This post analyzes the popular notion that records and other overall season performance measurements should not be considered in predicting rivalry games, specifically the Virginia - Virginia Tech rivalry, due to the emotion inherent in the matchup.
"Strap on your hard hats, hold on to your seats, and throw out the records folks, we've got a good old-fashioned hard-nosed battle between two historic rivals in Blacksburg, VA."
As you can see, I took the liberty of writing ESPNU's opening line for them. In any football rivalry, numerous commentators, coaches, and fans cling to the Lombardi-esque cliche1 that emotion can magically even the playing field. It's a tempting thought. It's especially tempting when your alma mater has perpetrated a mercurial season filled with more bad than good. Even the illogically rabid fan inside me, hidden under layers of numbers and graphs,2 stirs at the mention of an emotional victory over Virginia Tech.
The Hoos travel to Blacksburg this weekend in their annual attempt to end what has become a protracted losing streak to the fearsome Virginia Polytechnic Institute and State University Turkeys.3 Should we, as fans, heed the old-fashioned advice, "throw out the records,"4 and expect a Cavalier victory? Or should we guard our expectations, instead betting our emotional well-being on the big picture progress of the program?
I attempted to answer this question and evaluate the accuracy of the timeless saying by analyzing the predictive value of various pre-game metrics. The metrics, including team record, are indications of a given team's quality entering a game. These measurements usually used to predict a game winner are based on past performances and largely ignore the emotion that might arise for a rivalry game. In a true "throw out the records" rivalry world, none of these metrics should be more predictive than a series of random coin tosses.5
I limited my analysis to the last ten years of the rivalry to create a sufficiently large sample while avoiding the bias of known streak cut-offs.6 The below chart7 depicts the last ten years of the rivalry along with the scenario entering Saturday's game, including the respective team records, team Sagarin ranks, Vegas predicted winner,8 and eventual game winner.9
I then compared the correlation of the better record, the higher Sagarin rank, and the predicted Vegas winner to the actual winner. A "throw the records out" (TTRO) coin flip or guess would have an expected success rate of 50%. The following chart shows the success rates of the various prediction methods.
The probability of ten successful coin-flip guesses out of ten, to match the success of the Sagarin ranks and the Vegas picks, is .098%.10 The probability of seven of eight successful random guesses, to match the record-based prediction, is about 3.1%.
At least in the Tech rivalry, pre-game metrics are quite strongly predictive of the final winner, and are all very likely a better predictor than "‘throwing out the records" and guessing. This particular announcing cliche appears to just be a baseless saying aimed only at creating enthusiasm among the respective fan bases. Temper your expectations. These trends spell trouble for the Hoos this weekend.
1 while we're on the subject of commentators and their sayings, did everyone hear Jesse Palmer refer to throwing the football as "spinning the magic bean" during last Thursday's game? Jesse, stop trying to make "spin the magic bean" happen. It's NOT going to happen.
2 and these really are just a facade for holding back some rampant homerism
3 the proximity of Thanksgiving to the annual matchup with the Hokies simply proves that the Pilgrims were the earliest noted UVA fans. Why else would they have chosen to metaphorically celebrate an impending Hoos victory via a turkey feast just days before the game? By logical extension, most Americans, including some apparently self-loathing Hokies, are huge UVA fans.
4 which Tech fans probably wouldn't object to either
5 since the very concept of the "records" saying is that the seemingly worse team has a legitimate chance of victory.
6 such as the Hokies winning the last 8, and 12 of the last 13, but also 12 of the last 14. In other words, the sample could be chosen with an end goal in mind, so I just picked a random number.
7 CAUTION: the following chart contains grotesque imagery and is not for the faint of heart. Viewer discretion is advised.
8 the Vegas prediction is kind of the odd one out here, but it is also based on past performances, and thus should be subject to the same rivalry "records" error ... if it exists
9 it's like a blue glimmer of hope in a sea of ... whatever that other color is
10 or 1 in 1024