Only three upsets this week – two lead to new beatpath segments, and one leads to a CHI=>PHI=>DET=>CHI beatloop. Detroit beating Tampa Bay has the biggest effect on the graph and rankings, and it made IND@JAC less of a big game than it appeared. Buffalo beating Baltimore was not an upset.
And, the rankings:
| Rank | Team | Notes | Last Week | BeatPower |
1 |
 |
(Bye) Green Bay claims the top spot in the rankings this week. Remember that Green Bay has direct beatwins over both the Giants and Washington, which is making them more highly ranked than New England. Furthermore, if they split with Chicago later this season, their beatwin over San Diego could re-emerge. |
2 |
96.4(26/28 – 0/28) |
2 |
 |
(Beat JAC) Indianapolis beats a somewhat tough Jacksonville team, but the entire rest of the AFC South is looking a little weaker this week, since Tampa Bay lost to Detroit. When Indianapolis is put head-to-head against New England, however, Indianapolis’ opponents still look stronger. |
1 |
100.0(21/21 – 0/21) |
3 |
 |
(Beat MIA) No great reward for beating Miami, but yes, they do look scary scary good. Hosting Washington next week gives them a good chance to possibly move up some more. |
3 |
100.0(29/29 – 0/29) |
4 |
 |
(Beat MIN) The NFC East is again moving up this week, mostly because the top three NFC East teams have a beatpath to Detroit, which now has a beatpath to most of the AFC South thanks to its victory over Tampa Bay. |
7 |
96.4(27/28 – 1/28) |
5 |
 |
(Beat SF) Seems I had just read an article about pass rush being overrated in terms of contributing to team strength. Put an asterisk by that if the pass rush is as frightening as the Giants’. |
8 |
89.3(25/28 – 3/28) |
6 |
 |
(Beat ARI) Another beneficiary of Detroit defeating Tampa Bay. They’re the next team that gets to burn themselves against New England. |
9 |
85.7(24/28 – 4/28) |
7 |
 |
(Beat TB) Detroit is the biggest upward mover in this week’s rankings. They lose credit for defeating Chicago, but this week’s beatpath more than makes up for it. |
16 |
79.3(22/29 – 5/29) |
8 |
 |
(Lost to DET)Tampa Bay is pushing up, Detroit is pushing down, so they’re ranked right next to each other. I didn’t see the game – did it seem like Detroit was playing beyond its abilities, or Tampa Bay playing down… or are they just both two very good teams, with Detroit slightly better? |
4 |
74.1(20/27 – 7/27) |
9 |
 |
(Bye) San Diego rises on the bye, because their victory over Denver looks slightly higher quality now. |
12 |
85.3(13/17 – 1/17) |
10 |
 |
(Beat HOU) Tennessee slides on the win, because their beatloss to Tampa Bay looks more damaging now. |
5 |
69.2(18/26 – 8/26) |
11 |
 |
(Lost to IND) Jacksonville’s slide is entirely because of Detroit beating Tampa Bay – no big penalty for losing to Indianapolis. |
6 |
64.0(16/25 – 9/25) |
12 |
 |
(Beat PIT) The Broncos now have a beatpath to Pittsburgh, and rise a few slots because of it. They don’t rise as much as you’d think, though – Pittsburgh’s other victories weren’t high quality. Denver has a great opportunity to mess up the graph next Monday night. |
17 |
52.2(12/23 – 11/23) |
13 |
 |
(Lost to TEN) Most of the DET->TB effects are taken care of by now – Houston holds steady after the expected loss. |
13 |
50.0(10/22 – 10/22) |
14 |
 |
(Lost to DEN) Pittsburgh falls after losing to Denver. They’re still the best team in the bad AFC North, though. If we were to pick out AFC playoff seedings, we’d have Tennessee and Denver as the wildcard teams, screwing Jacksonville and Houston. |
10 |
45.5(10/22 – 12/22) |
15 |
 |
(Beat OAK) I got bored just looking at the scoreboard of this game. Kansas City climbs three after defeating Oakland. |
18 |
45.7(9/23 – 11/23) |
16 |
 |
(Beat PHI) Chicago manages to shed a damaging beatloss to Detroit by defeating Philadelphia. Although, the beatloss wouldn’t have been so damaging anyway given Detroit’s win over Tampa Bay. The real victim here is Philadelphia, though – they got thumped in the rankings this week. As for Griese, how many quarterbacks call their own signals, beyond audibles? It seems like it could be such an advantage. |
23 |
37.5(0/8 – 2/8) |
17 |
 |
