FanPost

Week 7: Blind Power Rankings

Cary Edmondson-USA TODAY Sports

Yes, I used the other name last week. I think I'll use this one going forward, though, as it's cleaner and provides a probably-justified warning to take the rankings with a grain of salt. To clear up any name-related confusion and provide a singular source to link to in the future, I'll explain the BPR algorithm again here.

BPR: BPR Primer Revisited

The only data BPR uses is, for each team, a list of teams that team has beaten and a list of teams that team has lost to (which can in turn be created from a list of the winner and loser of each game that has been played). Ties are handled by considering each team to have a win against the other. The essential algorithm is as follows:

1. Create a list of each team and how many losses that team has.

2. Each team with the fewest losses is put into the first 'tier'.

3. For each team in the first 'tier', subtract losses from each team still in the list equal to the number of times those teams have played.

(Interpretation: If a team has lost to a better team, it can be essentially ignored, and if they've beaten a better team, it makes up for one of their losses.)

4. Remove each team in the first 'tier' from the list and add them to the ranking.

5. Repeat Steps 2-4 until all teams are ranked.

This, specifically, is the algorithm for what I've been calling Loss-Column BPR. For Win-Column BPR, create a list of wins instead, start with winless teams, and reverse the final ranking.

The results seem to indicate that it's best to look at LBPR and WBPR together, so the main ranking statistic I'm using is CBPR, which is simply the average of a team's ranking by LBPR and WBPR.

Week 7 Rankings


LBPR WBPR INDIFF CBPR EXDIFF
1. Broncos (2)
1 2 1 1.5 0.5
2. Cardinals (1)
3 1 -2 2 -3
3. Chiefs (16)
8 4 -4 6 -8
4. 49ers (6)
5 8 3 6.5 -5.5
5. Chargers (3)
10 3 -7 6.5 -0.5
6. Colts (10)
6 9 3 7.5 3.5
7. Eagles (7)
2 16 14 9 6
8. Cowboys (4)
4 15 11 9.5 7.5
9. Patriots (20)
16 5 -11 10.5 2.5
10. Ravens (8)
7 17 10 12 3
11. Seahawks (5)
12 13 1 12.5 1.5
12. Bills (17)
19 6 -13 12.5 -3.5
13. Giants (13)
9 19 10 14 -7
14. Packers (9)
18 10 -8 14 8
15. Dolphins (23)
21 7 -14 14 -4
16. Bengals (15)
17 11 -6 14 1
17. Lions (11)
13 18 5 15.5 5.5
18. Panthers (14)
20 12 -8 16 -1
19. Texans (18)
15 20 5 17.5 -5.5
20. Bears (12)
22 14 -8 18 -1
21. Rams (19)
11 25 14 18 -6
22. Washington (24)
14 28 14 21 -5
23. Steelers (25)
25 21 -4 23 8
24. Vikings (21)
24 23 -1 23.5 -3.5
25. Browns (22)
27 22 -5 24.5 4.5
26. Titans (29)
26 24 -2 25 -3
27. Jets (28)
23 31 8 27 -2
28. Falcons (27)
30 27 -3 28.5 3.5
29. Saints (26)
31 26 -5 28.5 6.5
30. Jaguars (32)
28 29 1 28.5 -2.5
31. Raiders (30)
29 32 3 30.5 -1.5
32. Buccaneers (31)
32 30 -2 31 1

Interpretations

Teams are ranked by this week's CBPR (average of LBPR and WBPR), with last week's rankings in parentheses. LBPR and WBPR are, essentially, measures of how impressive a team's loss column is and win column is, respectively. INDIFF is just the difference between the two - a high magnitude here means the two metrics are conflicted about where to place the team, while a low magnitude means the two metrics give similar results.

A positive INDIFF number means a team has relatively few strong wins or weak losses - they beat worse teams and lose to better teams. A negative number means a team appears inconsistent - they combine losses to worse teams with wins over better teams. EXDIFF is the difference between CBPR and the ESPN power rankings (as the representative 'conventional' power ranking). A negative number here indicates a team that BPR considers underrated, while a positive number indicates a team that BPR considers overrated.

Ranking Notes

Hail to the Chiefs

The Chiefs, previously bogged down by their awkward loss to Tennessee, made a great case for themselves by beating the Chargers, previously undefeated except by the Cardinals. But this alone didn't account for their rise from 16th to 3rd. They also got help from the Dolphins. By beating the Bears, the Dolphins got their first signature win against someone other than the Patriots. Now that both the Dolphins and Chiefs look significantly better, the Patriots are more easily forgiven for losing to both. And, in turn, now that the Patriots are ranked relatively highly, their loss to the Chiefs makes the Chiefs look better. These two upsets provide most of the background for any seemingly-random ranking changes this week - as all the rankings are connected, a team's ranking can be affected by many games other than its own. The Patriots were an obvious beneficiary this week, as they shot up the ranks despite only taking care of business by beating the Jets. The Bills similarly did nothing too impressive this week but were helped by their schedule now looking significantly tougher. Looking a bit farther out, the Chargers' loss gave the Broncos a slight edge up to the top spot, and the Chiefs' win made San Francisco's win against them look much more impressive. Consider: the 49ers have played five games against the top 8 teams. They are 3-2 in those games.

