Here are the Week 14 Power Rankings. We’re 112-96 for the season. The main differences are that Buffalo is better than you think, and Pittsburgh’s rise is affecting a lot of other teams.
Right now, the graph suggests the following playoff seedings, which are quite different than the standings indicate.
AFC: IND, SD, BAL, NE, DEN, CIN
NFC: CHI, NO, SEA, NYG, MIN, ATL
For review, here’s the beatpaths graph:
And, the rankings:
| Rank | Team | Notes | Last Week | BeatPower |
1 |
 |
(Lost to JAC) JAC=>IND is considered a beatfluke due to its alternate beatpath. In addition, DAL=>IND is also now considered a beatfluke, thanks to Jacksonville’s beatpath to New Orleans (Pittsburgh’s victory helped with that). So Indianapolis also regains their direct (but currently invisible) beatwins to Jacksonville and Philadelphia. |
1 |
100.0(29/29 – 0/29) |
2 |
 |
(Beat DEN) San Diego now has a beatpath to Denver. Since that gives San Diego an alternative beatpath to Kansas City, KC=>SD is now considered a beatfluke. So San Diego’s victory enables them to regain their beatwins to Pittsburgh, Cleveland, and Cincinnati. San Diego’s other win over Denver is what enables them to shed their beatloss to Baltimore. |
4 |
98.3(29/30 – 0/30) |
3 |
 |
(Beat KC) Baltimore now has a direct beatwin over Kansas City, but they have an alternate path to them anyway. CAR=>BAL is now considered a beatfluke, so this also enables Baltimore to regain a direct beatwin to Atlanta. None of this is very relevant to Baltimore’s high placement, holding steady at #3. |
3 |
94.2(23/26 – 0/26) |
4 |
 |
(Lost to SD) Since San Diego is ranked so high, Denver isn’t hurt badly by the loss. Beatpaths looks at the entire season, not momentum, and the fact is that Denver has had a murderous schedule, and their quality wins (NE, PIT, BAL) have been enough to cancel out most of their losses. San Diego’s sweep makes them lose credit for defeating Baltimore, but the poor performance of other teams makes beatflukes out of KC=>DEN, SEA=>DEN, and STL=>DEN. Denver regains its direct beatwin over Kansas City due to KC’s loss. Seattle’s beatfluke means Denver regains credit for defeating New England. St. Louis’ beatfluke means that Denver regains credit for defeating Pittsburgh. For this week, Denver looks kind of like Dallas in 2005 – the overall season’s performance still warrants Denver making the playoffs, even though the normal NFL standings would have Denver sitting them out. |
8 |
91.9(28/31 – 2/31) |
5 |
 |
(Lost to MIA) Surprising shutout, but MIA=>NE is considered a beatfluke. SEA=>DEN being a beatfluke means that NE=>CHI/MIN=>SEA is restored to the graph, even while DEN=>NE reappears. |
2 |
90.0(27/30 – 3/30) |
6 |
 |
(Beat OAK) Since KC=>SD is now considered a beatfluke, SD=>CIN=>KC gets restored to the graph. No major change for Cincinnati this week. |
5 |
83.9(23/28 – 4/28) |
7 |
 |