(Bye) The NFC South falls slightly due to Tampa Bay’s loss. |
14 |
34.6(4/13 – 8/13) |
18 |
 |
(Lost to KC) Oakland also rises slightly after losing to Kansas City, mostly due to other teams looking slightly worse. |
20 |
36.4(8/22 – 14/22) |
19 |
 |
(Beat STL) Seattle’s beatloss to Pittsburgh looks more damaging now. The entire AFC West is ranked ahead of the entire NFC West. Speaking of teams being screwed, the NFC playoff seedings would have the Giants and Detroit as wildcards, screwing Washington, Chicago and Carolina. |
15 |
35.7(7/21 – 13/21) |
20 |
 |
(Bye) Cleveland rises slightly on the bye, mostly because of other teams looking slightly worse. Cleveland was already held down by Oakland, so they weren’t really negatively affected by their beatloss to Pittsburgh. However, Pittsburgh’s loss damaged other teams they have a beatpath to, which makes Cleveland look better in comparison. |
24 |
30.4(7/23 – 16/23) |
21 |
 |
(Beat ATL) No great reward for defeating Atlanta. |
19 |
12.5(1/12 – 10/12) |
22 |
 |
(Lost to DAL) Minnesota holds about steady on their expected loss to Dallas. I still haven’t had the pleasure of seeing Minnesota’s quarterback on tv. But I’ve read quite a few mentions this week about how stunningly horrible he looked. Do they even have a second-stringer? |
21 |
22.2(1/9 – 6/9) |
23 |
 |
(Lost to CHI) Philadelphia had the largest downward movement in this week’s rankings. They lost credit for defeating Detroit, which really hurt since Detroit looks so much better this week. But they really had no other quality wins to keep them afloat. My sense of this team was always that it’s a good team that has a bad ethic and plays down to its opponents. But maybe they’re just really a bad team. |
11 |
33.3(2/9 – 5/9) |
24 |
 |
(Beat BAL) Beating Baltimore was the expected result, so Buffalo doesn’t get a lot of help there. But their beatloss to Pittsburgh hurts a little more now since Pittsburgh has a beatloss to Denver. Buffalo falls a couple despite winning. |
22 |
31.6(6/19 – 13/19) |
25 |
 |
(Beat NYJ) No great reward for defeating the Jets. There really is a lot of bottomfeeding going on this year it seems. We’re due for some crazy upsets and long beatloops to flatten the graph. |
26 |
25.0(6/24 – 18/24) |
26 |
 |
(Lost to NO) Atlanta holds about steady on their loss to New Orleans. I like Leftwich as a quarterback here, though – I don’t recall his injury situation but if he can keep playing, I like Atlanta’s chances to improve. |
25 |
3.6(0/14 – 13/14) |
27 |
 |
(Lost to BUF) Baltimore holds steady on the expected loss. This was one of those outcomes that reinforces the occasional value of Beatpaths to me. The conventional wisdom was that Baltimore was a very good team thanks to their 4-2 record. I saw AP headlines about Buffalo “upsetting” Baltimore. But Beatpaths saw where they roughly slotted and showed that the conventional wisdom wasn’t necessarily true. Again, not a guarantee, never a guarantee, but it does give some valuable perspective at times. |
27 |
20.0(5/25 – 20/25) |
28 |
 |
(Lost to CIN) I’d say that there’s no big penalty for losing to Cincinnati, but it’s really more that there just isn’t much further to fall. The NFC West is just awful, and the Jets can’t go lower than Miami. |
28 |
4.3(1/23 – 22/23) |
29 |
 |
(Lost to NYG) San Francisco is pretty tamped down by their beatloss to Baltimore. |
29 |
6.5(1/23 – 21/23) |
30 |
 |
(Lost to WAS) These bottom four teams all pretty much played true to form, so there’s no real change in rankings. |
30 |
6.2(1/24 – 22/24) |
31 |
 |
(Lost to NE) A tough year for Miami gets even tougher with the loss of Ronnie Brown for the year. |
31 |
0.0(0/23 – 23/23) |
32 |
 |
(Lost to SEA) But at least Miami has hope…? I don’t get a sense that the Rams have much of a vision for the future here. Any good young players or evidence of a new generation with this team? |
32 |
0.0(0/24 – 24/24) |
Your tiebreaker is a bit screwy. If New England had beaten Dallas and San Diego, and tied every other game, they would be ranked #1. (If I understand your system)
I don’t think you meant to punish teams for winning games. It might need a rethink, maybe?