Faked Out

As is probably expected, the Seahawks dropped a few spots. And, of course, after their dramatic signature win, the Rams... dropped two spots. Now, this is slightly an unfair way of putting it, as the Rams actually have a better CBPR than last week (from 19 to 18). It just happens that there's a fat, ugly middle of the pack sitting right in front of them. As it happens, the win for the Rams mostly just counters their loss to Minnesota, meaning they still have some work to do before moving anywhere significantly up the ranks. Of course, as with the other upsets, this didn't only affect the two teams actually playing the game. In particular, look at the Cowboys and Packers. The Cowboys are looking quite impressive and got another win this week. But they dropped four spots. Why? Well, of course, people were discounting the Cowboys for playing a weak schedule until they got a signature win against... the Seahawks. Oops. So now the Cowboys are back to not having beaten any above-.500 teams. ESPN ranks them second. As for the Packers, their loss to the Seahawks now looks a bit worse and highlights the fact that they haven't actually beaten anyone better than Miami. ESPN ranks them sixth. This shift in turn causes another team to drop: with the Packers less of a power, the Lions have less of a signature win, and their loss to the Panthers is looking worse by the week.

Eliteness

For the first three weeks that I charted BPR, exactly five teams had a CBPR of less than 6, which makes some intuitive sense. This week, there are only two, followed by a pile of 19 teams separated by only 12 points of CBPR. Parity seems to be out in full force - most of the teams in the league haven't looked solidly good or solidly bad and it's difficult to tell how the many middle-of-the-road teams rank up against each other.

Consistency

I mentioned last week that the Bengals, Browns, and Steelers, all in the AFC North, were ranked by BPR as three of the league's five most inconsistent teams. While they're still seen as somewhat inconsistent, BPR seems mostly convinced that they just aren't very good (along with the Panthers and Saints, last week's two other most inconsistent teams) and has passed the torch of inconsistency to the AFC East, where the Patriots, Dolphins, and Bills now rank as the league's three least consistent teams. Will they meet the same fate as their forebears? On the other side of things, I mentioned last week that Houston, Washington, St. Louis, Kansas City, and Philadelphia were the five most ambiguous teams. Two of these teams went on to pull major upsets, and they were 4-1 overall. Not to say that this definitively means something, but it does make sense that these teams would be among the most likely to do unexpectedly well in a game. Of course, they could also be among the most likely to do unexpectedly poorly (which is arguably what happened to Houston this week). So for this week, keep an eye out for the unexpected involving Philadelphia, St. Louis, Washington, Baltimore, and the Giants.

Overrated/Underrated

A look at the five teams BPR considered most overrated last week: in short, the Browns, Saints, and Bengals certainly appeared overrated (ESPN now ranks the Bengals about where BPR does, moved the Browns closer to where BPR has them, and still overrates the Saints by the same amount). The Colts and Patriots had help from other teams in making their schedules look more convincing, and BPR adjusted them up accordingly (though BPR still thinks ESPN is slightly overrating both). For this week, BPR rates the Packers, Steelers, Cowboys, Saints, and Eagles as the five most overrated teams.

Of the five teams BPR considered most underrated last week, one (the Rams) won in an upset and two others (Vikings and Jets) came within seconds of upsets (the Cardinals won as expected and the Giants lost as expected). This week, BPR ranks the Chiefs, Giants, Rams, 49ers (see the end of the first note), and Texans as the five most underrated teams. And BPR doesn't even know how close the Texans theoretically are to being 6-1.

Looking Back

BPR was 10-5 this week in predicting games, whiffing on Houston/Pittsburgh, Seattle/St. Louis, Jacksonville/Cleveland, Miami/Chicago, and Kansas City/San Diego. Over the three weeks so far, that's a combined record of 31-13, which isn't too shabby, though, of course, it doesn't account for the spread.

Looking Forward

It looks like BPR actually has some upset picks this week (it missed the half-hearted upset prediction of Houston over Pittsburgh last week). Namely, BPR picks Minnesota over Tampa Bay, Baltimore over Cincinnati, Buffalo over the Jets, and... Green Bay over New Orleans? That's an upset? Now, it is significant to note that all four of these picks are of away teams over home teams, which could be partly just a result of BPR not knowing what teams are home or away when. So any confidence in these picks should be inversely related to confidence in the effect of home field advantage.

That wraps it up for this week - as always, I'd be interested to hear thoughts on the rankings and feedback or ideas for these articles.