Wednesday, September 9, 2015

Weekly Playoff Probabilities - Week 1

When I wrote this post a few weeks ago detailing my methodology behind predicting playoff participants, I planned on updating those numbers every week.  So far, so good, as I have successfully completed the update following the games of Week 1.  Without further ado, here are the rankings:

Rank Team FPI Rank POFF Prob Change
1 Ohio State 1 73.264% 5.2%
2 Baylor 3 48.156% -0.9%
3 Notre Dame 10 37.224% 20.8%
4 Texas Christian 7 37.026% -8.6%
5 Florida State 14 21.243% 6.5%
6 Georgia 5 20.862% -6.6%
7 Michigan State 16 18.962% 2.4%
8 Southern California 8 18.571% 9.6%
9 UCLA 13 17.928% 1.3%
10 Mississippi 2 15.248% 8.6%
11 Oregon 17 12.095% -5.0%
12 Texas A&M 6 11.887% 1.8%
13 Louisiana State 4 11.511% -6.6%
14 Oklahoma 12 10.097% -0.4%
15 Tennessee 11 7.582% 0.2%
16 Clemson 22 5.321% -5.1%
17 Alabama 9 5.050% 1.4%
18 West Virginia 20 4.410% 3.5%
19 Wisconsin 35 2.407% -5.0%
20 Oklahoma State 24 1.833% -0.1%
21 Georgia Tech 21 1.736% 1.0%
22 North Carolina State 42 1.725% -0.4%
23 Florida 18 1.515% 1.0%
24 Miami (FL) 27 1.430% -0.1%
25 Arkansas 15 1.313% -1.6%
26 Boise State 37 1.245% -1.1%
27 Northwestern 46 0.987% 0.9%
28 Virginia Tech 28 0.888% -2.7%
29 Iowa 58 0.766% 0.2%
30 Duke 51 0.747% 0.6%
31 Auburn 19 0.650% -0.3%
32 Utah 34 0.644% 0.5%
33 California 25 0.592% 0.4%
34 North Carolina 33 0.540% -4.3%
35 Appalachian State 61 0.521% 0.3%
36 Arizona 41 0.468% -1.1%
37 Marshall 60 0.421% 0.1%
38 Stanford 23 0.346% -6.7%
39 Michigan 29 0.341% -2.1%
40 Missouri 32 0.339% -0.7%
41 Louisville 38 0.297% -0.4%
42 Temple 50 0.245% 0.2%
43 Illinois 49 0.236% 0.2%
44 Kansas State 40 0.175% -0.2%
45 Cincinnati 55 0.168% -0.3%
46 Arizona State 30 0.116% -1.5%
47 Mississippi State 26 0.094% 0.0%
48 South Carolina 31 0.077% 0.0%
49 Nebraska 44 0.075% -1.4%
50 Louisiana Tech 56 0.074% 0.0%
51 Pittsburgh 54 0.071% -0.7%
52 Penn State 43 0.070% -2.2%
53 Memphis 57 0.069% 0.1%
54 Kentucky 47 0.055% -0.1%
55 Ohio 74 0.044% 0.0%
56 Texas Tech 39 0.044% -0.1%
57 Minnesota 48 0.044% -0.1%
58 Brigham Young 45 0.028% 0.0%
59 Western Kentucky 59 0.026% 0.0%
60 Rutgers 69 0.021% 0.0%
61 Houston 75 0.018% 0.0%
62 Toledo 68 0.012% 0.0%
63 Texas 36 0.008% -0.5%
64 Maryland 65 0.008% 0.0%
65 Navy 81 0.005% 0.0%
66 Washington 52 0.005% 0.0%
67 Middle Tennessee State 67 0.005% 0.0%
68 San Jose State 72 0.003% 0.0%
69 Virginia 53 0.003% 0.0%
70 Purdue 66 0.003% 0.0%
71 Syracuse 70 0.002% 0.0%
72 Colorado State 95 0.002% 0.0%
73 San Diego State 79 0.002% 0.0%
74 Florida International 86 0.002% 0.0%
75 Indiana 76 0.001% 0.0%
76 Boston College 73 0.001% 0.0%
77 Wake Forest 85 0.001% 0.0%
78 Colorado 71 0.001% 0.0%
79 Oregon State 92 0.000% 0.0%
80 Iowa State 64 0.000% 0.0%
81 Washington State 87 0.000% 0.0%
82 Kansas 109 0.000% 0.0%
83 Vanderbilt 62 0.000% 0.0%
84 Utah State 84 0.000% 0.0%
85 East Carolina 94 0.000% 0.0%
86 Rice 88 0.000% 0.0%
87 Northern Illinois 83 0.000% 0.0%
88 Air Force 82 0.000% 0.0%
89 Nevada 107 0.000% 0.0%
90 Ball State 89 0.000% 0.0%
91 Fresno State 105 0.000% 0.0%
92 Hawaii 99 0.000% 0.0%
93 Old Dominion 106 0.000% 0.0%
94 South Alabama 128 0.000% 0.0%
95 New Mexico 93 0.000% 0.0%
96 Buffalo 108 0.000% 0.0%
97 North Texas 111 0.000% 0.0%
98 Massachusetts 97 0.000% 0.0%
99 Miami (OH) 117 0.000% 0.0%
100 South Florida 90 0.000% 0.0%
101 Tulsa 101 0.000% 0.0%
102 Connecticut 114 0.000% 0.0%
103 UNC Charlotte 119 0.000% 0.0%