(Beat NYJ) The first clue here is that Detroit is a very bad team. Buffalo lost to Detroit earlier this year, and that one loss was keeping Buffalo from beatwins over Miami, Green Bay, and Minnesota. Buffalo also was kept from a victory over Jacksonville due to an earlier loss to the Jets. So defeating the Jets was a very key victory. It enabled them to form an alternate beatpath to Detroit through Jacksonville. The end result is that NYJ=>BUF and DET=>BUF are now considered beatflukes, and Buffalo now has credit for their victories over Jacksonville and Minnesota. Go check Buffalo’s game history again – they kind of snuck up on everyone, and their losses are to extremely good teams. Buffalo is still kept from their victory over Miami due to losing to Chicago. |
17 |
84.5(24/29 – 4/29) |
8 |
 |
(Beat IND) Jacksonville gains benefit by PIT=>NO re-emerging. While JAC=>IND is currently considered a beatfluke, so now are WAS=>JAC and both HOU=>JAC victories. As a result, Jacksonville regains beatwins to PHI, TEN, NYG, and DAL, while also suffering beatlosses to IND and BUF, but neither of them hurt very much. |
15 |
82.1(23/28 – 5/28) |
9 |
 |
(Beat CLE) OAK=>PIT and ATL=>PIT are now considered beatflukes, and all of Pittsburgh’s other games are now reflected in the graph. SD=>PIT re-emerges, as does PIT=>KC, PIT=>NO, and PIT=>CLE. It’s PIT=>NO that really helps them out here. |
21 |
73.3(22/30 – 8/30) |
10 |
 |
(Beat STL) Due to SEA=>DEN being a beatfluke, DEN=>NE=>CHI=>SEA re-emerges, which causes Chicago to fall a bit. CHI also regains a beatwin over Green Bay. |
6 |
75.0(15/22 – 4/22) |
11 |
 |
(Beat DAL) CAR=>NO is now considered a beatfluke, which means that all of New Orleans’ other games are reflected in the graph. In addition to NO=>DAL appearing, NO=>ATL and NO=>PHI re-emerge. |
10 |
67.9(19/28 – 9/28) |
12 |
 |
(Beat NE) While MIA=>NE is considered a beatfluke, Miami is helped by other factors. HOU=>MIA and GB=>MIA are both considered beatflukes, apparently mostly because of Seattle’s inconsistency. As a result, Miami regains credit for defeating Tennessee. Also, due to HOU=>JAC and STL=>DEN beatflukes, JAC=>MIA=>MIN and PIT=>MIA=>KC both re-emerge as well, which helps Miami more than it hurts. |
19 |
64.3(17/28 – 9/28) |
13 |
 |
(Lost to PIT) CAR=>CLE is considered a beatfluke, and all other Cleveland games are now reflected in the graph. CLE=>ATL, PIT=>CLE=>OAK, and SD=>CLE=>KC all re-emerge. After all that, Cleveland remains steady. |
13 |
63.0(17/27 – 10/27) |
14 |
 |
(Beat DET) GB=>MIN and SF=>MIN are both considered beatflukes, and all other games are now reflected in the graph. MIN=>ARI, BUF=>MIN=>DET, NE=>MIN=>SEA, and MIA=>MIN=>WAS all re-emerge. |
12 |
56.0(14/25 – 11/25) |
15 |
 |
(Lost to BUF) The Jets’ earlier victory over Buffalo is now considered a beatfluke. Buffalo had a beatpath to the Jets anyway, though – the difference is that JAC=>NYJ emerges, as does NYJ=>HOU. Not a big difference to the graph other than Buffalo leapfrogging them. |
14 |
55.6(14/27 – 11/27) |
16 |
 |
(Lost to BAL) KC=>DEN and KC=>SD are now considered beatflukes, and all other games are reflected in the graph. PIT=>KC, CLE=>KC, CIN=>KC, and MIA=>KC=>STL all re-emerge – this hurts Kansas City and they fall in the rankings. |
7 |
53.8(14/26 – 12/26) |
17 |
 |
(Lost to ARI) ARI=>SEA is considered a beatfluke, so Seattle isn’t hurt by that. But SEA=>DEN is also considered a beatfluke, so Seattle again suffers under Chicago and Minnesota. |
