A more accurate this week for the picks – Isaacson-Tarbell was 12-4, Beatpaths was 11-5, and ITB was 10-6. I was 2-0 on the personal picks, so I am now two ahead of ITB again. The BeatPicks were 8-1 this week, or 39-14 for the season.
And here’s the NFL Week 12 Beatpath Graph! A lot of larger loops got shrunk this week – full list of loops after the jump.
Splits:
- TEN/JAC
- MIA/BUF
- OAK/KC
- ATL/CAR
- SD/DEN
- NE/NYJ
- HOU/TEN
Loop: WAS=>STL=>DET=>WAS
Loop: ARI=>HOU=>SF=>ARI
Loop: HOU=>SF=>JAC=>HOU
Loop: HOU=>BUF=>NYJ=>HOU
Loop: NYG=>WAS=>DEN=>NYG
Loop: NYG=>DAL=>PHI=>NYG
Loop: NYG=>OAK=>PHI=>NYG
Loop: CHI=>PIT=>MIN=>CHI
Loop: TB=>GB=>DAL=>TB
Loop: KC=>PIT=>SD=>KC
Loop: DEN=>CIN=>PIT=>DEN
Loop: BAL=>DEN=>CIN=>BAL
Loop: DEN=>DAL=>WAS=>DEN
Loop: BAL=>DEN=>NE=>BAL
Loop: BAL=>PIT=>MIN=>BAL
Loop: BUF=>NYJ=>TEN=>BUF
Loop: OAK=>CIN=>BAL=>SD=>OAK
Loop: ARI=>JAC=>NYJ=>CAR=>ARI
Loop: ARI=>JAC=>BUF=>CAR=>ARI
A couple weeks ago I asked if there was ever a flatter graph that late in the season? Looks like it’s reversed now.Has there ever been a more vertical graph, ever? It’s odd how it can all turn around so quickly.
It looks like Arizona is the key to the Beatpaths season graph. I remember a few weeks ago they really shook things up and it appears their loss to TEN this week has created two different shelves again.
Back when he was using betflukes, there were more vertical graphs. But I don’t think I’ve seen a more vertical graph using TT’s standard algorithm.
By the way, you want vertical? Check out the iterative graph:
http://beatgraphs.com/archive/iterative/2009/I_2009_12_NO.php
The longest path is 22 teams long. There are only 8 teams that AREN’T on the longest path. (Denver, Indy, and New Orleans are all on the longest path by virtue of paths to New England.)
For both the iterative and standard (parallel) approach, the key game that stretched everything this week was Tennessee beating Arizona. Take that game out and things flatten quite a bit.
Looking at only week 12s as far back as my data currently goes, 2004 was equally tall with a longest paths of 15 teams. Week 12 of 2002 had a 16-team path from ATL to DET. But the longest of week 12 was in 2001 with a 17-team path from CHI to JAC.
So while the turnaround is dramatic over 2 weeks, a 15-team high graph isn’t that abnormal.
Now that the graphs are so vertical, it’s easy to see which teams there is significant disagreement on.
Houston: parallel has them as an elite team with a path to Cincy. Iterative has them as a mediocre team, wedged between the Jets and Oakland in the middle of the graph.
Baltimore and Denver: In parallel, neither team has any bad losses, but they both only retain wins to weak teams. In iterative, both still retain some of their strongest wins, and are up at the top of the graph.
Philadelphia: their best win is Chicago in both cases, but in parallel the only teams with beatpaths to them are San Diego and New Orleans. In iterative, Oakland retains its path as well.
I think those are the only dramatic differences.
Yeah, interesting. The rankings will be wacky this week. Pittsburgh rises, Baltimore falls. Baltimore was favored over Pittsburgh last week; this week it sees Baltimore’s victory as an upset.
As for Houston, I think they’re being propped up by something that will likely disappear if an expected season split happens. Don’t remember exactly what I was thinking when I noticed that last night, but that’s what occurred to me.
I’m not sure how that would work. The only season series Houston has not concluded is Jacksonville, who they lost to… a season split there would only help them.
What you might be thinking is that if CIN beats NYJ in the last game of the season, that will form a three-team beatloop that will wipe the game out. The issue with HOU->CIN is that same issue that often plagues the standard algorithm. That is, the two “strength of schedule” games are much harder to shake than any other game, because they can’t be in any three team beatloop except those involving the other two teams from that conference in that SoS tier (in this case, Oakland and the Jets).
As it happens, there aren’t even any four team beatloops involving that game currently. In order to wipe that game out, the iterative algorithm considers three six team beatloops, STILL retains the game, and only wipes it out thanks to two seven team beatloops:
HOU → CIN → PIT → TEN → ARI → HOU
HOU → CIN → CHI → SEA → JAC → HOU
HOU → CIN → GB → SF → JAC → HOU
HOU → CIN → BAL → SD → MIA → NYJ → HOU
HOU → CIN → CHI → SEA → JAC → NYJ → HOU
It’s going to be interesting to see what happens if Houston gets the season split with Jacksonville… that game may re-appear in interative as well. In that case, it will probably stick around in both versions unless/until Cincy can beat the Jets in week 17, and/or if Houston has a run of bad losses. (The only bad loss the Texans have at this point, assuming we consider Tennessee good, is the Jets.)
I looked in the How it works section of the site, but I don’t understand how the vertical position of some teams is set in the graph.
Specifically, I’m a GB fan and I was thinking about the matchup with Baltimore this weekend. GB is placed just under Minnesota and Cincy with a long path down to SF. Meanwhile, Baltimore is just above Cleveland and KC, with a long path above to Indy.
Why is GB so high and Baltimore so low? It seems that the vertical positioning along these long beatpaths is arbitrary.
The only thing I can think of is local support. GB has two paths above and only one below, so the above paths have precedence. Meanwhile Baltimore has one above and two below, so the below paths have precedence.
If this is the case, it seems some sort of weighting would be more accurate, so that GB would be on the level with Tennessee, while Baltimore could be on the SF or Atlanta level.
Arkaein, I think the reason the vertical placement is such as it is, is (wow is this a record for how many times is is in the same sentence? I think it is) because graphviz tries to minimize the arrows – it tries to make as compact and readable graph as possible.