IND9.810.1HOU98.22.423.61515 Although the Packers hadn’t replaced a coach at midseason since 1953, Sunday’s loss forced their hand. Now they’ll need to figure out who’s next, from a candidate list that includes big names among both pro coordinators (Josh McDaniels) and up-and-coming college coaches (Lincoln Riley). They’ll also need to hope Rodgers’s issues were more related to McCarthy’s offense and less to his getting older and less productive — basically, that the next Packer coach will be more Mike Shanahan to Rodgers’s John Elway than Jimmy Johnson to his Dan Marino. So while the Packers may not have much on the line over the rest of their games, this promises to be the most interesting offseason Green Bay has had since Favre was retiring and unretiring more than a decade ago.FiveThirtyEight vs. the readersMake sure to check out FiveThirtyEight’s Elo ratings using our NFL prediction interactive, which simulates the rest of the season 100,000 times and tracks how often each team should make the playoffs and win the Super Bowl. And did you know you can also pick against the Elo algorithm in our prediction game? Maybe you can also climb up our giant leaderboard (or, if you’re like me, fall down it with each passing week).Here are the games in which Elo made its best — and worst — predictions against the reader picks last week: Rodgers hasn’t won as much as he should haveTop 10 NFL starting quarterbacks by Yards Above Backup QB, 1990-2017, with their actual and expected Super Bowl appearances PIT62PIT54LAC 33, PIT 30+6.7– 1Peyton Manning1721,58544.15-0.15 PHI28.324.2DAL81.119.451.01578 3Drew Brees1617,25012.89-1.89 KC83KC89KC 40, OAK 33+0.4– Home teams are in bold.The scoring system is nonlinear, so readers’ average points don’t necessarily match the number of points that would be given to the average reader prediction. 6Ben Roethlisberger1410,94531.35+1.65 CAR17.89.9CLE0.30.422.01454 8Steve Young810,02211.65-0.65 4Brett Favre1913,04721.86+0.14 Over time, it became more and more difficult for the Packers to come within striking distance of the Super Bowl. In 2015, Rodgers slumped to career-worst numbers without top wideout Jordy Nelson, though the team as a whole was still good enough to get to the divisional playoffs before losing. In 2016, it was more of the same when Rodgers mused that Green Bay could still “run the table” — sparking an eight-game winning streak that saw the QB return to vintage form and left the Packers a win away from the Super Bowl.4Though their defense was shredded by Matt Ryan and the Falcons for 493 yards, ending the streak. By then Rodgers was 34 years old, so a sense of urgency was setting in when 2017 came and went without a playoff berth — even though that could be written off as the byproduct of Rodgers missing nine starts.The 2018 season was always going to be the real crossroads for McCarthy. With a healthy Rodgers leading the way, the Pack could always count on contending in the past, so this year’s expectations were no different. But Rodgers’s numbers have been merely good, not great. Brett Hundley isn’t around anymore to take any blame. And unlike in 2015, when Green Bay was talented enough to survive despite a downturn in its QB’s individual stats, there has been no answer from the team’s supporting cast this time around. It all came crashing down around McCarthy in the loss to Arizona as 13½-point favorites, Green Bay’s single most disappointing defeat since the merger according to Pro-Football-Reference’s point-spread data.We can visualize the Packers’ decline over McCarthy’s final few years at the helm using FiveThirtyEight’s Elo ratings. Specifically, I’ve been tinkering with an experimental version of Elo that keeps a separate adjustment for the primary QB in each game, similar to how we treat starting pitchers in our MLB ratings.5For those curious, it uses what I’m calling a QB’s Performance Value Added (PVA). That metric starts with an estimate of a primary QB’s value over replacement (according to ESPN’s expected-points data) in a game:PVA = -2.2*Attempts + 3.7*Completions + (Pass Yds / 5) + 11.3*Pass TDs – 14.1*Interceptions – 8*Sacks – 1.1*Rushes + 0.6*Rush Yds + 15.9*Rush TDsIt then compares that number to the per-game average allowed by the opposing defense in all other games that season (excluding the game in question). By definition, an average PVA is 0.0. Then a running mean of PVA is kept for both teams and individual QBs, which is used to modify the team’s base rating depending on which QB is used. (Interestingly, the predictions work best when the individual ratings update slightly faster than the team ones, which suggest QB “hot streaks” do contain extra information.) Debut QBs are assigned an expected PVA of -47, and they reduce a team’s effective Elo rating by 108 points relative to an average QB. Using this, we can trace how a team’s performance rises and falls independent of its QB — which is useful in cases like 2017, when Rodgers was hurt and Hundley started nine games. (For instance, by season’s end, the Packers would have projected to be a 1529 Elo team with Rodgers starting — compared to a 1427 team with Hundley. And remember, 1500 is average.) HOU77HOU69HOU 29, CLE 13-6.6– LAC96.03.6CIN0.91.39.11504 Game quality is the harmonic mean of the Elo ratings for the two teams in a given matchup. Total Change adds up the potential swing in playoff odds for every team in the league (not just the two teams listed).