2008 NFL Week 16 Beatpath Graph

2008-16-Nfl-Clean Well.

I have had quite the crisis of confidence this week with the beatpath graph and the algorithm, all because of the freaking NY Jets.

There’s no way the NY Jets should be ranked #1, I thought. There’s no way they should be at the top of the beatpath graph, with a beatpath to almost everyone else in the graph.

The Jets lost to Denver. To San Diego. To Seattle. To Oakland. And they’re kept from those losses, all because of Miami. And it’s got to be a fluke that Tennessee hasn’t beatlooped them away yet – Tennessee and Indianapolis – they’re powerful!

But here’s the thing. I did some digging.

The NY Jets being so high isn’t entirely because they’re especially good. If you look at the results for the week, you’ll see that the AFC had three very key victories over the NFC this week, and all of them were damaging in the beatpath graph. So the shape of the graph is more about the NFC East and South collapsing than it is about the AFC South and North rising to prominence and making the NY Jets look amazing. If you look at the actual quality of victories of these top teams, it’s not really a picture of dominance. Houston beat Tennessee. All these teams are beatable.

And finally, the NY Jets play Miami this week, and Miami has a chance to avenge their loss and finally cancel out that horrible snarl of a beatloop that is keeping the NY Jets from all those beatlosses. But what happens to the NY Jets if Miami wins?

Answer: Nothing. They still retain their beatwin over Tennessee, and it’s because all those teams that beat the Jets, that Miami beat? Well, Buffalo beat four of the five, and the Jets beat them. Twice. The other one is beatlooped through Arizona. So it turns out this NYJ=>MIA problem might really just be a big distraction, because they’re all redundant beatloops. It’s true that Miami can help itself quite a bit with a win, regaining beatwins over all these teams, but the teams aren’t great and it wouldn’t improve Miami’s ranking very much. It might very well be that the NY Jets have just had a few flukey losses this season that are making them look worse than they really are.

Overall, I’m fascinating about what will happen in the final week and in the playoffs. This might actually prove to be a case where Beatpaths is telling us something that we just are having trouble accepting – it doesn’t mean we’re right, though!

4 Responses to 2008 NFL Week 16 Beatpath Graph

  1. doktarr says:

    “So it turns out this NYJ=>MIA problem might really just be a big distraction, because they’re all redundant beatloops.”

    This strikes me as the algorithm driving your thinking, as oppose to your thinking driving the algorithm. I’m sorry, but there’s just no way that a team with one win over any team better than “pretty good” should be in the top spot despite five losses to mediocre-to-poor teams. This is a quirk; an artifact left by the way that we remove beatloops.

    The only reason the Jets keep the win over the Titans is because the schedule happened to land all their losses in smaller beatloops than the Tennessee win. That’s not a sign of hidden strength, it’s just a sign that a team with an odd pattern of wins and losses can occasionally benefit from our beatloop removal system.

    This is not really an issue of parallel versus iterative, at the moment. Iterative is leaving this win in as well. Again, just a quirk of the Jet’s schedule, that they can completely exhaust four of their wins to loop away 80% of their losses before the Tennessee game comes into play.

    Now, iterative is a little more methodical about matching wins and losses, and as such a Miami win on Sunday DOES have a big impact. I believe the Jets would be left with no beatlosses, but after 7 wins and 7 losses are loops away, their only remaining wins would be over KC and STL. (Ouch.) This would be the algorithm’s way of saying (correctly, IMO) that we don’t really know how good the Jets are, except that they are better than the really bad teams. But this ability to shift after the Miami game doesn’t mean I see the Jets’ current #1 position as “correct” in iterative. Intuitively, I think they should currently retain their win over Cincy and that’s it.

    The idea of the iterative approach was twofold:

    1) to treat all wins and losses equally – one win can only cancel one loss.
    2) to eliminate the most inconsistent wins/losses, allowing the most data to be retained.

    While these are being fulfilled in the strictest sense, intuitively we find the TEN game to be the most inconsistent. The question, algorithmically, is whether there’s an impartial, consistent way to choose loops that is better than “smallest loops first”, which will allow us to achieve more intuitive rankings.

    This is arguably related to the situation where the Broncos had that stratospheric ranking a few weeks ago. That case was a bit more defensible, though, because the Broncos actually had a bunch of strong wins to pair with even more profoundly head-scratching losses. There just wasn’t a good place to slot the Broncos – they looked wrong everywhere.

