Beatpaths was 9-5 last week, for a total record of 44-30. My “personal picks” that conflict with the Beatpaths picks were 0-1 due to Washington going away to defeat Philadelphia. So I’m 4-4 so far this year.
Detroit at MINNESOTA: Both teams are playing for a beatwin – Minnesota is favored.
CHICAGO at Atlanta: This match should verticalize the graph a bit, one way or the other. The teams are playing for a beatwin, and they have a couple of common opponents already. Chicago is favored.
Oakland at NEW ORLEANS: New Orleans is comfortably favored over Oakland – both teams are playing for a beatwin.
Cincinnati at NY JETS: The NY Jets are heavily favored in this game. I almost can’t tell from the graph, but it looks like both teams are playing for beatwins. If Cincinnati wins, crazy things happen to the graph – extremely vertical, and the entire AFC North gets ranked above the entire NFC West.
MIAMI at Houston: Two more teams playing for beatpaths – Miami is favored.
Carolina at TAMPA BAY: Lots of beatpath games this week! Here as well, and Tampa Bay is favored.
BALTIMORE at Indianapolis: Baltimore is favored in this game. I’ll personally pick Indianapolis due to home field advantage and “conventional wisdom”.
St. Louis at WASHINGTON: The most lopsided pick of the week. Washington has a beatpath to St. Louis, too.
Jacksonville at DENVER: Denver has a beatpath to Jacksonville – a very long beatloop could be created if Jacksonville wins in Denver.
DALLAS at Arizona: Both teams are playing for a beatpath, and they’re actually very closely ranked. Within two spots, and Arizona has home field advantage. Dallas is favored.
PHILADELPHIA at San Francisco: Both teams are playing for a beatpath, and Philadelphia is favored.
GREEN BAY at Seattle: Both teams are playing for a beatpath, and Green Bay is favored, although not by as much as you might expect. I will personally pick Seattle just because I think they’re due.
NEW ENGLAND at San Diego: New England is favored over Seattle – both teams are playing for a beatpath. But I think the undisciplined San Diego team will bounce back – I’ll personally pick San Diego.
NY GIANTS at Cleveland: The NY Giants are heavily favored and already have a beatpath to Cleveland.
The new DEN->TB->CHI->PHI->PIT->JAC->DEN beatloop is an interesting group of teams. It’ll be interesting to see which wins from that loop reappear by the end of the season.
Awful day picking yesterday due to a lot of high teams losing and low teams winning.
The Standard rankings match your picks perfectly and are 6-7 this week.
The Iterative method picked CAR over TB and is a ghastly 5-8.
Yet again, the Weighted method makes a case for itself, picking SD over NE and going 7-6.
All three have NYG at the top of their rankings and will be picking NYG tonight over CLE who ranks 29, 28, and 22 in the respective methods.
It might be interesting to look at power ranking volatility week by week. That is: take the difference between Week 5 and Week 4 rankings and sum them. Each week, the larger the differences, the larger the volatility. If changes in ranking are not as great, the rankings (and predictions) are more stable. If the beatpaths method is overall correct (save for fluke games), then stability of the rankings should increase over time. Here are the last three weeks (change in each team’s rank is summed):
Week 5
0+1+1+3+6+3+5+8+1+3+6+4+2+10+6+10+2+7+1+2+3+4+0+4+3+1+13+1+12+1+1+0 = 124
Week 4
0+1+1+2+5+11+8+0+7+1+12+17+1+4+10+9+4+5+5+1+3+8+12+1+5+5+0+0+7+4+1+2 = 152
Week 3
2+0+14+2+2+12+19+7+5+0+10+11+8+6+4+11+6+2+5+0+7+0+0+5+6+17+15+0+16+2+6+2 = 202
This limited sample seems to confirm that the ranking’s stability is increasing over time and that the volatility of the rankings is decreasing as the beatpaths method moves teams closer to their “actual” position in the pecking order.
It will be interesting to see if this trend continues when the power rankings for Week 6 are posted.
I think Tom’s idea is interesting. And although I can’t form a good argument off the top of my head for why, I know that most of the time in statistics you’ll see people square all those differences before summing them together.
Justin: Squaring the differences is how you form the statistical values of Variance and Standard Deviation, although that’s in a slightly different case than shown above.
I think we’ve known for a while that as the season progresses, the graph stabilizes. I think the only significant value of this exercise would be to see where the volatility is consistent from week to week as a point to say “this is the point in the season where BeatPaths should be making sense”. The other thing that might be interesting is comparing the stable volatility number for the 3 different graphing methods to see which is more consistent.