9 |
46.4(13/28 – 15/28) |
18 |
 |
(Beat HOU) Due to other beatflukes, JAC=>TEN=>WAS and MIA=>TEN=>HOU both re-emerge in the graph. The Miami reappearance probably hurts Tennessee slightly. |
16 |
50.0(13/27 – 13/27) |
19 |
 |
(Beat CAR) Aside from the DAL/NYG split, all the Giants’ games are reflected in the graph. JAC=>NYG=>WAS reappears, as does NYG=>HOU. |
22 |
40.3(12/31 – 18/31) |
20 |
 |
(Beat TB) ATL=>PIT is now considered a beatfluke, but so is DET=>ATL. All other games are reflected in the graph. ATL=>ARI, NO=>ATL=>CAR, CLE=>ATL, and BAL=>ATL all reappear. On balance this helps Atlanta rise slightly. |
26 |
34.5(10/29 – 19/29) |
21 |
 |
(Beat WAS) Other beatflukes simplify Philadelphia’s graph situation a bit, but not much. NO=>PHI, JAC=>PHI=>WAS, PHI=>HOU, and IND=>PHI all re-emerge. |
25 |
29.3(7/29 – 19/29) |
22 |
 |
(Lost to NO) DAL=>IND is now seen as a beatfluke. Due to the HOU=>JAC beatflukes, IND=>JAC=>DAL=>HOU re-emerges. It’s NO=>DAL that hurts more, though. Dallas falls significantly. Dallas just doesn’t seem to have much in the way of quality victories other than Indianapolis. |
11 |
47.8(9/23 – 10/23) |
23 |
 |
(Lost to NYG) Carolina’s wins over NO, CLE, and BAL are all seen as beatflukes (again). The only real difference is that ATL=>CAR is restored to the graph. |
18 |
24.1(6/29 – 21/29) |
24 |
 |
(Lost to CHI) STL=>DEN and SF=>STL are considered beatflukes, so the previous STL=>SF win is considered valid again, as is the earlier KC=>STL victory. |
23 |
15.4(4/26 – 22/26) |
25 |
 |
(Lost to PHI) WAS=>JAC is now seen as a beatfluke, so Washington now has beatlosses to PHI, TEN, and NYG again. |
28 |
13.5(2/26 – 21/26) |
26 |
 |
(Beat SF) GB=>MIA and DET=>BUF are now beatflukes, so the end result there is that CHI=>GB and GB=>DET are added back to the graph. Not a big difference. |
29 |
11.1(3/27 – 24/27) |
27 |
 |
(Lost to TEN) HOU=>JAC (x2) and HOU=>MIA are all considered beatflukes, so Houston suffers under TEN, DAL, PHI, NYG, and NYJ again. |
20 |
4.2(1/24 – 23/24) |
28 |
 |
(Lost to ATL) There are actually no graph changes for Tampa Bay this week – they have the same relationships as last week. |
30 |
4.2(0/24 – 22/24) |
29 |
 |
(Lost to CIN) OAK=>PIT is now a beatfluke, so Oakland suffers under KC and CLE. |
24 |
3.8(0/26 – 24/26) |
30 |
 |
(Lost to GB) SF=>STL and SF=>MIN are now considered beatflukes. This causes no visible change to the graph, but GB now has a beatpath to SF. |
27 |
7.1(1/28 – 25/28) |
31 |
 |
(Beat SEA) ARI=>SEA is considered a beatfluke. ATL=>ARI=>DET re-emerges. |
31 |
7.1(1/28 – 25/28) |
32 |
 |
(Lost to MIN) DET=>BUF and DET=>ATL are both considered beatflukes, so Detroit suffers under MIA, MIN, GB, and ARI. |
32 |
0.0(0/27 – 27/27) |
First off, thanks for all the hard work.
Secondly, I think you have the comment for CAR in the STL box, and some placeholder text in the CAR box.
Finally…I’m a Bears fan, and I’m sad to see their placement. I realize the NFC sucks, but it’s annoying to see that keep them ranked in the back. Well, that and the fact they went 2-2 against the AFC East.
