

    • The problem is that the way WAR is formulated it is promising unicorns and rainbows where they don’t actually exist. A good estimate of park effects is just like a unicorn. It would be wonderful and great fun if it was possible, but its not. Claiming to have found it, when you haven’t, is not the same as finding it.

      When WAR claims (Or people like the commenters on this blog claim) to have a quantifiable measurement of everything anybody needs to know in one convenient stat, they are really promising a unicorn. Something wonderful that isn’t real. In reality, WAR is anything but convenient and its claims are impossible to prove, like if someone found a raging bronco and said “Here I found your unicorn. You must believe in it.”

    • Phil says:

      I would say that’s a complete misreading of the body of WAR work, but that should be painfully obvious at this point. To each his/her own, which is what (should) make discussions and arguments fun!

    • Robert says:

      “Well, first, Invitro, I apologize because I didn’t realize you were an arbiter of truth on this blog with the right to say that someone’s opinion is just plain wrong. I should have simply asked you and then kept quiet I suppose. After all, whether my opinions are something you agree with or not is of utmost importance to me.

      But since you’ve deigned to respond to me, how about this for an answer. By “Impossible to quantify” I mean “Impossible to attach a number to that a majority of paying baseball fans will agree with.”

      I don’t mean to offend here, but some opinions are actually incorrect. If someone told you they were sure that the sky was purple, and that you just didn’t understand the assortment of shades of purple…you’d laugh off that opinion as silly.

      This leads us to the biggest disconnect (and feeling of condescension) of the anti-sabermetric people. Everything, literally everything, is quantifiable within some margin of error. That’s the definition of statistics within the discipline of mathematics. So, we can generally separate the world into two kinds of people, those that have some understanding of statistical principles and those that don’t. If someone doesn’t at least understand Gaussian Distribution and Standard Deviation, they’re not even equipped to have the debate. That knowledge is an essential tool for engagement.

      As it applies to baseball, this creates a huge chasm, because most people that understand statistics have a hard time rejecting sabermetrics entirely, because at it’s base it’s just a logical application of statistics to numbers that have happened in the past. Within the statistical community there is plenty of argument about the best way to approach it for predictive value, but there is no debate that important things (like park factor) should be excluded entirely b/c they can’t be calculated to zero error.

      Which leads to the biggest point of condescension for anti-sabermetric/non-statistics people. Without an understanding of statistics, it’s impossible to make good arguments against advanced calculation baseball stats, because that’s where all the room for criticism lies. And without SOME intellectual ability to understand the stats at a deep enough level to criticize them, many end up calling the sky purple.

    • BobDD says:

      I think it’s OK to rely on things that work that I do not personally understand. I am not a mechanic but I still drive a car.

    • jkak says:

      Robert – Not everything is quantifiable. Who is the better artist: Rembrandt or de Kooning – please quantify. Value judgments are not quantifiable.

      Let’s go back to Joe’s statement of the reason he has written so much here about Trout and Cabrera and things related: “I’m fascinated by the question of which one is better.” That’s a value judgment and is not objectively quantifiable. The WAR calculation does exactly what it is programmed to do, i.e., it generates the WAR number it is programmed to generate.

      If you believe WAR (whichever version you prefer) with the assumptions and biases built into it (the factors to which it attaches value and the way it weights those factors) is a formula that determines whether player A is “better” than player B, then you can believe yourself justified in saying that the player with highest WAR is the best. But many of us, even those of us who do know statistics and mathematics, believe it is foolish to proclaim that any statistic or calculation can objectively and conclusively determine who is “better”.

    • Robert says:

      Sure it is. I could quantify the better artist by figuring out who was more prolific (counting their works). I could endeavor to figure out which works sell for more money. I could sample people in an attempt to figure out who is more famous. I could create a formula that weighed all 3. And, since in none of these cases would I be sure I gathered every data point, I could apply statistics to determine my error and confidence level. It’s a bit of a silly example, because I don’t really have control data. My error would be very high, whereas in baseball we have literally thousands upon thousands of games for control driving our error down greatly.