 A couple of notes:

1. Everything is the same from the previous rankings.  I am still using ESPN's FPI as the input, and I am still using the 40% loss probability for championship games (which is probably too high for Ohio State, but oh well), and I am still using the same methodology to tie win-loss records to playoff likelihood.

2.  I spent a couple of paragraphs talking about the one final adjustment I had to make because the playoff teams only summed to 2.5.  I still had to make that adjustment this week, but now the raw number of playoff teams summed to 2.9.  I am guessing that this is because the results of Week 1 did enough to get FPI to separate out the great teams from the bad ones a little bit more.  #1 and #25 in the rankings are now separated by 15 points instead of 13.7, so there's evidence to back up that explanation.*

*Longer explanation: there are two sources of expected variance from preseason expectations.  One is randomness, which probably hasn't been reduced much after a weekend of FCS games.  The other is uncertainty about which teams are the best.  This uncertainty is probably lower due to the results of the first week of games, hence the slightly bigger spread in team ratings in FPI.

3.  I'll be subtracting teams from the ratings when they no longer "qualify."  That is, if a major conference team loses its third game, or any other team loses its first, I will eliminate them from contention.  As I noted in the preview post, no such teams has ever finished in the top four of the BCS or the playoff rankings.  If this changes, then so will I.  For now, we wave goodbye to the following: Central Florida, Georgia Southern, Arkansas State, Louisiana-Lafayette, Western Michigan, Central Michigan, Texas-El Paso, Texas State, Akron, Louisiana-Monroe, Bowling Green State, Wyoming, Florida Atlantic, Tulane, Texas-San Antonio, Southern Mississippi, Army, Kent State, Idaho, Troy, Nevada-Las Vegas, New Mexico State, Georgia State, Southern Methodist, Eastern Michigan.  Sorry guys.

4.  Ohio State moves up to 73% after a convincing win.  I was a little worried that they might approach or even top 100% because of the adjustment factor mentioned in the last paragraph.  I'm a little less worried now because of the spreading-out and because Ohio State looks so good that they would deserve it.

5.  Hey look, my team is the biggest mover of the week, more than doubling their playoff odds, and moving all the way up to third.  I guess quadrupling up a decent (?) team in yards is a good thing.

6.  TCU's lackluster effort makes them the biggest faller, although they still reside comfortably in the top four.  The next two biggest drops belong to Georgia and LSU.  As far as I can tell, the main reason is just subtle changes in the ratings of their respective SEC gauntlets.  It's going to be a fun year in the south.

7. Conference playoff likelihoods from this week remain about the same, with the only big change being Notre Dame and its weird friends overtaking the ACC (hey, isn't that Notre Dame's conference?):

Conference Exp Playoff Teams
Big 12 1.017
Big 10 0.972
SEC 0.762
Pac 12 0.508
Independent 0.373
ACC 0.340
Mountain West 0.013
CUSA 0.005
Sun Belt 0.005
American 0.005
MAC 0.001


Week 2 Preview

One bonus feature I plan on doing is finding the top ten games for the coming weekend via the same methodology I used in the preview post.  The one difference is that I am adjusting the formula slightly to account for the number of losses a team has.  This shouldn't change things too much, but it will likely favor games involving undefeated teams as the season progresses.

Home Away Home Win Prob       Playoff Teams Lost
Michigan State Oregon 59.8% 0.049
Tennessee Oklahoma 59.8% 0.030
Virginia Notre Dame 14.0% 0.017
Mississippi State LSU 23.4% 0.009
Vanderbilt Georgia 7.9% 0.005
Clemson Appalachain St. 87.3% 0.004
Ohio State Hawai'i 99.0% 0.002
Florida State South Florida 96.6% 0.002
BYU Boise State 49.7% 0.002
Baylor  Lamar 99.0% 0.002

Welcome to the only Week 2 preview featuring Lamar.  Feel free to replace that with Williams/Halep, should they both win tomorrow.  Or El Assico if they don't.  The inclusion of several likely blowouts indicates that Week 2 is pretty thin in quality, but it does still surpass Week 4 in terms of overall impact, due to the fantastic pair of games at the head of the list.

No comments:

Post a Comment