Re-Calibrating the Metrics
When Nate Silver of the then New York Times affiliated FiveThirtyEight correctly picked the winner in all fifty states in the 2012 Presidential election, he did so by aggregating polls over time and weighting them based on past performance. Two key tenants emerged from that success. Don’t place too much emphasis on newer polling data if it appears to be radically different from past polling data, and use multiple sources, taking into consideration their biases, to draw your conclusions.
Prognosticating on football, or any other sport for that matter, often tends to ignore both of these tenants. A big win by an underdog or a bad loss by a favorite in the last game often overshadows weeks of consistently bad or consistently good performance. And, we can become too attached to one source or one type of data and ignore others.
Over the past eight weeks, I’ve been using the advanced metric of Defense-adjusted Value Over Average (DVOA) to predict the outcomes of NFL games against the spread. What I’ve found is that DVOA appears to be a pretty good indicator of the winner of any given game, but not necessarily a good indicator of the final score. Over the last three weeks (since I’ve been tracking it closely), the more efficient team as measured by DVOA is 29-16 straight up, but 20-25 against the spread.
This week and going forward, I’ve decided to consider and weight a few more metrics along with DVOA in an attempt to bring up the accuracy against the spread. The Elo rating system, a measurement named after its inventor Armad Elo and used in a regular column on the new ESPN sponsored FiveThirtyEight, will be one component, while absolute point differential, and QBR (quarterback rating) will be the other two. A tie breaker, in some cases, will be home field advantage, as less efficient home teams made up four of the six instances where DVOA missed the winners last week.
The rationale behind using Elo is that it has also been about 65% accurate in picking winners (about the same as DVOA) and slightly better than DVOA in picking winners against the spread (even .500). The point differential should give some indication of how much better any given team has been than its opponents (ironically an element removed from the BCS formula early on to discourage running up the score), and, in a pass first league, the QBR should give an indication of which team has the advantage at the most important position in the NFL.
This Week’s Picks
For the column, as in the past five or six weeks, I’ll be concentrating on the NFC West and the Sunday night and Monday night games. If you’d like to see all of my picks, or pick against me, you can find me at ESPN’s Pigskin Pick’em in the Fans of Seattle Seahawks (spread) group under Brian Hight’s OSN Picks.
Also I would encourage interested readers to check out Football Outsiders for a more in-depth explanation of DVOA and FiveThirtyEight for information on the Elo rating system.
Here we go.
Arizona (6-1) at Dallas (6-2) -4
The X factor in this game is, of course, Tony Romo. Forced out of the Washington game with a back injury (a game I correctly picked, albeit to cover, not to win), Romo looked terrible when he returned in the fourth quarter to not save the day for the Dallas Cowboys. Owner, GM, and orthopedic surgeon, Jerry Jones assured the media this week that Romo would play against the Arizona Cardinals if he could tolerate the pain. Otherwise, Cowboys fans will see Brandon Weeden at QB. This would be a major downgrade, but so would a less than effective Tony Romo.
Dallas ranks 11th in the league in overall DVOA at 8.5%, while the Arizona Cardinals rank 18th with -1.0% for a still pretty close absolute differential of 7.5%. The Cowboys defense has regressed somewhat, now ranking 19th in efficiency, but should still match up well against a 25th ranked Cardinals offense. The real intrigue, however, comes with the matchup of the Cowboys 6th ranked offense versus the Cardinals 7th ranked defense.
The Cowboys have outscored their opponents by 46 in eight games, while the Cardinals have outscored their opponents by 25 in seven.
Dallas has the QBR advantage if Romo plays (77.8 to Carson Palmer’s 68.2), but that advantage disappears if Weeden has to play or Romo is less than optimum.
Elo has the Cowboys winning this matchup 56% of the time, but places the point spread at just -1.5.
Given all the question surrounding Romo’s health (and he did not participate in walk throughs or practice on Thursday), I’m taking the Arizona Cardinals.
Pick – Arizona Cardinals
St. Louis (2-5) at San Francisco (4-3) -10
I got burned badly last week by the St. Louis Rams. Coming off the upset of the Seattle Seahawks, I took them to cover 11 at Kansas City. I was only off by 16, as the Rams lost 34-7.
The Rams rank 30th in overall DVOA with -24.6%. Ranked 24th on offense, 31st on defense, and 24th in special teams, the Seahawks game looks like a total fluke.