      Now, you could criticize my criteria. You could say, “those have nothing to do with who was a better artist” and maybe you’d be right. But what you COULDN’T say is that I didn’t quantify the question with objective data. And it goes with WAR. When you say WAR isn’t objective, it’s hard to take that criticism seriously. It is NOTHING if not objective, it is but a consistent, player agnostic calculation of numbers. Park Factors, the point of this posting, is just an objective way to measure how much runs are worth in Colorado and Oakland and Detroit and Anaheim. Now, you are correct in that it’s not conclusive, but I know very few sabermetric folks that make that claim…particularly for small differences in WAR.

      And this is another example of the chasm I spoke about before. You should be criticizing the criteria in WAR that you think are poorly calculated (my selling price does not indicate artist quality example) not saying WAR is subjective when it obviously isn’t. For instance, I would have voted for Trout for MVP last year, because I think he was better. The offensive numbers were close enough, Trout plays in an intuitively harder place to hit, and the defensive differences made up the rest. This year, I’d vote Cabrera, because Trout’s WAR advantage lies within a part of the formula (defense) that I personally criticize due to sample size fluctuation.

    • Adam says:

      “It is NOTHING if not objective, it is but a consistent, player agnostic calculation of numbers.”

      You’re confusing two completely different concepts. “Objectivity” means that two different observers can reliably come to the same conclusion given the same evidence. In your art example, a count of the number of paintings by an individual artist is objective.

      That is quite different from “consistency,” which is a property of an estimator guaranteeing that the estimates it generates will converge to the true value of the variable being estimated as the sample size approaches infinity.

      Several commenters in this thread have pointed out plausible sources of inconsistency in standard measures of park effects. That doesn’t mean that PE isn’t a “natural” thing to try to correct for, but it does mean that it may not being properly corrected for with current techniques. And “properly” doesn’t mean “without error”, but “with bias.”

      It seems obvious that park effects ultimately operate at the level of the individual player (as in the Bill Mueller example above). Any calculation of an aggregate park effect across all players will be subject to change as the composition of the team changes, even though the effect on each individual player (or player type) remains the same. Attributing such composition effects to changes in the performance of specific players is incorrect.

    • jkak says:

      The number of home runs a player hits is objective. Trying to factor that number into an evaluation of that player’s value compared to another player, unless number of home runs hit is the sole criterion on which value is based, is subjective, even if you use the same objective formula to evaluate both players. WAR measures WAR. It is objective to that extent. However, to argue that higher WAR means “better” player obviously is not objective, and at root is not much different from saying Pujols is better than Trout because he is paid more. After all, using salary to determine who is the better player “is NOTHING if not objective” and in fact “is but a consistent, player agnostic calculation of numbers.” Objective, but essentially meaningless.

      As far as park effects, I don’t think anyone would argue that some parks are more favorable to run scoring than others. And of course the number of runs scored in all parks is easily quantifiable. The subjective leap is when you say PE “is just a objective way to measure how much runs are worth in Colorado and Oakland and Detroit and Anaheim.” Objectively, a run is worth one run in every park.

    • Adam says:

      To correct a mangled sentence in my 10:45 comment:

      That doesn’t mean that PE isn’t a “natural” thing to try to correct for, but it does mean that it may not be properly corrected for with current techniques. And “properly” doesn’t mean “without error”, but “without bias.”

  21. Vidor says:

    I’ve wondered before at how park factor changes from year to year, and now that Posnanski provides an explanation, it seems like park factor is nonsense. You add up the runs scored every year? Really? IT’S THE SAME PARK. Did ownership move the fences? Did Al Davis close in the outfield so he could sell more football tickets? Did they change out the artificial turf for grass? If the answer to these and other similar questions is “no”, then park factor should not change. And if it is changing, which it obviously does since the people that preach this gospel just add up runs every year, then it is measuring something other than the inherent characteristics of the park.

