Then and now: Myth-busting Rivalry Week's old "throw the records out" saying

It's Rivalry Week, and that means you'll hear lots of people say some version of this statement: "You can throw out the records when those two teams play."

Here's the reality, though: We probably shouldn't throw the records out the window. The old cliché revolves around the idea that with high emotions, bragging rights at stake, teams and players who are familiar with one another, maybe a chance for one team to salvage its season, then anything can happen.

We saw a good case of that earlier this year when Texas beat 16-point favorite Oklahoma after the Longhorns had just suffered their worst loss (52-7 to TCU) in more than a decade. And who can forget the way Auburn dashed Alabama's national title hopes as a double-digit underdog in the "Kick Six" game two years ago?

So is there truth to this belief? Are there more upsets than usual in rivalry games?

To answer this question, we polled a group of college football experts from ESPN's Stats & Information Group to get a list of the top 20 current rivalries. The only stipulations were that the teams must have met every year over the past 10 seasons and no team could have more than two rivals. Then we examined how the favorites in these rivalry games did compared with expectations and with all other games in the 10-year sample.

Although this exercise began with the premise that rivalry games truly are different from any other matchup, the results proved the opposite. In fact, an underdog team in a rivalry game has actually performed worse than expected against its nemesis than against all other teams.

Dating to 2005, the favorite in the 20 rivalries had a .739 winning percentage, a higher percentage than was predicted by ESPN's Football Power Index entering the game (.728). The favorites' point differential in those games (plus-10.1 points) was marginally greater than the predicted point differential (plus-9.7).

So maybe the favorite in those games performed slightly better than expected, but there sure seem to be more "big upsets" in rivalry games over the years, right?

Again, this does not prove true. In that sample, there were 46 "big upset chances," or games in which the favorite had at least an 85 percent chance to win. But only four of those games ended in an upset. That's about the number of upsets that would be expected given each team's chance to win entering the game. It's human nature to remember the biggest upsets in the biggest games of the year, but the reality is that underdogs in rivalry games are just as unlikely to win as in any other game.

Grouping all rivalry games, however, might be an overgeneralization. For example, Florida State versus Miami proved to be the most unpredictable rivalry game over the past 10 seasons, and the favorite in Michigan-Ohio State mostly held serve. Going back further, which would include Michigan knocking off No. 2 Ohio State in 1995 and 1996, the results could differ.

This weekend there are a number of rivalry games with conference championship and playoff implications. Let's dig deeper into the three of the biggest rivalry games of the weekend.

Bedlam: Oklahoma vs. Oklahoma State

Favorite's expected win percentage past 10 years: .747

Favorite's W-L past 10 years: 8-2

The game in Stillwater might have the biggest impact on the College Football Playoff of any matchup this week. Oklahoma fans are hoping for a different result than last year, when the Cowboys, who were coming off five straight losses, knocked off the 20th-ranked Sooners. Entering that game, FPI projected that Oklahoma had a 92 percent chance to win, making it not only one of the biggest upsets in the history of Bedlam but also the most unlikely win for an underdog in the 200-plus rivalry games included in this study.

In 2013, there was a similar though not as unexpected result when Oklahoma beat then-No. 6 Oklahoma State in Stillwater. Can we expect more mayhem in this year's Bedlam game? FPI gives Oklahoma a 68 percent chance to beat Oklahoma State. A win by the Sooners would give them the Big 12 tiebreaker and move them one step closer to a playoff berth.

The Game: Ohio State vs. Michigan

Favorite's expected win percentage past 10 years: .779

Favorite's W-L past 10 years: 10-0

To the general public, this game lost some of its playoff luster when Ohio State lost to Michigan State last week. To these respective fan bases, though, the 112th meeting between these rivals means as much as ever. Despite Michigan's hosting the game, ESPN's FPI gives Ohio State a 58 percent chance to win. That might not be good news for Michigan fans; the FPI favorite has won each of the past 10 meetings between these teams by, on average, nearly the exact scoring margin that FPI predicted entering those games.

The Iron Bowl: Alabama vs. Auburn

Favorite's expected win percentage past 10 years: .757

Favorite's W-L past 10 years: 8-2

This game looked as if it would mean a lot more at the beginning of the season, but it still holds weight in the state of Alabama. The two Iron Bowl upsets over the past 10 years have been two of most memorable games in recent memory -- Kick Six in 2013 and Cam Newton's valiant comeback in 2010. Both of those games featured matchups of teams in the AP top 10, but FPI gave the favorite at least a 68 percent chance to win in each of those games. In the other eight Iron Bowl matchups in the past 10 seasons, the favorite won by an average margin of about 19 points. It might be another lopsided game this Saturday, when Alabama heads to Auburn. ESPN's FPI gives the Crimson Tide a 78 percent chance to win, and if they do, they will clinch a spot in the SEC title game.