The study in the post’s title is the painstaking and thorough research I’ve done on teams that were in the position to choose to rest starters in the final week from 2003 to the present (In reality assembling this info was very painstaking, but the data’s certainly not thorough as an empirical study). Since information was more difficult to collect for earlier years, and since I found assembling it boring, 2003 just represents the point where the misery of examining this exceeded my pain threshold; there’s no principled reason to cut it off there, and I’d be deeply appreciative of anyone who thought this was interesting enough to look back further (i’ll post anything that anyone comes up with).
So what’d I do and what’re the results? I took every team whose record indicated they might have been able to rest their starters in the final week or two without an real negative impact on their playoff standings (so, for example, the Patriots and Bengals this year would count even though their loss affected their standing as the 3 or 4 seed, since neither position was more desirable). I then checked Wikipedia, their box scores, and some contemporary news reports to see if they had rested their starters and what the ultimate result of their season was. I coded teams reaching the conference championship as a good result and teams that didn’t as a bad result. Other ways of coding outcomes (did they lose to a better or worse seed? for example) seemed more difficult and don’t seem to have affected the final result (since the teams that didn’t rest perform extremely well on any measure)
The executive summary is that there’s not enough teams that chose not to rest their starters to prove it’s better as an empirical matter, but all of the non-resting teams achieved good outcomes whereas a bit less than half of the resting ones managed to pull off a good result. Here’s what happened: since 2003, there were 21 teams I identified who had the opportunity to rest their starters. Of those, fifteen clearly did so. Seven of them had good results, and eight did not. Of the six remaining teams, five clearly chose not to rest their starters , all of which had good outcomes, and one mostly chose not to rest (see below for details), and it also had a good outcome.
Here’s the breakdown of teams:
Rested with good outcomes: (the 07 Packers (Conference Championship), 06 Bears (Super Bowl), 06 Saints (Conference), 05 Seahawks (Super Bowl), 04 Steelers (Conference), 04 Eagles (Super Bowl), and 04 Saints (Super Bowl).
Rested with bad outcomes: 05 Bengals, 04 Colts, 07 Colts, 08 Titans, 08 Giants, 05 Colts, 05 Patriots, 04 Chargers)
Did not rest with good outcomes: (07 Patriots (Super Bowl), 06 Patriots (Conference), 08 Steelers (Super Bowl), 04 Patriots (Super Bowl), 08 Cards (Super Bowl))
Probably did not rest with good outcomes(05 Broncos). What did probably did not rest mean? The Broncos sort of rested players: they beat the 9-7 Chargers 23-7, holding them out of the playoffs, but their backup, Bradley Van Pelt came in in the 3rd quarter and threw 8 passes for 7 yards). On the other hand, their starting QB, Jake Plummer, passed 14 times, and the box score indicates that they played Rod Smith, Ron Dayne, Tatum Bell, and John Lynch; each of whom (as I recall) were important players on that team.
So what conclusion can we draw from these data: most obviously, teams should stop resting their starters. I’m not going to go so far as to say that past results prove that teams that play hard in week 17 will do better in the playoffs than teams that don’t: there’re definitely not enough data points to draw that definitive of a conclusion. I do think there’s enough info to say that there’s definitely no clear advantage to resting starters. If there were, one would think that the record since 2003 wouldn’t have gone so clearly in the other direction, and the evidence (as far as it exists) is fairly clear: when the Colts made it to either the Conference Champonship (in 04) or the Super Bowl, it was in seasons where they couldn’t rest their starters (because they needed to win to preserve seeding. Whenever they did rest, they did poorly. When the Patriots rested, they lost in the divisional round, when they didn’t they usually made it to the Super Bowl (even managing to beat the juggernaut 06 San Diego in a year after the Patriots lost Rodney Harrison in their last, meaningless, week 17 game). Even in the Steelers case, the years they won the Super Bowl were ones where either they couldn’t rest players (05) or they chose not to (08). They’re coded with a good result in their rest year of 04, but they barely beat a mediocre Jets team in overtime and then got embarrassed by the Patriots in the conference game.
If resting players doesn’t create a significant advantage, it clearly shouldn’t be done for reasons almost too obvious to state. Not trying in the last game deprives paying fans of watching a good game; it unfairly changes the results for teams still trying to make the playoffs; and most nebulously but also most importantly, it sacrifices a chance to play a real game of football for the phantom advantage of potential victory in the future.