A couple of quickies. Last week’s picks (calculated from combined 2005/2006 results) were 11-5. Here are the graphs for the conclusion of Week 2.
First, the graph for the 2006 games only. Not all that valuable yet, but interesting. I like these early graphs before all the beatloops come into play, because you can pretty clearly see who has beaten whom. People are pretty high on San Diego right now, but this early in the season, it might be that the teams they’ve beaten are among the crappiest teams in the league.
Next up, the graph for the combined 2005-2006 games. I haven’t started pulling out games yet – after thinking about it, I really don’t see the point, because a team that is increasing in quality will eventually have that quality reflected in the graph, anyway. For those that have asked, there really isn’t a way to decrease the “strength” of wins by percentage or otherwise, because we actually really aren’t dealing with “numbers” here. It’s all relationships.
That’s one of the more interesting graphs I’ve seen. Very haphazard looking. Not entirely useful in terms of reflecting the quality of teams this year, but so far, the picks don’t appear to be of horrible quality. This next week will be really fascinating, though – so many of the matches are between teams that are very closely ranked in this week’s power rankings. Actually, this week appears to have more interesting and compelling games than I’ve seen in a very long time.
Wow! After two weeks of games, there is a 6 team path from Seattle to Kansas City. I’m not sure it means much, but it looks kinda cool.
Love this site and appreciate the work you put into it.
As a SD fan: is there possibly a sadder sight than the Raiders at this moment? It’s great to see how your system (even with just the two weeks of this season) can puncture the fools like Peter King that are high on SD’s chances this year.
Have you decided when you are going to eliminate last year’s results from this years’ graphs? Are you just going to keep producing two graphs?
Thanks again for the site!
Seems like there has to be a time when games stop counting. For an extreme exmple, early 1990s games shouldn’t be propping up the Cowboys beatpaths… and Oakland’s 2003 season success is long no longer a factor. Perhaps should add the 2004 season to last years graph and see how it looks — would philly be a middle of the pack team? Would carolina be near the top?
Miles, yeah I have been wondering the same thing about 2004/2005. I might check that out sometime this week.
A possible way to discredit old relationships is to have this years results dominate in loops. For example, last year Team A beat Team B, this year the reverse happens. Normally this would be a ‘no relationship’ loop, but you could make the new result dominate in these kind of small loops. I don’t know how easy this would be to implement.
I’ve been considering something similar – perhaps taking out all the divisional matchups from last year since we know they’re all going to be replayed this year. I think there are going to be a lot of approaches to closely examine after this coming week, because that is when the graph will get really interesting!
I suppose another idea would be to only go back about “a year” — for example, at this point, you could take out the first two weeks of games from 2005.
I hate to suggest extra work, but it almost sounds like people are suggesting 2 seperate beatpaths graphs. One for the current season, since standings are reset at the beginning of the year, so it’s helpful to see how teams stack up against each other during the season. Another graph could take into account some amount of data from past seasons, since there is some continuity for the teams from season to season (plus the graphs with more data are more fun to look at).
Heck, if you’ve got tons of extra time and computing power, it could be interesting to load up all the games over several years and see how the graph shakes out. There would be lots of loops elimated, and some sort of ranking would fall out. Just seeing how all the teams stack up since 1967 could be interesting. I’m not really requesting this since I’m sure it would be tons of extra work, but maybe you’ll mull it over and get bored next offseason.
Where are the Week 2 power rankings?
Power rankings are up now!
I, too, like that early season beatpath graph. Interesting that after two games, not only are there no beatloops (not surprising), but there’s only one game that is redundant (the IND win over HOU). Every team except those two has two arrows sticking out of it, corresponding to its two games.
As far as when to discount old results, I like something like Chris Smith’s suggestion. Keep all old results (maybe all results less than three years old – every team plays every other team over the course of three seasons) until eliminating that game can resolve a beatloop, at which point that score is permanently thrown out.
Wow, I didn’t actually know that the nfl actually structured it so that every team played every other within three years. Cool!
Actually, it’s been every *four* seasons since division realignment. Three for the teams in your own conference.
Yes every 4 years for teams from the other conference. Each division is rotated. there’s 4 games. PLus of course the home and away from your own division. There’s 6 more. In your own conference, you play each division once every three years. There’s 4 more. That compromise the 14 common games within a division as one of the tie breakers. For the final two games, each team is ranked first through fourth in its division. All the firsts in each conference play each other, all the seconds and so on. There are four firsts in each conference (ties are broken to rank each team) resulting in 3 games, one of which is already accounted for in the year’s rotating divisional foes. 16 games. Plenty of opportunity for beat loops to form.