Are you using your beatfluke system yet? I ran my own version of your system and ended up with a similar graph, but two minor differences. I found these nine beatloops:
NO => SEA => TB => NO
CAR => ATL => HOU => CAR
MIN => CHI => KC => MIN
MIN => CHI => GB => MIN
KC => SD => CHI => KC
GB => SD => CHI => GB
PHI => DET => CHI => PHI
PIT => SF => ARI => PIT
SEA => SF => ARI => SEA
But because of TB => TEN => JAX => DEN => PIT => SEA the SEA => TB is considered a beatfluke which restored NO => SEA and TB => NO. This adds a line from NO to SEA on the graph for me.
Also because of the many alternate paths PIT and SEA have to ARI, ARI => PIT and ARI => SEA are considered flukes and PIT => SF, SF => ARI, and SEA => SF are restored. The only change this makes to the graph is that now instead of BAL splitting off to SF and ARI, the order goes specifically BAL => SF => ARI => STL on the graph.
Could we get a list of current beatloops being cast aside?
Poor Arizona. They are really much better than their placing. I expect them to rise eventually.
Random hypothetical: say the Broncos beat the Packers next week. Would that be immediately beatlooped (via the Chargers), or would that lead to GB=>SD beaing beatfluked, thereby causing CHI=>GB to emerge? My first instinct was the latter, but that would lead to a DEN=>GB=>NYG=>WAS=>DET=>TB=>TEN=>JAC=>DEN beatloop. I’m not sure how that would get broken. That DET=>TB game really screws with things.
Doktarr:
See my post(s) above for the beatloops. Also, if DEN defeats GB, I believe it would be considered a beatfluke itself considering that GB has a long and alternate route to DEN, and nothing would change.
Thanks. My proposed method of throwing out beatwins that appear in multiple beatloops would preserve some of those wins, notably KC and GB’s wins over MIN and SD, ARI’s wins over PIT and SEA, and PIT and SEA’s wins over SF.
Regarding Denver – yup, it would just lead to Denver shedding its beatloss to San Diego, but Denver would still be under Jacksonville’s thumb. It would also be considered a beatfluke (the official graphs aren’t using beatflukes this year though).
Yes, Jacksonville is really what is keeping Denver down. There is hope yet for you though. All other things staying the same, if DEN beats GB this week and DET next week it looks like DEN will shed both beatlosses, and the DET => TB => TEN will be removed on top of that. The apparent NFC dominance may not last much longer.
On a side note, I’ve looked at another way to rank the teams using this system. It counts all possible paths for each team where it is either the starting point or the ending point and subracts the latter from the former for a total “Path Score”. The ranks come up very similar and are a good measure of the stability of each team’s position. The top 3 are the same, but in a different order that you have with scores IND = 914, GB = 694, and NE = 555. I think this is a good representation of how IND has shored up it’s spots with multiple victories against the teams directly below them, whereas NE can fall a long way if it loses it’s beatwin over DAL.
