We’re a quarter of the way through the NFL season, and Week 4 brought its fair share of surprises. The Buffalo Bills upset the Atlanta Falcons on the road, a last-second Carolina Panthers field goal delivered Tom Brady’s New England Patriots their second home loss of the year, and Andy Dalton was actually good.

Through four weeks, very few teams are following the course predicted for them at the season’s outset. Many pundits thought a Los Angeles team would be sneaky dangerous this season — they just didn’t think it would be the Rams. Others assumed a New York team would start 0-4 — they just didn’t pick the Giants. But in a season marked by surprises, the most perplexing start may belong to the Dallas Cowboys.

After Week 1, Dallas had every reason for optimism: The expected suspension of running back Ezekiel Elliott was delayed by the courts, and the team that was 13-3 last year was fresh off a dominating win over the New York Giants. Dallas’ Elo rating climbed, and many analysts (most of them in Texas) viewed the Cowboys as a surefire Super Bowl team. As it turned out, the Giants are terrible and Dallas would spend the next three weeks showing that it was not the same team it was last year — even with Elliott in uniform.

“America’s Team” had an Elo advantage of 160 points heading into Sunday’s game with the Los Angeles Rams, meaning they were favorites to win the game by about six points.Every 25 Elo points is equivalent to 1 point in an actual game.

“>¹ The Cowboys lost and fell to 2-2 and, in the process, saw their chances of making the playoffs drop to 40 percent — their lowest of the season — according to our 2017 NFL Predictions. For Dallas — which dropped to a tie for eighthWith the Seattle Seahawks

“>² in our power rankings — a 4-in-10 chance of making the playoffs with three-quarters of the season left to play may not sound all that bad. But compared with the rest of the teams in the top 10, the Cowboys are noticeably out of place.

Most of Dallas’s metrics are roughly the same as they were last season, with the major exception of the passing game (and some faulty special teams). Last year, in QB Dak Prescott’s first NFL season, he led the Cowboys to the sixth-most expected points added (EPA) through the air of any offense in football — a highly effective complement to Dallas’s second-ranked running game. But although the Cowboys still rank among the top five in rushing EPA, Prescott and the passing offense has fallen to 19th in the league in EPA. Prescott’s traditional numbers aren’t bad, but he’s all but stopped throwing the deep ball and the Dallas receivers are picking up very few yards after the catch. The result has been an aerial attack that struggles to move the chains, keeping the Cowboys from scoring as many points as they should.

The bad news doesn’t stop there for Dallas. The Cowboys face a tough road if they hope to win the division, because even though the Giants are floundering, the rest of the NFC East has played better than we predicted in the preseason. And based on FiveThirtyEight’s Week 5 NFL Elo ratings, Dallas has the toughest remaining schedule of any team in its division; the Cowboys’ opponents have an average Elo rating of 1525, while the teams New York is facing average 1514, Washington’s opponents average 1499, and the rest of Philadelphia’s schedule averages 1496.

The Cowboys and Eagles’ next five games could further increase the gap between those two teams — the Eagles currently have a one-game lead. Dallas’ next five opponents have an average Elo of 1549, and the Cowboys will face three of those teams on the road. Compare that with Philadelphia’s next five games — in which the Eagles’ opponents average an Elo of 1490 and only one game will be on the road — and the Cowboys’ playoff hopes could be essentially done by Week 10.

FiveThirtyEight vs. the crowd

Weeks 3 and 4 in our NFL prediction game — in which we invite you to pick football games and try to outsmart our Elo algorithm — were triumphant victories for machine over man. Over a combined 32 games, our Elo model beat you, the readers, by a whopping 146.2 points! You can blame the New York Jets for your two largest net losses. The Jets upset the Miami Dolphins in Week 3 — which both our model and the readers got wrong — and this weekend the Jets pulled out an overtime victory over the Jacksonville Jaguars, which our Elo picked right and you readers got wrong. (Note: We don’t blame you for not picking the Jets.)

Despite Elo correctly predicting the winners in only two more games than readers through Week 4, our model holds a lead of 181.7 net points. This is because our game rewards not just correct picks, but also how much confidence the model and our readers had in the outcome. For example, despite our bashing the Cowboys here, both Elo and the readers predicted Dallas would win at home over the Rams this past weekend. Though we both predicted incorrectly, the readers netted 5.5 points on that game because Elo had more confidence in Dallas winning (78 percent) than readers did (72 percent).