    • BobDD says:

      Of course. So if 90% of the changes are easily explainable like with the examples you give (new dimensions, new stadiums, added seating, large weather change etc., would that make the last 10% worth more to a fan or less?

    • Vidor says:

      But those things usually don’t change. A park’s dimension remains constant from year to year. So the inherent characteristics of the park don’t change, so the “park factor” shouldn’t change. If the park factor is changing, it must be measuring something that has nothing to do with the park. The second possibility is that it is simply a deeply flawed statistic that isn’t measuring anything, at least not accurately.

      I am five feet, eight inches tall. If I start measuring myself on a monthly basis and one month I’m 5-10 and one month I’m 5-7 and then I’m 5-8 for two months in a row but then I’m 6-0 and then I’m 5-6, the logical reaction would not be to ponder just what is making me grow and shrink from month to month. The logical reaction would be to assume that either my ruler is bad (or I’m using a bunch of different rulers and some of them are bad) or that I am simply doing a really bad job of measuring my height.

    • OK, so let’s call it weather factors instead! Wind is a contributing factor, as well as the temperature during each game. The amount of rain can affect the infield and how fast ground balls travel.

      In addition, there are also team-specific factors, such as when Colorado started using humidors to affect the baseballs. That affected the park factors for Coors Field. Another team-specific factor example would be the SkyDome in Toronto, and how often the roof is closed vs. open.

  22. tarhoosier says:

    Comerica is MVP(ark)

  23. Matty Boy says:

    another factor here might be the larger context of parks throughout the league.

    i believe (i’m not sure and haven’t checked) that, while comerica has not changed, most changes to OTHER parks have made them more pitcher-friendly. i believe that most new parks since comerica opened have been pitchers’ parks, too.

    if comerica has stayed the same, but everyone else has gotten more pitcher-ey, then comerica would shift along the scale to being more hitter-friendly.

    just a theory.

  24. JD in KC says:

    Great article and discussion.

    I find many things easier to understand when I think of them in extremes. I have a 12-year old little leaguer who plays in parks with fences ranging from about 210 to 250 feet. Let’s say a field with fences at 225 feet was the home park for the Tigers. How many home runs would Miggy hit there? I don’t know… a lot. Let’s say he hits 125 home runs there, and 25 on the road for a total of 150 home runs.
    That would be great fun to watch. But if we were measuring his value vs. a player who played all his games in a MLB park, we would have to consider the park effects. This is just plain common sense. It wouldn’t be “punishing” Cabrera for hitting all those home runs at home, it would simply be common sense. While a little league park vs. a MLB park is a huge and ridiculous chasm, it is an extreme example. If you imagine those 2 fence lines in your mind getting closer and closer the hugeness and ridiculousness fades… but the general idea doesn’t go away, it just grays.
    I like Joe’s simple formula he shares for basic understanding. Whether you want to have “faith” in the further adjustments or not, the basis of the adjustment is common sense.

  25. kehnn13 says:

    I guess my concern is that park factor isn’t really park factor if it changes in the absence of any changes to the actual park. If the dimensions of the park are the same, but the team has some fluke hits at home, or gets some bad breaks in away games, the stat changes. I’d call that a statistic that does not indicate what the name implies that it should indicate (and it also turns it into a stat that includes a fudge factor to account for fluke hits and bad breaks, etc)
    In addition, as was already mentioned..AL teams have stronger offensive lineups in their ballpark for all home mixed league games (and weaker lineups in all away mixed league games) This is another factor now being accounted for by “Park Factor.”

    All in all, a very poor name for a pot luck of a statistic.

    • invitro says:

      What would you name it?

    • BobDD says:

      Even Imperfect information is not worthless. It is very difficult to not let additional information not be helpful.

      A fluke hit or bad break can easily change any of your favored stats, so I see no detraction for such here.

      “In addition as was already mentioned” – InterLeague games are accounted for just as well as home/away games because there are not more interleague games from one league than from the other.

      I wish “resistance is futile” came into play here.