I tried to figure out ways that the NFC could get back up in the next few weeks, but there aren’t that many inter-conference games left, and most of them would have to be considered miracles if they happened. So unless the AFC gets less well-ordered, looks like the NFC isn’t going to be improving its lot anytime soon…
Right you are about CAR/STL – fixed.
It’s funny, because it seems to be a flaw of the parity approach. If the NFC has somehow become clearly worse than the AFC, then the limited number of interconference games means that the NFC teams’ records will continue to be slightly inflated, while the AFC teams’ records will be lower, which could mean the AFC teams will continue to have an unfair draft advantage, further exacerbating it. Or maybe it’ll just solve itself somehow, I don’t know.
Maybe I’m missing some small part of how beatflukes and beatloops work, but it would seem to me that the immediate beatloop you remove from both Indianapolis and Jacksonville would be JAC=>IND=>JAC. Therefore, shouldn’t the Colts’ win over the Jaguars earlier in the year also not count? Maybe it doesn’t and I’m misreading something, or maybe it wouldn’t reflect on the rankings at all. Could you maybe shed some light on this. Thanks.
P.S. I really enjoy this site, it is quite unique and always a good read.
A season split is considered just another beatloop. So after JAC=>IND=>JAC is taken out, then it looks to see if one of those teams still has an alternate beatpath to the other. IND had an alternate beatpath to JAC, so it contradicted and knocked out JAC=>IND.
If the alternate beatpath didn’t exist, then the JAC=>IND=>JAC beatloop (split) would just be taken out.
I do want to ask about the JAC/HOU situation. I understand that the system is considering both HOU=>JAC wins as flukes since JAC has at least 1 other beatpaths to HOU. My question was, if JAC has 1 alternate beatpath to HOU, is that enough to qualify both games as flukes? Or does JAC have multiple alternate beatpaths to HOU?
Well, it considers them one round at a time. This is a pretty complicated scenario. I find small beatloops first, then bigger ones. In the non-beatfluke variant for this week, both HOU=>JAC victories are taken out. First it takes out season splits, and then takes a fresh look at the graph. It finds a bunch of three-team beatloops and takes them out, which removes one of the HOU=>JAC victories. Then it goes through the graph again to find more beatloops – all the three-team beatloops are gone, but there might be longer ones. And it finds a longer five-team beatloop, which takes out the other HOU=>JAC victory.
So then it goes through the successfully “directed” graph and sees if JAC has a beatpath to HOU. It does, so it declares ONE of the HOU=>JAC victories a beatfluke. What this means is that the system would not have found that longer five-team beatloop. So, that is a new beatpath segment that is restored to the graph, which means that JAC then has another alternate beatpath to HOU.
So honestly, I can’t think of a scenario where it would declare one of the victories a fluke and not the other. It seems to be that either one of the victories would be in the graph (a HOU=>JAC) beatpath, or they’d both just be beatlooped, or both victories would be considered fluke.
It’s a very good question. But you know, it’s funny – doesn’t it just kind of feel weird to consider the possibility that one HOU=>JAC victory is a fluke while the other one isn’t? Beatfluking isn’t so much about whether to invalidate a game, it’s more about determining whether a team is better than another despite losing to them. And in that sense, both victories are accounted for by finding the differently-sized beatloops, after which JAC *still* has a beatpath to HOU even after all these loops are taken out. JAC is seen as just plain old better than HOU, despite both losses… so… using that logic, it would make sense to declare both of them fluke.
You can tell I’m not sure whether that’s a rationalization or not. Let me look at the non-beatfluke version… right now we’ve got them at #8 and #27 respectively. In the non-beatfluke version we have them at #7 and #21 . I think I’m still okay with it working this way.
(Wow… the non-beatfluke version has Buffalo ranked #4.)
I guess that makes you the only person outside of upstate NY who called the Buffalo/Miami game correctly.
Yeah, I’m trying to decide whether it means that beatpaths is actually reliably valuable for judging a certain type of matchup, or if it’s a broken clock thing.