STL is at the bottom with a -1465 score.
I’ll chime in with my usual set of responses to TT’s questions about the NFC
:
- I saw bits and pieces of DET-TB…the impression that I got was that the Bucs were making some big mistakes (blocked punt, Garcia fumbles in the red zone) that the Lions took advantage of. In other words, the Bucs gave the game away, but I don’t know that they would have won if they hadn’t made those mistakes. Personally, I don’t think much of the Lions, and I think the Bucs are decent, but that’s subjective, and the truth is that they’re probably both only mediocre at best.
- On Griese: Keep in mind that he was playing against a pretty weak prevent defense on that last drive–the Eagles left the sidelines free up to 8 yards or so, and the middle of the field about 15 yards. So there was plenty of room for him to work with. Still, I’m happy with the results
- Tavaris Jackson (MIN QB) is awful. Does not belong in the league, much less starting. Their backup (IIRC) is Kelly Holcomb, late of the Bills and Browns. He’s at least somewhat competent.
- The problem with PHI–especially this season–is that they have absolutely no playmaking ability on offense. Westbrook is their only real threat, and even he can be contained. Also, the O-line seems to have regressed this year, and the D has slipped somewhat, too.
Their worst games have been with Westbrook out. But even with him, they basically have to dink and dunk their way down the field, which is A.) Hard and B.) means they stall in the red zone. So they end up with few points and end up losing games due to attrition. They could be a respectable team again, if they could force teams to respect them downfield (*cough* Terrell Owens *cough* Donte Stallworth *cough*).
-Re: Rams–Their stars are Bulger (QB, 30), Holt (WR, 31), and Steven Jackson (RB, 24). So, that core should be good for a few years more. Other key players (Pace, LT, 32 but has been in the league for 11, and Little, DE, 33) are older and might have to be replaced soon. There’s a lot of talent on this team, actually, but injuries and awful awful awful OL play has basically killed them.
that’s interesting about the possible-number-of-pathways. Kind of an extension of beatpower, which someone else suggested. Beatpower more just has to do with the number of teams.
re: #1 – no, they wouldn’t be ranked #1 in that case. But you’re right about a large number of ties leading to some seemingly strange results. But it’s not inconsistent with the aims of the site – to represent things with only wins and losses. A tie just means it’s unclear who’s better than who. At any rate, I have a hard time worrying about that particular scenario since… when’s the last time we’ve even had a tie in the NFL?
Looks like MOOSE might be able to answer some of the questions I don’t get to in time!
MOOSE, maybe you’d want to volunteer to write for the site if you get some interesting ideas.
doktarr, keep reminding me, I want to take a closer look at what you’re suggesting.
re: #1, 11 – I guess I was wrong that it was the *average* strength of the teams beaten that mattered.
MOOSE’s comment (#9) about stability of a team’s position sounds about right. New England would fall farther if you removed its best beatwin, so it’s not unreasonable to say it is therefore “weaker” – until new evidence comes in.
Interesting as always.
I’d say that the primary problem with this total beatpaths idea is that the numbers are too close at the beginning of the season and get ridiculous by the end of the season. The amount of paths for the top and bottom teams rise exponentially. Last week the top three teams were GB (259), NE (204), and IND (171), while STL only had -379.
I’m going to see if I can find a formula I like that will keep the numbers in check while still giving a decent spread.
As far as writing anything, I’ll drop my ideas here and if there’s anything worth expanding on we can see about it. Right now I’m not fully comfortable that my version of the program is bugproof. I’m matching this season’s results well, but I come out with somewhat different graphs for last year’s data. My program also doesn’t automatically graph anything either yet, so I have to draw my graphs manually with the output.
Here was the specifics of my idea again:
- The graph becomes a weighted graph. If a team sweeps another team on the season, that link gets a double weight.
