clock menu more-arrow no yes mobile

Filed under:

Response to Kyle's reponse to my response to his response to my post

In my last post, I wrote:

I think two broad conclusion that can be drawn are that a) it's pretty difficult to do just a cursory examination of home field advantage given the large sample sets required to draw any significant conclusions from them, and b) someone needs to actually do it!

Well, hats off to Kyle: he has done the work, and went back to 1892 to determine UGA's official record in games played in Atlanta. Turns out they don't have such a huge home field advantage after all - they're 22-29-5 over that span in Atlanta. Once again, I encourage readers to check it out.

Couple things though. Kyle notes:

How is it possible to square that fallback position with the initial boldfaced claim of home field advantage? And the Valley Shook offered an assertion (that treating a game played at the Georgia Dome as "a Road/Neutral game for the Georgia Bulldogs" "may be a bit of a stretch"), then backtracked and insisted that "there's no way we can draw anything close to a factual conclusion" based upon the games to which I referred . . . which are the only games there are upon which to base his initial declaration.

Saying I "backtracked" is disingenuous. I offered a qualitative assertion that it's a home field advantage, and when finding out that there are only six games in the sample which he'd presented, concluded that we have to treat it as a qualitative rather than a statistically significant result. I defy you to find me one statistician on the planet who'll be satisfied drawing significant conclusions from a sample set of six. (For instance, had WVU only gone up 3 TDs early rather than 4 in the Sugar Bowl, UGA could well have come back and won it, given they only lost by 3. Then UGA's winning percentage in Atlanta jumps to 0.667 in those games, and suddenly they look great.)

I get the feeling Kyle was in a grumpy mood after having done all the work, so he added this little zinger:

Given the foregoing facts, I feel quite comfortable reiterating my assertion that A.T.V.S.'s initial assertion is sheer nonsense. A.T.V.S., however, has stated that, due to the small available sample size, "this becomes a purely qualitative exercise rather than a statistical one."

This is a convenient way of saying that, when historical facts and quantifiable results don't support an obviously baseless conclusion, over a century of history and over 55 games may be cast aside as irrelevant to an amorphous expert assessment of the advantages enjoyed by a school I attended in a geographic area in which I have lived for nearly 38 years.

That's just false, and probably beneath Kyle on most days. He's arguably the best writer in the entire SBN network, but I have no clue where this came from other than perhaps sheer exasperation on his part. I never once knew anything of a sample set of 55 games (that is, unless one were to impute knowledge of the entire history of Georgia Bulldogs football to my LSU fan memory), just Kyle's having given me six games to work with in his prior post - and that's where I drew my conclusion. Hence my suggestion, as quoted above, that it would take a lot more work and somebody ought to do it. I say again, hats off to Kyle for actually doing it.

My point about 30 games being generally too small for a meaningful conclusion was in reference back to SMQ's analysis that I was discussing in my original post. In fact, now that I'm looking into that position of mine specifically, I should note that I think I'm actually wrong and SHOULD backtrack here - I think it's fair to draw a relatively solid statistical conclusion from a sample set of 30. See here for a t-distribution table and note that the value (say at a 95% confidence interval) for 30 degrees of freedom is actually not that far off from infinite degrees of freedom.

Couple other points. Kyle:

I am curious, though, why A.T.V.S. ever ran "the actual numbers" at all. Earlier, he said that "a sample set of 30 is generally too small," as a result of which "this becomes a purely qualitative exercise rather than a statistical one." If that's true, why is A.T.V.S. rebutting Sunday Morning Quarterback's numbers with more numbers?

Agreed. I didn't have time to go compile the actual numbers going back multiple decades, so I just went with what I had to work with, making a couple minor adjustments. It was quick (20 minutes or so) and to the point. That's it. The point of the whole exercise was taking a stab at what would be the sort of analysis I'd use over a larger data set.

From Kyle:

Treating the Red and Black's games in the Georgia Dome as Bulldog home games is arbitrary and insupportable, particularly in light of the fact that A.T.V.S. didn't indicate that he was also treating the Bayou Bengals' games in the Superdome as L.S.U. home games

First, LSU played no SEC games in New Orleans over SMQ's time span. In fact, I don't know if we ever have. Secondly, I will say that only making those adjustments and not for Bama, for instance, does leave the analysis incomplete. Again, it was a time issue and I had no desire to go through each and every team to sort that out, which I made abundantly clear in my original post:

Since we're working with percentages here rather than absolutes, and since I don't have the time to go aggregate the data for the 11 other schools, I'm going to drop LSU's SEC record from those two seasons (2-6 home, 1-7 road).

More from Kyle:

A.T.V.S. asks about "the distribution of Red vs Purple shirts in the stands for any of the LSU-UGA battles in Atlanta." I don't know what it was, but I know what it could have been: 50/50. One of the reasons neutral sites are neutral sites is that the tickets are allocated evenly between the opposing teams.

It's not a question of selling your own allotment. Each team gets about 16,000 tickets. The remainder go to the SEC (evenly distributed), Georgia Dome sponsors and suite holders. The odds that THAT contingent - notably here the Georgia Dome sponsors - would be evenly split between, say, LSU or West Virginia versus Georgia are rather low, in my opinion.

So then, that brings me to: why would we non-UGA fans be so silly to believe that UGA has any sort of home field advantage in Atlanta? Maybe it's because we've been fed it, by everyone from writers to the Georgia Bulldogs themselves. From an Associated Press 2002 SEC Championship Preview titled "Georgia has definite homefield advantage in SEC title game":

"It's almost a home game for us," receiver Terrence Edwards said...

Both schools received 15,900 tickets, but the Razorbacks sold 2,200 from their allotment to Georgia. In addition, nearly all the remaining seats at the 72,000-capacity dome probably will be filled with fans of the home-state team.

Georgia's very own Terrence Edwards would strike me as the sort of expert that Kyle is looking for. And I don't think it's a stretch at all to assume that the 40,200 tickets not allotted to either team wind up finding their way to more Georgia fans than those of the opponent (given that sponsors and suite holders at the Georgia dome likely swing for the local team - I don't think this is an absurd theory). Not enough? Ok.

2005 SEC Championship Preview (cached, sorry) from the Augusta Chronicle (actually I gather this particular writer is syndicated across Georgia, as the artle was also in OnlineAthens), titled "Dome Field Advantage for Dogs":

As long as the game is in Georgia, Bulldogs safety Greg Blue says it tilts the way of the Bulldogs.

"It's a good advantage," Blue said. "They've got to come to us. If we play them at Valdosta, it would be a home game for us."

Same game, different preview from ESPN.com:

While the Tigers are the designated home team, they know that most of the fans -- two-thirds or more, if past games are any indication -- will be wearing red and black.

Hmmm. But this writer could be suspect. How does he know if "past games are any indication" on the subject? Let's then go to a writer who actually did go to a past game. From the 2003 SEC Championship Game Recap on ESPN.com:

The Bulldogs were playing just an hour's drive from their campus, cheered on by a crowd that appeared to be three-fourths red and black. At the end, most of the Georgia fans were gone.

Ok this post has gone on long enough, I don't really have anything else to add.

Conclusions:

  1. The world has had very legitimate reasons to believe Georgia derives a home field advantage from playing in Atlanta, making it ridiculous to label anyone who believes that to be speaking "sheer nonsense."
  2. The data, however, show a different story, as Kyle demonstrated that UGA has compiled only a 22-29-5 record in non-Georgia Tech games in Atlanta since 1892. Good enough for me.