July 8, 2013
Analyzing Rogers Centre's Roof
Earlier this year, there was some talk in the Toronto papers about Rogers Centre and its status as a new “home-run haven.” The linked article is mostly a case of failing to stop writing once you have to say “it’s early, but...” However, the ideas described within it are consistent with conventional wisdom surrounding the stadium. Namely, that there are differences in batted-ball outcomes depending on whether the roof is open or closed.
When not in use, the shell-like components of the Rogers Centre roof are stacked together and stored above center field. (If you’re unfamiliar with the setup, see here for a good picture.) Theoretically, this shape causes a “wind scoop” with the gusts coming off the nearby lakeshore. That wind then swirls around in the outfield, blows down onto the field, and suppresses the distance of fly balls that would otherwise go over the fence. During early-season closed-roof games you will invariably hear visiting broadcasters describe this fly-ball-suppression effect, and most of them accept the theory as stated. After all, it makes sense. But does it actually exist?
All information here comes from the Retrosheet game logs and event files. (Glory and praise to Retrosheet.) Each game is already marked as an indoor or outdoor game, and I wanted to extract the game-time temperature too, but some games have missing information in that field. However, most indoor games in any given year had the exact same temperature (where available), so I filled in the missing games with the season average: 72 °F prior to 2004, and 68 °F for 2004 to 2012. The outdoor games without temperature information were discarded.
This leaves us with 1826 regular-season and 18 playoff games where it is known whether the dome was open, and what the first-pitch temperature was. There are 1905 games in total, so we have about 97 percent coverage here.
Let’s test that wind-scoop theory. Some home runs that are pushed down by the wind would turn into doubles or singles, presumably, but if the theory is true, it’s flies with longer hang times that are most at risk of carrying over the wall—balls that would otherwise be caught at the track, say. So in addition to home runs, we’ll also take any fly ball scored as an outfield putout or a dropped ball error by the outfielder (i.e., 7, 8, 9, E7, E8, E9). That leaves out all singles, doubles, and triples, but they aren’t always broken down into fly balls and ground balls in the Retrosheet data, and we’re specifically looking at theoretically-catchable flies that carried or did not carry over the wall. (For what it’s worth, if we add all the hits in there, treating them as fly balls where no batted-ball info is known, the results don’t change.)
This isn't much of a difference at all, statistically or practically speaking. The average game at Rogers Centre over the years has about 15.5 such fly balls, and 35 games per year are played with the roof closed, which implies less than one additional home run per year in Toronto due to the closed roof. Or not even half a homer per team, since we're looking at both the Jays and their opponents here.
But maybe we can find evidence that a closed roof increases offense in other ways.
First, though, let's acknowledge what's obvious to any Blue Jays fan: Toronto’s April weather is not pleasant. That’s the entire reason that the stadium even has a roof. Which means that games where the roof is open are played, disproportionately, on nice days.
Here is the distribution of game-time temperatures for roof-open games. Remember that all indoor games are played between 68 and 72.
Less than one game in 10 is played with a first-pitch temperature below 63 °F / 17 °C. That is not the case in other northern cities—e.g., nearly half of all games played in the new(er) stadia in Pittsburgh, Cleveland, and Detroit had a first-pitch temperature below 52. This may suggest that Rogers Centre's status as a hitter's park when closed is due to the fact that the weather is warm or better, and not necessarily because of any particular quirks of the roof or wind.
With the previous finding in mind, let's look at three sets of games at Rogers Centre: all games with the roof closed, all games with the roof open, and roof-open games where the temperature was between 68 and 72.
After controlling for weather, we see there are fewer home runs in roof-open, moderate-weather games than in roof-closed games, but only to a small extent: 0.8 more HR per year per team. Maybe if the Jays have an extreme fly-ball pitcher, it’s best to pitch him in road games where possible until the roof is open in May. Even then, though, the effect is very small.
Anything interesting in the batting lines? Counting stats are per 600 PA:
All the same, basically. There's no evidence that a closed roof, on its own, increases offense relative to an open roof with the same temperature.
Unfortunately, the comparison we really want to make—closed-roof game vs. the cold-temperature outdoor game that would be played if there were no roof—isn't possible. Rogers Centre opened in 1989, and since then the Jays have never played a game with a temperature recorded below 50 °F. (There’s this game, but I treated that as 20 °C, not 20 °F, which is impossible.)
As mentioned, I suspect the main reason that the Dome plays as a hitter's park has nothing to do with wind scoops, but simply the fact that having the roof eliminates any cold-weather games. Essentially, the Jays and Brewers are the only teams in the north who can play games in climate-controlled environments; everyone else is obliged to deal with cool weather when it occurs. This makes Rogers Centre look more hitter-friendly by comparison. Or, to put it another way: if you played every Cleveland April home game in “72 and sunny” conditions, I daresay the Jake would appear more favorable to hitters early in the year, too.
We'll have to live with these limited results for now, which suggest almost no roof-closed effect when the outdoor weather is pleasant, and tell us nothing about the effect of closing the roof on a cold day.