I was feeling unloved so I delayed posting the graph to make y’all whine. No, seriously, my schedule just got away from me this week. Here’s the beatpaths graph for Week 14! We were 9-7 last week. Overall this has been a lousy year for picks, but I guess that was never really the point in the first place. Teams just keep behaving differently than they have so far throughout the season.
This week’s beatpaths graph is really surprising. I’ll have a lot of explanation to give about Denver in this week’s power rankings.
We have quite a few games classified as beatflukes this week:
ARI=>SEA
ARI=>STL
KC=>DEN
NYJ=>NE
MIA=>NE
JAC=>IND
NYJ=>BUF
TEN=>IND
SF=>SEA
DET=>BUF
TB=>CIN
CAR=>NO
CAR=>CLE
CAR=>BAL
KC=>SD
OAK=>PIT
ATL=>PIT
ATL=>CIN
WAS=>JAC
HOU=>JAC
SEA=>DEN
STL=>DEN
GB=>MIN
SF=>STL
SF=>MIN
DET=>ATL
HOU=>MIA
HOU=>JAC (again)
GB=>MIA
DAL=>IND
And only 8 beatloops left over:
ARI=>SF=>OAK=>ARI
CAR=>TB=>PHI=>CAR
CAR=>TB=>WAS=>CAR
TB=>PHI=>DAL=>TB
BAL=>SD=>DEN=>BAL
TEN=>PHI=>DAL=>TEN
BUF=>MIA=>CHI=>BUF
MIA=>CHI=>NYJ=>MIA
The three highest ranked teams are Chicago (k), New Orleans (k), and… Minnesota? That’s really unexpected. I guess this would put them on the inside track for the sixth seed. You should say that in the power rankings and then look super smart when they win out with that cupcake schedule.
And I would think more explanation would be needed for Buffalo than for Denver. Didn’t that team lose 40 – 7?
Yeah, Buffalo will actually be ranked #7 this week. Wow.
And you’re right, MIN #3 in the NFC. The AFC looks so dominant in the graph this week that it’s easy to overlook the NFC. The NFC #6 seed (according to the graph) will be ranked #20 in the power rankings. That’s just sad.
It should be no surprise that almost the entire AFC is ranked above the NFC; by my count, the NFC is a pathetic 21-35 against the AFC (led by the horrendous NFC North at 4-11). Only 3 NFC teams have winning records against the AFC: Dallas, Carolina, and Seattle. On the flip side, only Oakland, Denver, Houston, and Cleveland have losing records against the NFC.
Looking forward to the ratings/explanation. Clearly from the graph NE is “artificially” propping up Denver. If Cin->Den next week, I presume that will become a beatloop and disappear. In fact, the whole DEN->*BUF->PIT->NO chain looks “suspicious” from what we know otherwise. However, like I say to the DVOA guys, it is good that you publish the objective measure and use the commentary to question when you think the measures are mistaken. That allows us access to the data and we can apply our own judgment as to why things “don’t look right”. I must admit that the ordering is almost a wish fulfilment for me for this season–it would be so great if it were true.
BTW, in the off-season, I’ll talk to you about playing with your software to do things “differently” (like alternate loop breaking strategies). You had some ideas where one could use weights (and I think I could implement them for you), and I think one could use age as a weighting factor to “weaken” old beats, so that teams that are improving or falling don’t get propped up by their old wins quite so much (e.g. Den->NE and NE->CIN in this graph)–the graph might look the same if their isn’t contradicting evidence, but some of the flukes and loops might get resolved differently.
It seems that the Titan’s win over the Giants is what’s really killing the NFC. Take that away, and the graph is probably a *lot* flatter.
Gah, Titans’ win. I hate misplaced apostrophes.
Poor Houston. You beat a team twice in a season, and both times it’s considered a fluke.
Regarding the “aging” of wins, it really depends on what you want the rankings to indicate. If you want the rankings to show the best team right now, then aging wins could be appropriate. But the season is 16 games long, so if the goal is to have the rankings indicate the best team over the season, then a win at the beginning of the season should count as much as a win at the end of the season.
8: Yes. exactly. Those are two separate questions, and they are likely to have different answers, and it would be nice to know both answers. I’m not suggesting eliminating what the beatwins report now. I just think it would be nice to augment it with more info.
I mean as a Broncos fan, I’m psyched to see them so high up in the chart, but I don’t think that really describes how good of a team they are. I think it overemphasizes that they beat NE before NE got better (and has since regressed somewhat) and underemphasizes that they just lost 4 games in a row, including 2 considered flukes, which maybe weren’t flukes at all.
Is it possible that beatflukes just aren’t done correctly? Looking at the chart, about 12% of the season or so right now is considered to be a fluke. Doesn’t that seem kind of high? Maybe instead of 1 alternate beatpath to create a fluke, 2 should be needed instead?
#10: There was one other time this season that I wondered the same thing, but both this time and last time, I checked the non-beatfluke version of the graph and things really aren’t a lot different. The beatfluke version is never extraordinarily different than the non-beatfluke version, just slightly more vertical, with a handful of teams that have rankings that differ more than a couple of notches.