Monday, June 21, 2010

Orioles Attendance

Yesterday, the Baltimore Sun had a big article stating that, despite the Orioles miserable start, attendance at Camden Yards has actually increased since last season.  Here's their assertion: "Even with the Orioles' ineptitude, attendance is up 9.5 percent, from 21,653 to 23,720, through 32 home games compared with the same number of games last year, according to Major League Baseball."  Technically, that is true, but it certainly doesn't tell the whole story.  The Sun does offer one caveat, noting that the Orioles have had an attractive slate of opponents in the early going, hosting "the Boston Red Sox in two weekend series, the Yankees in two midweek series and the New York Mets on a weekend,"  while at this point last year the club "had hosted two Yankees series — one on the weekend and one during the week — and no Red Sox games," but they did nothing to quantify the effect.  Since they couldn't be bothered, I will do so for them!

First, let's try and build a model that would predict attendance for any single game.  What sort of factors might you be interested in?  I identified several variables to test: 1) Day of the Week, 2) Attractive Opponent (defined as Yankees, Red Sox and Mets), 4) Opening Day, and 5) Summer (defined as between Memorial Day and Labor Day).  Now, obviously, you could also add in things like win loss record, game time temperature, and day game vs. night game, but I'm going for a quick and dirty model here (plus, this is the data I can easily access).  After running the regressions on 2009 data*, I came up the following statistically signficant variables (95% confidence interval): Attractive Opponent, Friday, Saturday & Sunday games, Opening Day, and Summer.  Thus, games taking place on Tuesday, Wednesday, and Thursday do experience a change in attendance from Monday nights (the baseline case).  I think those results are very inuitive.  Here's the regression equation:

Attendance = 13,803 + AttOpp(12,185) + Friday(9,538) + Saturday(12,557) + Sunday(6,519) + OpenDay(22,617) + Summer(4,256)

Using that equation, we would expect attendance for the first 32 games of 2010 to be 25,889.  Instead, actual attendance came in 8.4% below that number.  So where did the Sun go wrong in its analysis?

First, all those attractive opponents make a huge difference.  In 2009, the Orioles played a total of 21 home games against the Red Sox, Yankees and Mets (average attendance of 32,747 including Opening Day), but had played just six of those games by the 32nd home date (average attendance 35,340); they've already played 15 games against those teams in 2010 and averaged just 28,299 fans.

Next, the use of 32 games--and not the calendar date--is also significant.  In 2009, the Orioles played home game number 32 on Thursday, June 11.  On Friday, June 12, the O's welcomed in the Atlanta Braves for the start of a summer weekend series that averaged 28,295 fans.  This year, home game 32 occurred on Sunday, June 13.  If you include that Atlanta series to normalize the date, the year on year comparison is less favorable, with the Sun's 2009 number moving from 21,653 to 22,222.

The use of game 32 (and the omission of another weekend series) highlights another important point: the day of the week.  The 2009 Orioles played 20 of their first 32 home games a Monday, Tuesday, Wednesday or Thursday; the 2010 O's contested just 17 of 32 on those days of the week.

Thus, when adjusting for opponent, day of the week, and the calendar, the Orioles 2010 attendance is almost assuredly running lower than last season, and I suspect that upcoming series (serieses? serieii?) against the Marlins, Nationals, Athletics, Blue Jays, Twins, Angels, White Sox, Mariners and Rangers (the O's next nine home series) will reveal just that.

*If you would like to see the data set, please email me.

4 comments:

dmbfanmd said...

I see many typos.
-Srikar

The Oriole Way said...

nice... what is the t-statisTICK?

Michael said...

What kind of R-square are we talking about here?

The Oriole Way said...

R-squared of 0.693