    • nscadu 9 says:

      Traditional stats like BA, rbi are riddled with fluke hits or bad breaks or screaming liners yet we still use them. Park dimensions are not the only factor. Perhaps there is construction surrounding the park that changes the winds, weather is always a factor for how the ball carries and it isn’t constant every year. I wonder if Fenway has changed because of the seats above the monster. Dimensions haven’t changed, but the park has. I imagine infield clay and outfield grass settles as it ages, particularly in recently constructed parks. Funny how virtually everyone can agree that park is a factor, but for some unless they see it boiled down to a perfect number they don’t want to hear of it. I’d rather have a stat that attempts to quantify it to some degree rather than not quantifying it at all.

    • Rob Smith says:

      You don’t seem to understand how statistics work. If you get her enough data, “fluke hits” and “bad breaks” even out. So, what statisticians call “outliers” are averaged out. “Enough data” is defined by the designers of park factors as three years. Like ALL statistical models, they will never give you a 100% accurate number. But that doesn’t mean the number isn’t very accurate. In addition, to get closer to 100% accuracy, they add tweaks to the formula…. What you call “fudge factors”. That is, again, how statistical models work and how they are improved. Some people have asked some very food questions about this model, and some have pointed out flaws. But rejecting it just because the park factor stays the same is kind of a flat earth approach. Each year is a new year. This year the wind could blow out more than in. Next year the opposite. This year the turf is perfect, next year it’s patchy. This year it’s rainy and the air is heavier, next year, it’s dry and the ball carries better. This year the ball flies, next year they add a humidor. this year its cool and they leave the roof open more often. next year its hot and they close the roof more and play in air conditioning The dimensions may not change, but the atmosphere and everything else about the stadium can and does.

  26. Herb Smith says:

    I think JoePoz was simply trying to give a short, “thumbnail” look at park factors, and WAR in general.

    Do casual fans understand that WAR only includes things that happen on the field, and are in direct relation to how a player performs? WAR doesn’t include a lot of what many fans think are important.
    WAR is oblivious to:
    -an unadjusted ERA
    -ANY post-seaon heroics
    -ANY awards, including the league MVP, the World Series MVP, the Cy Young, being a starter in the All-Star game, etc.
    -whether your team won the pennant, or was even in the race at all

    Some folks can’t seem to fathom this, and think that Derek Jeter is a fine fielder because he’s won a plethora of Gold Gloves.
    Anyway, park factors are so obvious that I think people who pretend not to see them are not unlike the prison warden in Shawshank. (purposely obtuse)

    • Ross Holden says:

      I think there are 2 types of park factor skeptics:
      1. Those that question how the number is calculated because the logic is having the results assume that the park factors drive the results. Also they question it because they don’t see how park factors could change if the park doesn’t change. Using 3 years vs 1 helps remove this issue but doesn’t remove it completely.
      2. Those who don’t want to consider park factor at all. This one I can’t defend.

    • Vidor says:

      “Those that question how the number is calculated because the logic is having the results assume that the park factors drive the results. Also they question it because they don’t see how park factors could change if the park doesn’t change.”

      Exactly. Thank you. I phrased this differently in a post above–if I measure my height every month and my height keeps fluctuating an inch or two, I probably shouldn’t wonder why I grow and shrink.

  27. What about Park Factors by handedness? Most parks are good hitters parks for lefties or righties, no?

    • Ross Holden says:

      Yeah, and as Joe said there are better parks for singles vs better for homers, etc. So the beneficial park effects that help teams score more runs in Detroit may or may not be helping Miggy. So it’s a crude adjustment. Kinda like grading on a curve in school. And one could assume that other factors would cancel out, but like in school some class years are more competitive than others. Back to your original question, I wonder if Detroit helps righies or lefties more.

    • The baseball simulator game, Diamond Mind Baseball, each season spits out LHB and RHB park factors for 1B, 2B, 3B, and HRs for each park.

      Using the one-number, runs-based park factor is the quick & dirty park factor to use. But you likely account for ~90% of the variation at ~10% of the time and effort.