*Average change is weighted by the likelihood of a win or loss. (Ties are excluded.)Source: ESPN.com Elo’s dumbest (and smartest) picks of Week 13Average difference between points won by readers and by Elo in Week 13 matchups in FiveThirtyEight’s NFL prediction game Playoff %Playoff % The dismissal of Green Bay coach Mike McCarthy — who was let go after the Packers’ stunning home loss to the Arizona Cardinals on Sunday — wasn’t exactly a shock. Perennially tabbed as a Super Bowl contender out of the NFC, McCarthy’s team had gone just 11-16-1 over the past two seasons, including a disappointing 8-9-1 in games that featured future Hall of Fame quarterback Aaron Rodgers as Green Bay’s primary passer.1Meaning he led the team in attempts for the game. It was time for a change along the sidelines that Vince Lombardi once roamed.Things weren’t always so bleak on the frozen tundra. The McCarthy era had its high points, particularly early on — when he and Rodgers appeared to have Green Bay positioned on the cusp of a potential dynasty. But between postseason near-misses, roster changes, injuries and coaching miscues, McCarthy’s Packers never fulfilled that promise. Instead, it’s fair to wonder whether Green Bay squandered the prime of one of the most talented QBs in NFL history.The Packers team that McCarthy inherited in 2006 from Mike Sherman2Because every Packer coach must be named “Mike”. was one in transition — and that meant navigating some heavy-duty Brett Favre melodrama in his first two seasons at Green Bay’s helm. However, McCarthy quickly found that he had an all-time great on his hands in Rodgers, who, when he took over the starting job at age 25, was just entering his best years as a passer. The McCarthy-Rodgers marriage sputtered to a 6-10 finish in its first season but yielded great results shortly thereafter: an 11-5 playoff campaign in Year 2, then a Super Bowl crown in Year 3 and a 15-1 regular season (with Rodgers winning MVP) in Year 4. The sky seemed to be the limit for McCarthy and his star QB.Since the end of the 2011 regular season, however, the Packers have gone just 5-6 in the playoffs; by comparison, Tom Brady and the postseason Patriots are 13-5 over the same span. Green Bay’s record includes a crushing home defeat against the New York Giants two weeks after that 15-1 season ended and another loss in which they watched helplessly as ex-49er Colin Kaepernick destroyed their defense in 2012 — still one of the greatest individual QB games in playoff history. The Packers’ postseason circumstances have not always been ideal: For instance, that Giants game was actually the only time since 2011 that Green Bay lost in the playoffs while favored — meaning the rest of the losses were as underdogs. But at the same time, the Pack have also had clear chances to return to the Super Bowl, and they came up short in each of them.All told, it remains mystifying that a quarterback of Rodgers’s stature hasn’t won more frequently. If we run a simple logit regression between a QB’s Yards Above Backup in a season and whether his team made the Super Bowl,3Minimum six starts during the season, using data since 1990 (the era of the current playoff setup). we’d expect Rodgers to have made 1.86 Super Bowls in his career through 2017 — roughly twice as many as he’s actually been to. (Meanwhile, other contemporary QBs — such as Ben Roethlisberger, Eli Manning and even Brady — have gone to more than twice as many Super Bowls as we’d expect from their individual stats.) BAL65.0%+/-15.1KC100.0%+/-0.031.81628 TB1.72.0NO100.00.06.91570 9Matt Ryan108,25111.14-0.14 DEN20.911.1SF0.00.024.71427 LAR68LAR79LAR 30, DET 16+4.1– CHI69CHI72NYG 30, CHI 27-6.1– TEN78TEN72TEN 26, NYJ 22-4.4– TEN18.910.6JAX0.10.121.91478 GB73GB79ARI 20, GB 17-10.5– PIT83.811.0OAK0.00.023.71453 OUR PREDICTION (ELO)READERS’ PREDICTION 1990 was the first season of the NFL’s current playoff format. Expected Super Bowls are based on a season-by-season logit regression between a QB’s Yards Above Backup and whether his team made the Super Bowl.Source: Pro-Football-Reference.com MIN59.719.9SEA87.711.841.41572 GB1.41.3ATL220.127.116.11469 7Philip Rivers1210,72101.54-1.54 MIA6.67.4NE99.70.516.21537 Super Bowls Made NE67NE65NE 24, MIN 10-3.8– 10Tony Romo98,19201.11-1.11 IND51IND62JAX 6, IND 0-14.0– NO64NO74DAL 13, NO 10-15.3– CIN52%DEN59%DEN 24, CIN 10+8.5– CHI94.44.3LAR100.00.010.81615 BUF0.00.0NYJ0.00.02.41377 PHI69PHI66PHI 28, WSH 13-4.1– After a