NFL Week 12 Beatpaths Power Rankings

The tiebreaking algorithm of the week is one that looks at the average beatpower of the top n direct beatwins of each team being compared, where n is the number of direct beatwins for the team with the least. 41 3-team beatloops this week, and 2 splits. We’re still not using the beatflukes graph, but if we were, it would cancel out nine beatloops.

Rank Team Notes Last Week BeatPower

1

(Beat PIT) Last week, Denver had slightly stronger direct beatwins, but Indy’s win over Pittsburgh solidifies their hold on top. At this point, the Denver fan in me wants Indy to remain undefeated so Denver can play them for the AFC Championship. I don’t think anything else would get the monkey off their backs.

1

100.0

(23/23 – 0/23)

2

(Beat DAL) A great victory for Denver keeps them in second. Denver’s high placement has been partially dependent on the apparent strength of their direct beatwins, and Washington’s collapse has been steadily weakening Denver’s position. Denver needed this victory.

2

92.3

(22/26 – 0/26)

3

(Beat ARI) As long as Jacksonville keeps winning, there isn’t a lot of reason to rank them lower, despite most sportswriters’ belief that they’re only a second-tier team. Their game against Indianapolis will be quite interesting.

3

88.1

(18/21 – 2/21)

4

(Beat NYG) Seattle still appears to be the class of the NFC. However, the beatfluke variant of the rankings has Seattle ranked a few slots lower, because of their loss to Jacksonville. It doesn’t count here because of Jacksonville’s loss to St. Louis.

5

91.2

(14/17 – 0/17)

5

(Lost to IND) The Steelers don’t get penalized much for losing to the top team in the league, and they’re helped by the collapse of Washington.

6

81.8

(17/22 – 3/22)

6

(Beat BAL) The Bengals basically hold steady, using their offense to make up for their defense.

7

80.0

(16/20 – 4/20)

7

(Lost to CHI) Here’s the strange behavior of the week. Tampa Bay loses to Chicago, and Tampa Bay rises while Chicago falls. How could this be? Here’s a hint – the placement of each team is very dependent on their relationship to other teams. No beatpath was formed with Chicago because it created a TB=>WAS=>CHI loop, but what was more important was that TB kept their beatpath to MIN.

10

83.3

(12/18 – 0/18)

8

(Lost to KC) The Pats basically hold steady, although KC was helped by their victory.

9

73.1

(7/13 – 1/13)

9

(Beat WAS) The Chargers and the next few teams rise a couple of slots due to Washington’s collapse and Chicago’s readjustment.

11

73.1

(6/13 – 0/13)

10

(Lost to DEN) The Cowboys don’t get hurt badly by their loss to the Broncos, but they have some key games coming up that they’ll have to win.

12

73.1

(7/13 – 1/13)

11

(Beat CLE) Here’s the key to the rankings movements this week. By defeating Cleveland, the Vikings were able to develop a beatloop that let them shake off their beatloss to Chicago. Combine that with Minnesota’s history, and there’s been a real paradigm shift with the Vikings. It turns out they’ve only lost to very good teams, and they’ve beaten some pretty good teams as well. If they keep it up, they could make the NFC North pretty interesting. At this point, they are actually in a slightly stronger place in the graph than Chicago is.

13

61.9

(11/21 – 6/21)

12

(Beat TB) Chicago falls in the graph because they lose their beatwin over Minnesota, which had a fairly large chunk of the graph, including the Giants.

8

58.8

(8/17 – 5/17)

13

(Beat NE) The Chiefs help themselves with their win over the Pats, but they’re going to have to improve their consistency to go much higher.

16

63.6

(4/11 – 1/11)

14

(Beat DET) The Falcons don’t get much help by defeating a weak opponent.

14

53.8

(4/13 – 3/13)

15

(Beat BUF) The Panthers don’t, either. This team has been confusing all year. When I generate the graphs of power rankings placement, this team’s variance will be pretty apparent. It’s probably one of the highest.

15

50.0

(6/19 – 6/19)

16

(Lost to SD) A pretty spectacular collapse, one of the largest of the season. I’ve been trying to find a tiebreaking algorithm that would have led Washington degrade a bit more gently, but the problem always has been that they really did have that beatpath over Seattle until this last week. Washington had just seen their reinforcements be chipped away little by little over the previous weeks. In the beatfluke graph, Washington would get one beatwin back, but it would be over San Francisco. So this is where Washington is right now – bad enough that they don’t have any convincing dominance over any other team, but good enough so that their only beatloss is to Denver.

4

45.5

(0/11 – 1/11)

17

(Lost to SEA) Their loss to Seattle is what led to Washington collapsing, so it’s something of a silver lining.

17

47.2

(6/18 – 7/18)

18

(Lost to MIN) This appears to be a pretty comfortaable place in the rankings for the Browns.

18

38.5

(2/13 – 5/13)

19

(Beat GB) This victory by itself should have been enough to help Philadelphia climb a few slots, but enough other teams performed well enough to keep the Eagles down.

19

40.9

(1/11 – 3/11)

20

(Lost to MIA) Both Miami and the Raiders seem to have warring personalities, deciding on a week-to-week basis which team to be. This time, Miami played the part of the good team.