    • invitro says:

      I’ve got a question about that. To do the more fine-grained park factors, do you just change the data that’s going in from R to 3B by LHB? Because that seems easy :), compared to coding up all the corrections. Am I misunderstanding?

      Now applying the park factors may involve more than trivial programming to put in the finer numbers into the RAA (or whatever) calculation. But I am worried about extreme LHB/RHB parks… I believe parks exist that are 105 for LHB and 95 for RHB, and batters’ runs for those parks would be 5% off if handed park factors are not used. But I’m just guessing from a few things I don’t remember well or even correctly.

    • Yes, you’d just change the data … so 3B by all LHBs, both teams, home and away, obviously with one team being fixed for that “park” factor.

      Seems easy, I suppose, to calculate, if you already have all the necessary data at hand. So then perhaps using only runs-based park factor accounts for ~70% of the variation at ~50% of the effort. My percentages are poorly educated guesses.

      I’ve seen LHB/RHB outcome (HR, 3B, etc.) splits that are considerably beyond 105/95. See Baker Bowl, Philadelphia, with its HR factor for both LHB and RHB. DMB has them both at 155 for the 1937 season.

    • invitro says:

      Yes, I liked using the Baker Bowl occasionally when I played that game. It seemed odd that the HR factor would be identically so high (227) for both LHB and RHB given how asymmetric the park is.

      I’m looking for a park that would have a combined (1B, 2B, 3B, and HR combined into R) run park factor of 105 for LHB and 95 for RHB (or vice-versa), but I can’t find one that is unbalanced for 1B and HR in the same way. Wait, here’s Marlins Park, 2012:

      1B 2B 3B HR
      LHB 101 91 179 68
      RHB 110 96 89 74

      Which on b-r is:

      multi-year: Batting – 101, Pitching – 102 · one-year: Batting – 99, Pitching – 100

      And so I wonder if the b-r park correction is helping RHB and hurting LHB more than it should.

      (I’m just curious… definitely not criticizing… I know it’s small potatoes)

    • invitro says:

      Now I’m curious why WAR doesn’t use the more detailed park factors. It seems that to be as accurate as possible in converting to a neutral park, you’d want to use them. I don’t have a feel for how the magnitude of this “correction” would compare to that of away parks or pitchers not having to face their team’s batters.

  28. Joe says:

    This is why America can’t fill it’s science and engineering jobs. “I don’t understand this math at all, so I’m going to give up trying to learn something new and just ridicule it!”

  29. invitro says:

    baseball-reference has a one park factor for hitters and one for pitchers. I don’t understand why there are two instead of one, and what the differences in calculation are. The about page just says “Calculated separately for batters and pitchers.” I feel like I’m missing something obvious… what is it?

  30. Sorry to chime in late on the discussion, but I had a little spark of thought about what Joe said and what I think Mark Daniel was getting at above.

    Let’s assume for the moment that because of familiarity or whatever, teams score 10% fewer runs on the road than they do at home, and allow 10% more runs on the road than at home. Let’s suppose for ease of calculation the average team scores and allows about 4 runs per game. With a mathematical fudge, their average home score would be 4.2 runs to 3.8 runs, and on the road the average score would be reversed. The end result of average teams playing in their average parks would be, over 100 games (again using 100 home games and away games (100 chosen to make the math easy), at home 800 runs would be scored and on the road 800 runs would be scored. The park factor would be 100, as expected.

    But let’s alter this so that you have an otherworldly offense on a very good team. Let’s suppose you have a team that averages 6 runs a game while allowing 4 runs per game. At home they would score 6.3 runs a game, and on the road that drops to 5.7 runs a game. The runs allowed would be the same. In this instance, over 100 home games you would have 630 + 380 runs scored, for 1010 runs total, while on the road there would be 570 + 420 runs allowed, or 990 runs total. Voila, you’ve created a park factor of about 102 simply by giving the home team a larger absolute spread in runs scored. You’ve done nothing to the park.

