THE BOOK--Playing The Percentages In Baseball

Very interesting.  One question I had was whether this may be capturing some day/night differences.  That is, are game-time temps higher for day games than night games?  It would probably be good to look at night games only to see if patterns remain as strong.

Great point.  Back when I was growing up (the 80s), the day/night issue was a much bigger deal than it is today, because of the sheer number, and the lighting of the day.

Another point, it doesn’t look like Chris adjusted for the quality of the pitcher/batter.  While unlikely as an effect, I’d still like to see an assertion as to the quality of the players in each pool.

Day/night issues are interesting and complex.  Day run scoring and the individual stats are similar to night run scoring and individual stats.  That is counterintuitive.

The reason is that the pool of batters is different.  Players get rested in dat games (usually after night games).

If you control for the players, you will see run scoring and the individual components being different/higher.

What I found is that temp and better lighting (presumably) BOTH contribute to the difference between day and night games.

Also, night games used to be a lot different in the old days.  For example, power pitchers USED to have a big advantage in night games.  Not any more.  I assume that is because the quality of the lighting is much better now.

Great point about the day/night difference.
The article was just a bunch of descriptive statistics to get the conversation started. There are a bunch of things I need to do to really isolate the temperature effects; controlling for time-of-day effects is one of them.

The quality of players in each ‘temperature pool’ doesn’t seem significantly different from one another. I’ll be more specific in a follow-up article, though. Also, just to clarify; I excluded postseason games from the sample because playoff games would give good players too much weight in the cold weather categories.

Because of the seasonal bias in cold/hot weather studies, you really have to be careful about the player pools AND you have to be careful about seasonal differences independent of weather. 

There is some evidence, for example, that certain, or perhaps most or all, players may perform differently at the beginning, middle, and end of the season.

"There is some evidence, for example, that certain, or perhaps most or all, players may perform differently at the beginning, middle, and end of the season.”

Right - I’m already working on a model that controls for time of year. Now I need to throw in other things such as time of day, player type differences, and regional differences (particularly because the temperature/humidity relationship varies across locations).

I’m also doing similar things with wind speed/direction but that’s a whole other story.

Also, night games used to be a lot different in the old days.  For example, power pitchers USED to have a big advantage in night games.  Not any more.  I assume that is because the quality of the lighting is much better now.

Lighting is better.  Also, night games start earlier.  I think most teams start them around 7 PM local time now.  Back when I was growing up in the 70’s and 80’s, they usually started at 7:30 or 8PM.  I’m not sure if this makes much of a difference, though.  Twilight may be more difficult to hit in than daylight or full night lighting.

A followup from Chris:
http://www.hardballtimes.com/main/printarticle/the-truth-about-april-home-runs/

A batted ball has a 4.0% chance of leaving the park during a game played in 70 degree conditions, but only a 3.5% chance of becoming a home run in a game played in 50 degree conditions.

FWIW, Last night, I just happened to be toying with something called Hit Modeler at NASA’s website (http://www.grc.nasa.gov/WWW/K-12/airplane/hitmod.html). I stumbled on it looking for another hit modeler I used in the past (I searched google for: batted ball simulator).  For temperature, they have a “Specify” option, which I never got to work, but using their Hot Day (90 degrees farenheit), Normal Day (60 degrees farenheit), and Cold Day (35 degrees farenheit), I came up with .31 feet (3.71 inches)per degree farenheit, at least using their preset for all other data points. I can’t load the site at work, but planned on trying the “weather only” one to see if I can get the “specify” option to work.  I’m not sure if that all sounds right though, a 20 degree difference would be 6.2 feet more.  Would that really be enough explain a .5% increase in HR/Batted Ball?  It sure doesn’t seem like it, but that’s only for balls hit at a 45 degree angle at 700 feet elevation, so who knows.  Also, I’m not sure of the accuracy of the program, but I bet if there were anybody that would know how weather and atmospheric conditions would affect the aerodynamics of an object moving through the air, it’d be NASA. 

The same page has several pages regarding Aerodynamics in baseball : http://www.grc.nasa.gov/WWW/K-12/airplane/short.html.  If you scroll to the bottom there is section titled Baseballs.  The page (K-12 is a give away) is aimed at school kids, but the different modelers (like the curveball modeler), calculators, and formulas are pretty cool, to me at least.

I realize the obvious flaws in using a model like that to determining the effects of temperature, accuracy not withstanding.  For one, balls are hit at a lot of different angles than 45 degrees, which looks to be calculated as an optimal “home-run” angle (or I’d guess so).  I suspect you’d need to get the distribution of the angle the balls are hit at.  There are other factors as well, like how the hitters do in the cold (as well as the pitchers), or what effect the temperature or humidity has on the pitch itself.  Also modeler only goes to 108 mph for the ball speed, which (if I read the site correctly) is the speed of the ball.  If that’s the speed off of the bat, that doesn’t seem right.  That’d make more sense if it was pitch speed, but it doesn’t look like it is.

BTW, Nate Silver had interesting piece on this recently, that referenced Chris’ work:  http://www.baseballprospectus.com/article.php?articleid=6139