Ranking Confidence – Week 15

Tom here with the “confidence” rankings of this week’s picks, based on comparing the BeatPower scores of each match-up:

(out of 100)
BeatPower comparison
(predicted winner – predicted loser)
Indianapolis-Detroit 76.1 76.1-0.0 CORRECT
NY Jets-Buffalo 74.8 97.9-23.1 CORRECT
New England-Oakland 51.3 60.7-9.4 CORRECT
Washington-Cincinnati 46.2 50.0-3.8 WRONG
Tennessee-Houston 45.7 95.7-50.0 WRONG
Philadelphia-Cleveland 31.2 73.3-42.1 CORRECT
Tampa Bay-Atlanta 29.4 90.9-61.5 WRONG
Minnesota-Arizona 27.1 77.1-50.0 CORRECT
Denver-Carolina 26.7 95.7-69.0 WRONG
Green Bay-Jacksonville 24.7 31.2-6.5 WRONG
San Diego-Kansas City 19.7 21.9-2.2 CORRECT
New Orleans-Chicago 16.3 54.5-38.2 WRONG
Pittsburgh-Baltimore 13.6 85.0-71.4 CORRECT
Dallas-NY Giants 7.6 92.0-84.4 CORRECT
Miami-San Francisco 4.1 34.1-30.0 CORRECT
St. Louis-Seattle 0.1 4.3-4.2 WRONG

One caveat about these picks: the Chicago win over New Orleans this past Thursday. When this data is included, Cleveland outranks Philadelphia in the normal Beatpath rankings. It’s not clear if the BeatPower scores would change enough to affect the confidence rankings of the picks above.

Edit: corrected the Dallas-NY Giants pick, which I had reversed. The picks got it right, my typo got it wrong.

UPDATE: Not a great showing this week for the “confidence” rankings, as the top half of the chart (the higher confidence picks) did only slightly better than the bottom half of the chart (the low confidence picks).

4 Responses to Ranking Confidence – Week 15

  1. doktarr says:

    iterative gets St. Louis-Seattle right, and Carolina-Denver right (although that’s based on power rankings as no beatpath exists). It got Minnesota-Arizona wrong, though.

  2. Tom says:

    I’ll be excited to see how the various methods compare at the end of the regular season. Maybe then we can weigh how various tiebreakers and rankings actually stack up against one another.

    I’ve been thinking about a loop-breaking method in which you simply break large loops by removing the oldest link(s) in the chain. Might be worth playing around with.

    Looks like the picks are getting hammered this week, although if Philly can manage to win, the confidence rankings will still do roughly twice as well in the top half of the graph than in the bottom half.

    I wonder if there are always a greater number of upsets once the wildcard race heats up?

  3. ThunderThumbs says:

    Breaking the oldest link. That’s a really interesting idea.

  4. doktarr says:

    With the iterative approach, you could just add an exponential decay function to all the link weights. I’ve wanted to look at what that does for a long time.

Leave a Reply

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