The Patriots Steal Back The Top Ranking

So much for settling in at the top. A week ago in this space, I noted that the Kansas City Chiefs had vaulted to No. 1 in our NFL Elo ratings, leapfrogging the previously top-rated New England Patriots with a win over the Pats in Week 1. Well, K.C.’s reign in first place turned out to be brief: The Pats retook the top slot in Week 2, bumping the Chiefs down to second.

I know what you’re thinking: Kansas City won on Sunday, so what gives? You can dig deep into Elo’s methodology here, but the basic premise is that it assigns each team a power rating that can be used to predict the outcome of any game. Once that game is in the books, Elo takes rating points away from the loser and gives them to the winner, in proportion with how unlikely the victory was (upsets shift the ratings more than routine wins) and the winner’s edge on the scoreboard (big wins are worth more, although there are diminishing returns to running up the score).

In the case of the Chiefs and Patriots, K.C. was a heavier favorite to win Sunday — 73 percent at home versus Philadelphia, as opposed to New England’s 64 percent chance on the road against New Orleans — but ended up winning by fewer points — only 7, as opposed to the Pats’ 16-point margin. So even though the win boosted Kansas City’s Elo rating by 10 points, New England gained 19, enough to erase the Chiefs’ slight Elo edge going into the weekend. (The margin between the teams is still extremely small.)

Early in an NFL season, it’s rare to see two teams pass the Elo baton of No. 1 back and forth like this. Before 2017, the last time a preseason No. 1 was overtaken in Week 1 and then reclaimed the top slot in Week 2 was in 1993, when the Dallas Cowboys and San Francisco 49ers fought over the top ranking. It was a harbinger of things to come that season: The Cowboys and Niners faced off in the NFC championship game, and Dallas went on to win its second-straight Super Bowl.

There was a similar situation in the previous season: Washington, the reigning Super Bowl champions, and San Francisco scrambled for No. 1 in the season’s first month. Washington would stumble that year to a 9-7 record but would eventually meet the Niners in the divisional round and lose.

But that kind of duel doesn’t usually last much beyond September. In all but one such case since 1970, one of the two teams quickly seized the baton and ran away with it, dominating the top ranking for most of the rest of the season. The exception was in 1974, when John Madden’s Oakland Raiders and Chuck Noll’s Pittsburgh Steelers swooped in and stole No. 1 from both Miami and Minnesota. Either Oakland or Pittsburgh ended up holding the top slot for all but one of the season’s final 17 weeks. (“The Steel Curtain” would defeat the Raiders in the AFC championship game, win its first Super Bowl and dominate the rest of the decade.)

Given its dynastic pedigree and Tom Brady’s rebound performance against New Orleans, New England might seem the likely candidate to tighten its grip on No. 1 and hang onto it the rest of the way. But take heart, K.C. fans: Even though the Chiefs are No. 2 in the ratings, Elo gives the Chiefs a better chance than the Patriots of making the playoffs (83 percent to 76 percent) and winning the Super Bowl (14 percent to 12 percent).

FiveThirtyEight vs. The Crowd

In Week 2 of our NFL prediction game — in which we invite you to pick football games and try to outsmart our Elo algorithm — FiveThirtyEight’s readers fared slightly better than the computer model, scoring some big wins. One instance in which our readers trounced the model was the Tennessee-Jacksonville game, where the average player picked the Titans to win on the road with 52 percent confidence. The model, which failed to factor in the Jaguars’ inherent Jaguar-ness, had the home team winning at 59 percent. The Jags lost, and it wasn’t close.

Readers were also more likely to fade the sad excuse for a football team known as the New York Jets. The Elo model gave the Oakland Raiders a 74 percent chance of winning at home, while readers had 88 percent confidence in Oakland — rightly knowing that the Jets (who systematically shed most of their talent this offseason) stood little chance in the Raiders’ home opener.

Elo had its wins, too. The model was more confident than readers were that the reigning NFC champion Atlanta Falcons would win at home against the Green Bay Packers; it gave the Falcons a 63 percent chance of winning, while the average reader saw the game as a coin flip (50 percent). Likewise, Elo predicted the Denver Broncos to stand their ground against the Dallas Cowboys despite their being a home underdog according to the Vegas betting line — while the average player (wrongly) backed Dak Prescott, Ezekiel Elliott and company.

Here are all the games from the second week of the season, in order of how many more points FiveThirtyEight readers earned, on average, than the model (or vice versa):

The Week 2 winner is …

Congratulations to Tristan Smith from Nova Scotia, Canada, who scored 282.2 points in Week 2. Tristan, a financial analyst by trade, correctly picked 15 of 16 winners, including picking the Cardinals, Raiders, Seahawks and Patriots at 100 percent confidence.

Remember: You can start playing the prediction game this week, even if you didn’t get your picks in Weeks 1-2.

Check out our latest NFL predictions.