    I don’t think that this is something that would add up to all of the difference you would see. But I would expect the park factor to incorporate the home team’s makeup (whether a good pitching or good hitting team), especially if the home field advantage is for some reason particularly large.

    • invitro says:

      This is interesting. I’m going to first repeat your example with different numbers. Let’s suppose we have a team that in a neutral park would score 420 runs and allow 293 runs in 81 games. Let’s then assume that home field advantage for every team in the league results in scoring 8.8% more runs at home, allowing 8.8% less, and on the road, scoring 8.8% less and 8.8% more (I hope this is consistent). This team would have these numbers:

      at home: score 1.088(420) = 457, allow 0.912(293) = 267.
      on road: score 0.912(420) = 383, allow 1.088(293) = 319.
      total: score 840, allow 586.

      The raw park factor would be (457+267)/(383+319) = 1.03, when it seems that it should be 1.00.

      The 1975 Cincinnati Reds had an extreme home field advantage in terms of wins; they went 64-17 at home and 44-37 on the road. They were an incredible offensive team, and a very good defensive one.

      Here are their home/road run numbers:

      Scored Allowed
      Total 840 586
      Home 457 275
      Road 383 311

      That’s the same as my fake team, except the Reds allowed 8 more runs at home, and allowed 8 fewer on the road. The runs scored are the same.
      (Now 8 is 3% of 289, which means something, but I’m not sure what.)

      The actual b-r park numbers are
      multi-year: Batting – 102, Pitching – 99 · one-year: Batting – 104, Pitching – 101

      Ok, I don’t know what to make of this, but it’s interesting. I hoped that this Reds team would serve as a nice real example of this effect… but I worry that 8 runs and 3% are too large, and that the park effect in this case is doing just what it should be doing.

    • Phil says:

      This comment has been removed by the author.

    • Phil says:

      In the work I had published by SABR about measuring relative team quality using standardized run differential adjusted for league size, one conclusion was that the size of the differential is all that matters — not its percentage of offense or defense. It’s a little controversial, as it considers a 10-9 victory as meaningful as a 1-0 one: which makes some sense, as it’s two tied teams separated by the thinnest of margins. My point is that in a different park, a team will have the same actual differential: not the same proportional one. So instead of “10% more runs at home” (or whatever), it’s more consistent to say “1.5 more runs per game at home” — and then maybe the park factors will be consistent. I would argue that you don’t give a team the same proportional spread in a more offense-friendly park: preserving the same differential actually keeps the team consistent, relative to its competition.

    • Quite possibly. I didn’t mean to suggest that this was something that necessarily happens, but that it could be what people think of when suggesting that park factors follow along with the differing skills of the home team. I doubt it’s a huge portion of it (how many teams score two runs more than they allow?) but it might be what they were considering.

  31. brhalbleib says:

    Maybe the more complicated ways of determining park factors takes care of this, but Joe’s simplified way does NOT take of the problem that, when comparing one team to another, their defenses’ capabilities in comparison to each other can change, depending on the park. In other words, if park factors were truly a constant, it shouldn’t matter which park Team A and Team B play, if Team A is 30% better defensively than Team B, that 30% difference would be apparent.

    But at least with outfielders, you can see that having Adam Dunn as your regular LF if your home games are in Fenway Park isn’t nearly as damaging to your defense than having him play LF for you if your home games are in the Astrodome, or Kaufmann Stadium or Coors Field. In other words, having Brett Gardner in LF over Adam Dunn is more important in those big ballparks than one with a small LF area (because some of the balls that Gardner gets to that Dunn doesn’t at Coors Field are off the Monster anyway) If you play in a big ballpark and you have an Adam Dunn type LF, some of his crappiness is going to be reflected in the park factor (which we don’t want)

  32. […] park (and it’s not the ‘Miguel Cabrera’ phenominon as Joe Posnanski explains here).  Don’t only use this for your draft, if you’re trying to decide on which player to […]

  33. Twinkle Toes says:

    I feel like 2, maybe 3 people have ever studied statistics (or even math for that matter) in this post.

Leave a Reply

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