Was Benching Bradford The Right Move?

LBVikings   |  

Sep 19, 2017

• Share on Facebook
• Share on Twitter
• Share on Reddit
• Share on WhatsApp
• Share via Email
• Copy Link

  Link Copied

It’s blasphemy in the NFL to suggest the sacrifice of a win for the greater good. Whether it’s tanking for a draft pick or resting your starters for the playoffs, it’s bound to be a controversial topic. Benching a player on purpose in week 2 should be a non-starter, but there’s an argument to be made when a fragile knee is involved. After an MRI, a lot of panic from Vikings nation, and a pre-game work out, the Vikings pulled the plug on Sam Bradford, and effectively their chances at a 2-0 start. Sunday morning, the Vikings grappled with an odd decision point: is it worth risking future games to put your best foot forward in the current one?

The results of the decision were obvious – throws that Bradford could hit with ease were overthrown or otherwise inaccurate. Case Keenum only generated 167 yards of passing offense, a passer rating of 65.9 and a pitiful QBR of 24.7. Ostensibly, the decision to bench Bradford was a decision to torpedo the week 2 game and live to fight another day. Even if you believe the Vikings could win with Keenum (and you’d have an argument), it’s a worthy thought experiment to consider such an ultimatum. The Vikings had to weigh the probability of Sam Bradford aggravating his knee, the effect of that, and the value of this particular game. It’s that last one where things get interesting.

I’m about to go deep into the math weeds here, so put on your nerd glasses and strap in. How much is a game worth before it’s even played? We can leverage win probability into expected wins by simply adding the percentages up. If a team is 50/50 to win a particular game, that game is worth half an expected win. If they’re 75% favored to win, that’s worth three quarters of a win. Add those two together, and those two games are worth 1.25 “expected” wins. Add up an entire season this way, and you can get a reasonable idea of how well a team played over a season, and how much each game meant against expectation.

I calculated each team’s 2016 season to establish a baseline for this exercise. There are a lot of options to use for win probability, but I decided to use the Vegas betting odds, since those theoretically reflect community opinion. Apply some crude math to convert Vegas lines into a scale from 0%-100%, and a little baby win probability model is born! Here’s how each team stacked up in 2016:

So by this model, Arizona was “expected” to win 9 or 10 games, while a team like the Jets was only expected to win 6 or 7. This model requires a huge margin of error to represent reality, so it’s not quite the best measure of quality. But it’s a decent enough approximation to measure the value of single games against a season.

For example, the Vikings’ season shook out like so:

So the Halloween Bears game was worth 61% (ish) of an expected win, as well as the ensuing Lions game. Losing both had a much bigger impact (1.225 expected wins) than, say, losing in Lambeau or against the Cowboys. Put another way, the Vikings already had 39% of a loss chalked up, and adding the other 61% hurts more. Again, this isn’t a very precise way to evaluate teams unless you’re using a better win probability model, but it can illustrate the impact upsets have on your season.

Applying this to the Steelers game, the Vikings were 5.5 point underdogs before the Sam Bradford news hit. That fell pretty quickly as the week rolled on, but we’ll roll with that as the “if Sam plays” number. That’s worth about 36% of a win by this formula- pretty small. Against a season’s aggregate total of 7-9 expected wins, a third of one win seems fairly insignificant. That’s not to say it’s irrelevant, however; these games add up. But a loss to the Steelers will hurt half as much as, say, a loss to the ailing Bears.

To properly evaluate the decision, we have to look at the other side of the risk. If Bradford played, got hit, and tore his fragile knee up, that likely ends the season. For simplicity’s sake, we’ll say that costs us four games we would have been able to win with Bradford (again, we don’t need precision here). Obviously, that’s a heavily impactful outcome. But it’s not 100% guaranteed. If there were a 50/50 shot of a Bradford injury, that’d be worth two expected wins, and so on.

So to weigh this week’s loss of 0.36 expected wins, you’d need to believe that there’s less than a 1-in-12 chance of Bradford sustaining a more severe injury. Unrealistic. Perhaps you think a Bradford injury loses us fewer than four games we would have otherwise won. Even if you think Keenum wins only a game or two fewer than Bradford, you’d still have to believe he was invincible to think it’s worth the risk. It only becomes viable when you have Keenum in for long enough to lose the same amount of games a Bradford injury costs you. But for Sunday, the value just isn’t there to hang the entire season in the balance of a game the Vikings were probably losing anyways.

Thanks for reading!