  2. Tom says:

    I’m not too bothered by the Jets’ losses, because they seem to be more an issue of geography than the team’s actual strength. All the bad teams they lost to are West Coast teams. There’s no other good reason that the Jets, who can beat quality teams located elsewhere in the country, should lose to the genuinely bad West Coast teams, except for geography. Arizona has the same problem playing East Coast teams.

    While losing because of jet-lag certainly reflects on a team’s maturity, does it make sense to remove the Jets from a top position because of a fluke of geography in their schedule?

  3. ThunderThumbs says:

    I don’t know… the thing about beatpaths is that it deliberately keeps us from judging any particular win or loss based off of how intuitive we think that win/loss is. The whole point of the system is to point out possible flaws in conventional wisdom. And the thing is, the Jets demolished Tennessee. I don’t think that’s evidence that the Jets are really good, it just is more reason to look at the rest of the teams as beatable. It’s not beyond the realm of possibility that the Jets should be favored above all these other teams in a head to head matchup. The Jets are inconsistent and can lose to bad teams, but there’s no way we’d actually rank Seattle ahead of the Jets all things considered.

    And honestly I think that, despite the post-week-17 differences, if Iterative still has the Jets ahead of Tennessee, it just underscores that the Jets’ placement isn’t as wholly ridiculous as it would intuitively seem – or at least, that neither Iterative nor parallel is really the answer here.

    The NY Jets had opportunity of being in a three-team beatloop with Tennessee by losing to Cincinnati, but they beat Cincinnati. Now, there was only the *one* opportunity for the two teams to be in a three-team beatloop. So there’s an inequitability here that I think the system assumes isn’t there.

    One other possibility is to look at – does Tennessee have beatpaths to the teams that have beaten the Jets? There’s “probability of beatloops” there. And, do the Jets have beatpaths (aside from the Tennessee-spawned beatpath) to the other teams that have beaten Tennessee? (Houston?) This is quite likely problematic, but it might be another way to identify flukes.

  4. mm says:

    “I’m not too bothered by the Jets’ losses, because they seem to be more an issue of geography than the team’s actual strength. All the bad teams they lost to are West Coast teams. There’s no other good reason that the Jets, who can beat quality teams located elsewhere in the country, should lose to the genuinely bad West Coast teams, except for geography. Arizona has the same problem playing East Coast teams.”

    If playing on the west coast was that difficult, the 2 western divisions would have several more home wins. I think the evidence points to the two West divisions being weak this year, with the AFC East being overrated in our minds because they get to play 8 games versus those teams (and 6 games against each other).

    The NFC North and South are each slightly overrated in our minds (though not as much as the AFC East is) because they each got to play one of those divisions. The NFC North, AFC North, and AFC South are each slightly underrated because of the limited games they’ve gotten playing those divisions.

    This gets to one of the problems with Beatpaths and the current NFL schedule (and that Tennessee-Jets game). The AFC East teams have 12 games of (head to head) data playing each other too sort them out, and 16 games of data against the AFC West to sort out how they compare to them, but only 4 games of (head to head) data against the AFC South and 4 against the North.

    This would not be as much of a problem if the good teams were more or less equally spread across the league. However, with the Western divisions significantly inferior to the other divisions, those 8 games of head to head are drastically magnified in value, as otherwise we can only say the AFC East is clearly above the 2 western divisions.

    What’s the only AFC East that went 2-0 in their games vs. the North and South (including a win against the #1 seed in the AFC that looks flukish)? Why, the Jets of course!! Most of their other 14 games will find themselves swallowed up in beatloops, leaving that 2-0 record determining their placement. Note, New England and Miami have a better record in common games (the other 14 each played), yet they went 0-2 in non-common games. New England and Miami are getting rewarded in real life with a better chance to win the division, but the Jets get rewarded with a better beatpaths ranking!

    I’d like to suggest that you guys try recalculating the beatpaths (in each method) by removing all the strength of schedule games and see what it looks like (perhaps in the offseason, when there’s more time). Removing data generally isn’t a great idea, and each team be left with only 2nd order links connecting it to 5 other divisions, but we’ll be removing the links that are most likely to produce flukes that beatpaths can’t correct. The rankings of the teams in comparison to their own division, and to the 2 divisions they played, however, might be more accurate.

Leave a Reply

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