Coming off a bye week for this game, it’s worth noting that the San Francisco 49ers beat the Rams just three weeks ago in St. Louis 31-17. But still the 49ers have been mediocre at best this season, ranked 20th in overall efficiency with -3.6% and a -7 point differential. This second meeting with the Rams is definitely an opportunity to improve.
Elo has the 49ers winning 81% of the time and matches Vegas by putting the line at -10.
Normally, my rule of thumb is to take the points when it’s double digits, but in this case, with the Rams being so bad, I’m reluctantly taking the 49ers to win and cover.
Pick – San Francisco 49ers
Oakland (0-7) at Seattle (4-3) -15
In beating the Carolina Panthers on the road last week, the Seattle Seahawks did not look particularly sharp, but their overall DVOA remains high at 3rd in the league with 18.8%, as it is adjusted for schedule. The Oakland Raiders rank 29th at -19.4% and stand 38.2% behind the Seahawks.
Seattle ranks substantially better than Oakland on both sides of the ball, 10th on offense to the Raiders’ 25th on defense, and 6th on defense against the 27th ranked offense from the bay area.
The Seahawks are +25 in points through 7 games, while the Raiders are -76 in losing all 7 of their contests. But, the Raiders have lost by more than this line of 15 only twice this season and the Seahawks greatest margin of victory came back in week one, 36-16 against the Green Bay Packers.
While not great, Russell Wilson carries a 63.0 QBR into Sunday. Rookie Derek Carr, with fewer weapons than Wilson, sports a 48.8 QBR.
Finally, Elo has the Seahawks winning the game 92% of the time and sets the line at -17.5.
All the numbers point to a Seahawks victory, but I do not like the 15 point line at all. I can see Seattle winning by more than that, but I can also see them winning by a lot less. That said, the Seahawks have the advantage in every metric I’m applying here, so I’m going to stick with the numbers and take Seattle at home to rout the Raiders.
Pick – Seattle Seahawks
Baltimore (5-3) at Pittsburgh (5-3) EVEN
The line on the Baltimore Ravens at the Pittsburgh Steelers is anywhere between even and -1.5 for the Steelers. I’m going with even money because the majority of sites ESPN uses have it that way, but if you like the Ravens, you can find points.
The AFC North may be the best division in football, with the Baltimore Ravens ranked 2nd in overall DVOA at 26.0%, the Pittsburgh Steelers ranked 8th with 10.0%, and the Cincinnati Bengals ranked 13th at 6.7%.
The Ravens narrowly lost in Cincinnati last week and now must face a Steelers team that dismantled the Indianapolis Colts. Baltimore is a well-rounded team, ranked 7th on offense, 5th on defense, and 6th on special teams.
Pittsburgh, despite a 5-3 record, is not very good on defense, usually a hallmark of Steelers teams, ranking 26th in defensive efficiency. This Steelers team is driven by offense, ranking 4th in efficiency and led by Ben Roethlisberger and his 77.7 QBR.
The Ravens have the advantage in overall efficiency and point differential (+86 to +9), while the Steelers have the advantage at QB (77.7 QBR to Joe Flacco’s 67.6) and home field, which accounted for 4 or the “upsets” last week when a more efficient team lost.
The Elo ranking system has the Steelers winning 55% of the time, with a line of -1.5.
With 3 out of 5 indicators going to the Steelers, I’m taking the Steelers.
Pick – Pittsburgh Steelers
Indianapolis (5-3) -3 at New York Giants (3-4)
These matchups between Olivia and Archie Manning’s sons are always such a treat. Oh, wait a minute. Never mind.
The Indianapolis Colts got destroyed by the Pittsburgh Steelers last week, 51-34, in a game I also lost. But, while the Colts’ defense (that I started in one fantasy league) took a hit, dropping to 21st in efficiency, Andrew Luck and the offense just don’t quit and come in this week as the 8th ranked offense in efficiency.
I have to admit I was surprised to see that Eli Manning actually has a slightly higher QBR than Andrew Luck (78.5 to 77.8), but it sure hasn’t translated into points, as the New York Giants have a -15 point differential compared to the Colts with +63.
With DVOA and point differential favoring the Colts, QBR essentially a wash, but home field to the Giants, Elo plays a big role in my pick. Elo has the Colts winning this game 54% of the time but places the spread at just -1.5. In what most would consider an upset, I’m taking the Giants at home on Monday night.
Pick – New York Giants
Good Luck
As always, these picks are not meant to be used as a substitute for a good retirement plan. That’s what the $284 million Mega Million drawing tonight is for.
Have fun, good luck, and enjoy the games.
Add The Sports Daily to your Google News Feed!