- If there are beatloops, follow the following approach:
* Start with beatloops of size (N=)3.
** Look at every link that is in a beatloop (of size N), and count how many beatloops it is in.
** Find the link with the lowest ratio of (link weight)/(number of beatloops containing that link). This lowest ratio becomes the value you will deduct from each loop.
** for every loop, subtract that ratio from all links in that beatloop. So a link that is in 2 loops gets that number deducted from its link weight twice.
** remove all zero-weight links.
* If there are still N-team loops, repeat the process for loops of size N. after all the 3-team loops are resolved, address all remaining 4-team loops the same way, and so on.
Example: take the current set of 3-team loops.
NO => SEA => TB => NO
CAR => ATL => HOU => CAR
MIN => CHI => KC => MIN
MIN => CHI => GB => MIN
KC => SD => CHI => KC
GB => SD => CHI => GB
PHI => DET => CHI => PHI
PIT => SF => ARI => PIT
SEA => SF => ARI => SEA
Several links have a ratio of 1/2, so decrement every link of each loop by 1/2. This leaves us with
KC => MIN
GB => MIN
KC => SD
GB => SD
ARI => PIT => SF
ARI => SEA => SF
NO => SEA => TB => NO
CAR => ATL => HOU => CAR
PHI => DET => CHI => PHI
The bottom three loops are then completely wiped out, as all links have equal weight are are in the same number of loops. So the preserved victories (with weight=.5) are
KC => MIN
GB => MIN
KC => SD
GB => SD
ARI => PIT => SF
ARI => SEA => SF
When we move on to 4-team beatloops,
ARI=>SEA=>CIN=>BAL=>ARI
… will get broken at ARI=>SEA, because this is already at half weight from the previous stage. So SEA=>CIN=>BAL=>ARI is preserved.
Make sense?
As an aside, a very obvious power ranking that arises from this is (total link weight below) – (total link weight above).
13: Does the fact that the number ranges vary so wildly from week to week really matter? It’s just a relative measure within that week, after all. It doesn’t seem that important to me that the top score could go from 600 one week to 1200 the next.
If you were so concerned and really wanted to compare weeks, you could always normalize the numbers in some way…but it doesn’t seem interesting/useful to me.
BTW, ThunderThumbs–could you repost/restate what the power rankings tiebreaker is, and maybe explain it, too? I couldn’t remember it–I know it’s not beatpower anymore, exactly–and I was trying to find it in comments, but didn’t see it.
As another wacky idea, has anyone considered trying to build the power rankings from the bottom up? I’d guess that would result in less information used at the top, which would probably be less interesting…just something that occurred to me.
I’d say that it matters as far as having people unfamiliar with the system being able to look at the data and see numbers that are worth comparing. By week 17 the number of paths will be so huge that distance between #1 and #2 could be several thousand, while in the middle of the pack, the distance between #14 and #18 could be under 100. It doesn’t mean that #1 and #2 are less close in terms of ability, and I’d like a scale that better represents the distance between the teams. Something that’s consistent from week to week would be a bonus. Maybe something along the lines of percentage of paths owned.
Unless I’m missing something, the numbers in your “beatpath power” wouldn’t be directly correlated with time–just with the relative “thinness” of the beatpaths graph. Right? And that graph’s thinness can go up and down over the course of a season.
The thing is, I don’t know that the number value is going to have any comparitive worth week-to-week, since the hole thing is related to structure of the graph. Although if there is a solution, I think your “total paths owned” seems like the most promising.
I’ve read through the rankings explanations a few times and still don’t see how the current rankings come from the system as it’s explained. Stop me where I’m misunderstanding: GB, IND and NE all have no beat losses, so they go to tiebreakers. The first tiebreaker on the “how do the rankings work” page linked is beatpower. Indy and NE have 100 and GB has 96.4. Shouldn’t Indy and NE be 1 and 2 then GB.
I love the idea behind the site and really want to understand the rankings, I’m sure I’m missing something, somewhere. Thanks.