series of narrow wins at midseason, the algorithm handed the readers their worst loss (-55.2 points on average) since Week 3. Some of the blame can go to the subject of this column — the Green Bay Packers, whose loss not only cost Mike McCarthy his job but also cost users 10.5 points on average. But readers were also burned by the Jaguars’ win over the Colts and the Cowboys’ upset victory over the Saints. Add it up, and Elo has beaten the average reader 12 times in 13 weeks this season.But congrats to Mike Edelstein, who led all users in Week 13 with 137.0 points, and to one of my favorite leaderboard names, Greg Chili Van Hollebeke, who maintained his No. 1 ranking on the season with 1,002.1 points. Thanks to everyone who has been playing — and if you haven’t, be sure to get in on the action! You can make picks now and still try your luck against Elo, even if you haven’t played yet.Check out our latest NFL predictions. 5Aaron Rodgers1010,98811.86-0.86 CAR59CAR64TB 24, CAR 17-7.8– WSH26.411.7NYG0.10.124.91435 The best matchups of Week 14Week 14 games by the highest average Elo rating (using the harmonic mean) plus the total potential swing for all NFL teams’ playoff chances based on the result, according to FiveThirtyEight’s NFL predictions ATL53ATL50BAL 26, ATL 16+1.5– Team ACurrentAvg. Chg*Team BCurrentAvg. Chg*Total ChangeGame Quality PICKWIN PROB.PICKWIN PROB.ResultREADERS’ NET PTS DET0.30.3ARI0.00.04.21412 2Tom Brady1619,73583.63+4.37 SEA83SEA83SEA 43, SF 16-1.5– MIA58MIA57MIA 21, BUF 17-2.3– QuarterbackYears StartingYARDS Above BackupActualExpectedDiff. At the beginning of 2015, the Packers had an effective Elo of 1622, which included a 73-point boost from having Rodgers at QB and a 49-point boost from his teammates. By the end of the year, Green Bay’s effective Elo was still in the same neighborhood (1597), despite Rodgers’s adjustment actually dropping to negative 11, because the rest of the team carried a larger share of the weight (+108). Meanwhile, at the peak of the Packers’ run-the-table surge in 2016, the team’s 1657 effective Elo arose out of a 61-point boost from Rodgers and 97 additional points (relative to league average) from the rest of the team.But fast-forward to now, and it’s clear how much the Packers have crumbled around Rodgers. His own adjustment is 16 points of Elo above an average QB, the lowest it’s been since Week 10 of the 2016 season. But he’s still expected to be above average; his supporting cast, by contrast, has fallen to a negative-67 score relative to the average team. That’s the worst they have been in Rodgers’s entire NFL career, and it isn’t especially close. Keeping QB play constant, the Packers’ Elo has dropped by a total of 139 points since the end of the 2016 season, which is essentially the difference in current Elo ratings of the 11-1 Los Angeles Rams and the 6-6 Carolina Panthers.The reasons for the slide are varied, but many can be traced back to a series of poor drafts under former general manager Ted Thompson, who was replaced by current front-office chief Brian Gutekunst in January. As Sports Illustrated’s Kalyn Kahler pointed out last week, only three of Green Bay’s 17 draftees from 2014 and 2015 remain on the current roster. While no team can avoid dry spells in the NFL draft if given enough time, the Packers also — largely by design — did little in the way of enlisting outside help as a backup plan.6In addition to the lack of starters from recent drafts, only two of the team’s current starters were drafted by a team other than Green Bay. Again, this was the byproduct of the Packers’ religious devotion to drafting and developing their own prospects, but that plan only really works when you draft well. Combine those infrastructural problems with criticisms of McCarthy’s offensive scheme (criticisms of a perceived lack of innovation that Rodgers apparently shared), plus legitimate complaints about Rodgers’s own decline in performance, and a season like this was bound to happen to Green Bay sooner or later.Even so, it came contrary to preseason predictions. Going into the schedule, you might have penciled in this week’s matchup against the Atlanta Falcons as a marquee game with playoff implications. Instead, it will be the third-worst game of the week, according to our combination of matchup quality (i.e., the harmonic mean of the teams’ Elo ratings in each game) and game importance (how likely it is to swing every team’s odds of making the playoffs):7A little note is in order here, to explain how these game importance ratings are now calculated. All season long, I have been crunching the numbers on how a given game affects the two teams involved. But this week, with the playoffs on the horizon and certain matchups taking on more importance because of their outside implications, I changed the formula to add up a game’s potential swing in playoff odds for every team in the league — including those not participating in the game itself.