If you came to this page looking for the latest insight into who should start week 8 in your fantasy league, I have to apologize, as that what I'll be discussion here has very little to do with that.
What this post is about is how I was enamored the other day when I was looking at the NFL Power Rankings posted on the ESPN website. And how, after flipping through the weeks a couple times, I drew the conclusion...
These guys have no idea what they are talking about...
I approach these rankings in a pretty simplistic way, the better of a team you are, the higher you go up on the rankings scale. Pretty simple concept, right? I also make the leap of logic that, more times then not, the team that is rated higher in the preceding week should probably win the following week if they play a lower ranked team. Sure, there will be upsets, that's why you play the game, but I figure if ESPN is going through the effort of recalculating these things like a Garmin every week, and since when the #1 power ranked Texans fell to the Packers in week 6 the Packers didn't just automatically become the #1 team, there are many factors that are weighed into this system, by experts, that should give you a good indication on who will win in the following week.
I put this logic to the test.
Drafting up a pretty simple excel spreadsheet, I input the players rankings for each week, then entered in the win/loss results from each week, drafted up a simple equation that determined who was higher ranked between the two teams, and tallied up how often the ranking systems 'predicted' the winner the following week.
The results were very interesting
After week 1, they where 8 for 16, or 50%. Fair enough, new season, getting a feel for who's out their etc. Week two the same, 8 for 16. And once again in week 3. So, after 3 weeks, the power rankings predictions are holding true at 24 of 48 or an even 50%. Then whatever makes these experts started to take hold, or maybe reducing the schedule to 14 games helped out, either way the Power Ranking record went to 10 and 4 in week 4, and 9-5 the following week. Week 6 was a step in the wrong direction as they went 3-11, wiping out most of the positive gains the prior 2 weeks. But rebounded with a strong week 7 at 11-2. Again raising the question if these ranking experts are just over burdened with the number of games played week to week.
In review, after seven weeks, using these power rankings, ESPN is touting a little better then random guessing 57-46 or 55.3% winning percentage.
Better then random guessing... how can we test this?
I came up with a hypothesis put together a random listing of teams and declare it the power rankings for the year, and see how it fairs against the 'experts' listing.
1. 49ers 2. Bears 3.Bengals 4.Bills 5. Broncos 6.Browns 7.Buccaneers 8.Cardinals 9.Chargers 10. Chiefs 11. Colts 12. Cowboys 13. Dolphins 14. Eagles 15. Falcons 16.Giants 17. Jaguars 18. Jets 19. Lions 20. Packers 21. Panthers 22. Patriots 23. Raiders 24. Rams 25. Ravens 26. Redskins 27. Saints 28. Seahawks 29. Steelers 30. Texans 31. Titans 32. Vikings
Not a bad listing, got some good teams at the top, and some bad ones on the bottom, but also the Browns making the playoffs well ahead of the Texans, along with a dozen or so other improbables, pretty random.
So how do the power rankings hold up in the random scenario?
a pair of 10-6 weeks to kick things off, followed by a 7-9 and then another 10-4, so after 4 weeks, the random selection was 3 games over .500 and 3 games ahead of the 'experts'. A bad week 5 at 4-10 put them just under the power rankings, but rebounded in week 6 with a 7-7 showing, leaving it at 48-42 after 6 weeks, which was 2 games ahead of the experts. But alas, week 7 was much kinder to the experts then the random's 6-7 record. Here is the full table;
| Week | Expert W | Expert L | % Correct | Random W | Random L | % Correct |
| 1 | 8 | 8 | 50.0% | 10* | 6 | 62.5% |
| 2 | 8 | 8 | 50.0% | 10* | 6 | 62.5% |
| 3 | 8* | 8 | 50.0% | 7 | 9 | 43.8% |
| 4 | 10* | 4 | 71.4% | 10* | 4 | 71.4% |
| 5 | 9* | 5 | 64.3% | 4 | 10 | 28.6% |
| 6 | 3 | 11 | 21.4% | 7* | 7 | 50.0% |
| 7 | 11* | 2 | 84.6% | 6 | 7 | 46.2% |
*=Weekly Winner (Tie week 4)
Leaving the expert Power Rankings with a record of 57-46, and the random one at 54-49 so far on the season.
I'll post some updates to this little experiment weekly, here are the predictions for week 8
Expert predictions based on ESPN power rankings;
Vikings over Buccaneers*
Chargers over Browns*
Bears over Panthers
Seahawks over Lions*
Packers over Jaguars*
Colts over Titans
Dolphins over Jets
Patriots over Rams
Steelers over Redskins*
Falcons over Eagles*
Raiders over Chiefs*
Giants over Cowboys*
Broncos over Saints
49ers over Cardinals
*Denotes that 'random' guess picked the other team to win
If you've read this far, and like what you see, don't be afraid to 'like' or 'G+' or 'tweet' this with your friends to encourage me to keep up with this little experiment through the football season. Quick buttons are on the right margin. Thanks Everyone!
This is awesome. I always wondered how something like this would work out.
ReplyDelete