2010 NFL Week 13 Beatpath Graph

A more accurate week, and maybe it’s just me coming around to what beatpaths has been telling me, but this one is the first week where I felt like the upsets really were upsets. The rankings were 12-4 this week. Once again it’s SEA->CHI that is sticking out like a sore thumb. I haven’t taken a close look at the rest of the season, but it seems there has got to be a big rankings shift coming soon.

2010-13-nfl-clean.png

14 Responses to 2010 NFL Week 13 Beatpath Graph

  1. Tom says:

    The graph has gotten a lot less vertical over the past four weeks. I guess parity really does remain the theme this season.

  2. ThunderThumbs says:

    That is common, right? As the season splits start happening?

    At some point I should figure out how to animate how a graph changes over a season. It seems you can control the movement of team boxes, take snapshots, etc. I wouldn’t be surprised to see a common pattern of vertical, then flat, then slowly vertical again – the nba and mlb graphs always seem to end up in roughly the same dimensions.

  3. dbt says:

    Not that I want to get deep into what-ifs, but if SF beats Seattle this weekend would that finally clear their win over the Bears?

  4. The MOOSE says:

    I’ll take a look through my histories and compare this week’s height to previous years after week 13. But in general the height has a lot to do with the method used to remove loops. The Standard Method removes a lot of loops which shortens the graph.

    Also, the height of the graph doesn’t necessarily have a lot to do with parity, because it’s really all about whatever the longest path is. If the longest path is 10 teams long, the graph has to be 10 teams high, but that doesn’t say much about how the other 22 teams are spread out.

    This week, I have Standard’s longest path at 6, compared to Iterative’s 17-team path and Weighted’s 20-team longest path.

    While it would be difficult to check historically how graphs change “as the season splits start happening” since this is the first year where division games have been spaced out in a way to make that measurement possible, I actually thing it works opposite of that as a loop that would eliminate 3 games will restore 2 to the graph when one of the 3 links turns into a split, adding height to the graph instead of removing it.

  5. The MOOSE says:

    After looking through the histories, it turns out that this year’s Week 13 Standard graph is the shortest one ever. There have been 4 years (1971, 1983, 1991, 2003) where the height at this point was 7-teams high.

    According to the numbers, there doesn’t appear to be any affect at all with parity. If you compare pre-salary cap to post-salary cap as your split point for parity, the average graph height has actually increased from a pre-cap 10.17 to a post-cap 10.75. But now both the longest (16-team 2001) and shortest (6-team 2010) happened post-cap.

    With regards to movement from vertical→flat→vertical, I don’t think there’s a way to compare NFL to the other sports since teams play each other a lot more and there’s no time during which splits could be considered “in effect”.

  6. JT says:

    That SEA->CHI path is interesting. Since the NFC West is playing the NFC South this season, and the NFC North is playing the NFC East, that makes this one of the strength of schedule games, and just a bit harder to loop away than a division game or a game against an opponent in the scheduled opposing division.

    Both Chicago and Seattle beat Carolina (Chicago’s other SoS opponent) and both lost to the NY Giants (Seattle’s other SoS opponent). Those are the two common opponents between the two team’s schedules and about the only way you’re going to get a 3 team beatloop including these two teams. So the only way this is going to get looped is if a 4+ team loop develops and survives the season splits and 3 team loop removal. And if I’m reading MOOSE’s page right, currently there is only 2 loops of 4+ teams that are surviving long enough to be removed.

    It might be interesting to label each path with the type of game (division, intraconference, extraconference, strength of schedule) and see which paths are surviving. I wouldn’t be at all surprised if a larger percentage of strength of schedule games than other games survive loop removal.

  7. JT says:

    For an example of what I’m talking about, I looked at the teams in the NFC West and their strength of schedule opponents, and how many remain in the graph.

    Arizona lost to Minnesota and has not yet played Dallas. MIN → ARI looped in ARI → NO → MIN → ARI, based on Minnesota’s other SoS opponent.

    San Francisco lost to Philadelphia and lost to Green Bay. Both are still in the graph.

    Seattle beat Chicago and lost to New York Giants. Both are still in the graph.

    St Louis lost to Detroit and beat Washington. STL → WAS remains in the graph, DET → STL is removed in BUF → DET → STL → SD → JAC → BUF.

    The St Louis one is the most interesting, as it includes JAC → BUF, which is also a SoS game.

  8. The MOOSE says:

    @JT: Yes you’re reading that correctly. The interesting thing about those two 5-team loops is that neither existed before this week. You’re also right that it is particularly difficult to remove a game involving your SoS opponents, due to a lack of common opponents.

    Labeling the graph with type of game wouldn’t be helpful though as many arrows are hidden due to redundancy.

  9. JT says:

    I guess I don’t suggest actually displaying in the pretty graph you guys display, just if it were possible to add it to the raw data and look at if different types of games survive the loop removal process better than others. It’d give some insight into how persistent paths could be in the graph in future seasons.

    It’d actually be best done on your past seasons, at least back to 2002 when the Texans joined the league and the current schedule formula was put in place.

    With something like that, you might be able to note that a path based on a SoS opponent has a 15% chance of being looped away, while a intraconf win has a 40% chance of being looped (numbers purely made up).

  10. The MOOSE says:

    That sounds like a good end-of-season activity. It would also correlate with interleague games in the MLB graph, possibly the other leagues too. Today’s NHL graph has nearly the entire Western Conference in a straight line beside the more jumbled Eastern Conference.

  11. JT says:

    The real correlation in MLB would be the Interleague rivals (Mets/Yankees, Cubs/White Sox, Giants/Athletics, etc) that are the interleague games that certain teams play every year no matter the interleague division pairings. These, like the strength of schedule games in the NFL, are the only ones that are determined on an individual team by team basis rather than a division vs division basis. Thus, they have less common opponents.

  12. The MOOSE says:

    Somewhat, but not exactly. Each time has their “rival” and each division is supposed to play an opposing division. But because the divisions aren’t even, it doesn’t work out well. For example, last year in the AL East was to play the NL West. Here is who played whom.

    BOS: rival – PHI, NLW – LA, COL, SF, ARI
    BAL: rival – WAS, NLW – SF, SD .. NLE – NYM, FLA
    NYY: rival – NYM, NLW – ARI, LA .. NLC – HOU .. NLE – PHI
    TOR: no rival, NLW – ARI, COL, SD, SF .. NLC – STL .. NLE – PHI
    TB: rival – FLA, NLW – SD, ARI .. NLC – HOU .. NLE – ATL

    So while NYM was NYY’s rival, BAL also played them. Similarly, while the AL East was to play the NL West, neither BOS nor NYY played SD.

    I think the thing that really throws a wrench into the comparison though, is the fact that teams play significantly more games against each other. A 3-team loop doesn’t break the same way that it does in football because Team A may have 5 direct wins over Team B, B have 2 over C, and C 1 over A. That would just break the C → A link and leave A → B → C.

  13. Kenneth says:

    I want to say that I’ve been following this discussion but haven’t had time to comment on it, but I think a lot of it sounds cool.

  14. ThunderThumbs says:

    dbt’s comment got caught in a spam filter, but he’s right – SEA->CHI was finally obliterated this week.

Leave a Reply

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