21

31.8

(1/11 – 5/11)

21

(Lost to CAR) Their win over KC becomes a distant memory.

22

34.6

(1/13 – 5/13)

22

(Lost to ATL) Interesting that everyone seems to be agreeing that Detroit’s problem is that they haven’t been developing their young players, but that there isn’t much agreement in the Lions’ organization as to whose fault it is. Well, it appears they’re agreeing it’s Mooch’s fault, but most people outside the organization want to blame Millen, too.

20

34.2

(5/19 – 11/19)

23

(Beat HOU) The Rams had played badly enough that Houston was actually favored in this game, but everyone’s new favorite Harvard darling gives the Rams one of the biggest ranking climbs of the week.

30

25.0

(2/14 – 9/14)

24

(Lost to PHI) Most of the movement below this point is just because of minor movements in the respective teams’ beatpaths.

27

19.2

(0/13 – 8/13)

25

(Beat OAK)

23

21.9

(0/16 – 9/16)

26

(Lost to JAC)

24

15.6

(2/16 – 13/16)

27

(Lost to CIN)

26

13.3

(1/15 – 12/15)

28

(Beat NYJ)

31

13.9

(0/18 – 13/18)

29

(Lost to STL) That really was a heartbreaking loss by the Texans. I was pretty dumbfounded refreshing the GameView.

25

5.9

(0/17 – 15/17)

30

(Lost to TEN)

28

8.3

(0/18 – 15/18)

31

(Lost to NO)

29

10.5

(0/19 – 15/19)

32

(Beat SF)

32

10.9

(0/23 – 18/23)

8 Responses to NFL Week 12 Beatpaths Power Rankings

  1. Pat says:

    Well, to be honest, Washington going from 3-0 to 5-6 was a pretty spectacular collapse, as well. 🙂

    I’m really constantly amazed at how well such a simple algorithm does at ranking teams. It’s really impressive. I’m still not so impressed with the college rankings, but I figured things would be weird there even before you did them, thanks to the poor connectedness of the teams.

  2. ThunderThumbs says:

    Thanks – I’m hoping that these algorithm variants that focus more on the strength of beatwins will make the college rankings settle out more. While the college graphs aren’t as connected, there’s more gradations between the length of beatpaths, and the number of beatwins, that might help make up for that. Plus, I’m hoping that the beatfluke graph will help with that as well because that will help create at least a few more connections in the graph.

    It’ll probably be a few more steps before I can seriously start analyzing college performance again – these variants are more complicated and until I can get some snazzier coding in there, each test takes around five minutes (as opposed to ten seconds for the nfl). Still a vast improvement over the 45 minutes from a couple of weeks ago, but ouch.

    In the meantime, I’m hoping to get an NBA graph up soon. The main question there is how to organize splitting the games out – I’ll probably just pick an arbitrary time period of a week or so.

  3. Pat says:

    Weird thing: as far as I can tell, there were no beatfluke losses in the Big Ten this year. At least, none related to intraconference play.

    Makes sense, though, as from a “common sense” point of view, there weren’t any either. The most ambiguous loss was of course Michigan over PSU, but Michigan never lost to a bad team all year, just a lot of good ones. Ended up with a whole ton of beatloops involving Wisconsin, Iowa, Michigan, Penn State, Ohio State, and Northwestern, but that’s kinda to be expected when they’re all good.

  4. Richie says:

    I was going to sarcastically suggest that you do beatpaths for MLB next year, assuming it an impossibly difficult project, but since you are going to do NBA, maybe it’s not crazy.

  5. ThunderThumbs says:

    Yeah, hard to say. I think it’s the number of teams more than the number of games that really adds the complexity. Basketball and Baseball are going to present their own challenges with all the redundant beatloops though. I have to be really careful to make sure that I’m canceling them out correctly.

  6. Pat says:

    The other problem with baseball and basketball is you have to deal with the issue “is playing a team 10 times and going 9-1 better than going 6-4?” There’s indirect benefit to doing that, of course – you’ll break more beatloops involving that team – but once all the beatloops are broken, right now I don’t think there’s a benefit to having multiple “net wins” over another team other than the extra stability it offers you. That should probably factor into the rankings somehow.

  7. Thok says:

    I’d think the obvious solution is that you remove games in groups of loops. For example if team A is 5-0 against team B, team B is 3-0 against team C, and team C is 2-0 against team A, then you reduce down to a graph with 3 lines from A to B, and 1 line from B to C. (Basically, remove exactly two copies of the
    A->B->C->A loop).

  8. ThunderThumbs says:

    Re #7: Yeah, that’s how it works. What I do is find all three-team loops, and then remove them once. Then I recalc. So if there’s a loop left over, it would find it again.

    So it would go from 5-3-2, to 4-2-1, to 3-1. A would have a beatpath to B, and B would have a beatpath to C.

    (This comment is based off of a very lazy memory of how I constructed my own algorithm, I might actually be wrong about myself. 😉 )

Leave a Reply

Your email address will not be published. Required fields are marked *