Ohio Football Topic
Topic: Week 11 OPPA++ Conference Rankings
Page: 1 of 1
The Situation
General User
Member Since: 7/13/2010
Location: Columbus, OH
Post Count: 957
mail
The Situation
mail
Posted: 11/5/2014 12:31 PM
RANK ---- CONFERENCE/TEAM

35 ---- SEC
1 ---- Mississippi State (8-0)
2 ---- Auburn (7-1)
3 ---- Mississippi (7-2)
4 ---- Alabama (7-1)
14 ---- Louisiana State (7-2)
17 ---- Georgia (6-2)
21 ---- Missouri (7-2)
40 ---- South Carolina (4-5)
43 ---- Florida (4-4)
49 ---- Arkansas (4-5)
51 ---- Tennessee (4-5)
59 ---- Texas A&M (6-3)
66 ---- Kentucky (5-4)
121 ---- Vanderbilt (3-6)
45 ---- Pac-12
6 ---- Oregon (8-1)
8 ---- UCLA (7-2)
16 ---- Utah (6-2)
19 ---- Southern California (6-3)
20 ---- Arizona State (7-1)
24 ---- Arizona (6-2)
56 ---- Stanford (5-4)
57 ---- California (5-4)
72 ---- Washington (6-3)
79 ---- Washington State (2-7)
86 ---- Oregon State (4-4)
98 ---- Colorado (2-7)
48 ---- ACC
5 ---- Florida State (8-0)
15 ---- Clemson (6-2)
18 ---- Georgia Tech (7-2)
31 ---- Boston College (6-3)
33 ---- Virginia Tech (4-5)
35 ---- Louisville (6-3)
36 ---- Miami (FL) (6-3)
45 ---- Duke (7-1)
49 ---- Virginia (4-5)
53 ---- Pittsburgh (4-5)
61 ---- North Carolina (4-5)
75 ---- North Carolina State (5-4)
99 ---- Syracuse (3-6)
113 ---- Wake Forest (2-6)
50 ---- Big 10
10 ---- Ohio State (7-1)
22 ---- Nebraska (8-1)
23 ---- Wisconsin (6-2)
25 ---- Penn State (4-4)
32 ---- Iowa (6-2)
34 ---- Michigan State (7-1)
38 ---- Minnesota (6-2)
47 ---- Maryland (6-3)
54 ---- Northwestern (3-5)
71 ---- Rutgers (5-4)
81 ---- Illinois (4-5)
84 ---- Purdue (3-6)
85 ---- Indiana (3-5)
88 ---- Michigan (4-5)
51 ---- Independent
7 ---- Notre Dame (7-1)
29 ---- Brigham Young (5-4)
74 ---- Navy (4-5)
95 ---- Army (2-6)
51 ---- Big 12
9 ---- Oklahoma (6-2)
12 ---- Kansas State (7-1)
13 ---- Texas Christian (7-1)
37 ---- West Virginia (6-3)
41 ---- Baylor (7-1)
65 ---- Texas (4-5)
67 ---- Iowa State (2-6)
69 ---- Oklahoma State (5-4)
94 ---- Texas Tech (3-6)
107 ---- Kansas (2-6)
71 ---- MWC
11 ---- Boise State (6-2)
26 ---- Colorado State (8-1)
27 ---- Nevada (6-3)
45 ---- Air Force (6-2)
47 ---- Utah State (6-3)
87 ---- Wyoming (4-5)
89 ---- New Mexico (3-5)
90 ---- Hawaii (2-7)
96 ---- San Jose State (3-5)
101 ---- San Diego State (4-4)
108 ---- Fresno State (3-6)
123 ---- Nevada-Las Vegas (2-7)
80 ---- C-USA
28 ---- Marshall (8-0)
42 ---- Louisiana Tech (6-3)
44 ---- Western Kentucky (3-5)
60 ---- Texas-San Antonio (2-6)
68 ---- Texas-El Paso (5-3)
77 ---- Rice (5-3)
78 ---- Middle Tennessee State (5-4)
91 ---- Florida Atlantic (3-6)
93 ---- Florida International (3-6)
105 ---- Alabama-Birmingham (5-4)
108 ---- Old Dominion (3-6)
120 ---- Southern Mississippi (3-6)
126 ---- North Texas (2-6)
83 ---- AAC
30 ---- Central Florida (5-3)
39 ---- Houston (5-3)
52 ---- Memphis (5-3)
62 ---- East Carolina (6-2)
73 ---- Temple (5-3)
82 ---- Cincinnati (5-3)
104 ---- Connecticut (2-6)
112 ---- Tulane (2-6)
117 ---- South Florida (3-6)
119 ---- Tulsa (1-7)
127 ---- Southern Methodist (0-7)
93 ---- MAC
55 ---- Toledo (5-3)
58 ---- Ball State (3-5)
64 ---- Western Michigan (6-3)
70 ---- Central Michigan (6-4)
80 ---- Bowling Green State (5-3)
96 ---- Northern Illinois (6-2)
100 ---- Akron (4-4)
106 ---- Massachusetts (2-7)
110 ---- Buffalo (3-5)
111 ---- Ohio (4-5)
114 ---- Miami (OH) (2-8)
116 ---- Kent State (1-7)
124 ---- Eastern Michigan (2-7)
100 ---- Sun Belt
62 ---- Louisiana-Lafayette (5-3)
76 ---- Georgia Southern (7-2)
83 ---- Arkansas State (5-3)
92 ---- Texas State (5-3)
101 ---- Louisiana-Monroe (3-5)
103 ---- Idaho (2-7)
115 ---- New Mexico State (2-7)
118 ---- Georgia State (1-8)
122 ---- South Alabama (5-3)
125 ---- Troy (1-8)
OhioCatFan
General User
Member Since: 12/20/2004
Location: Athens, OH
Post Count: 15,697
mail
OhioCatFan
mail
Posted: 11/5/2014 2:44 PM
Very interesting. Seems generally to be a reasonable alignment to me.
The Situation
General User
Member Since: 7/13/2010
Location: Columbus, OH
Post Count: 957
mail
The Situation
mail
Posted: 11/7/2014 8:05 AM
bump
PhiTau74
General User
PT74
Member Since: 8/6/2010
Location: Columbia, SC
Post Count: 458
person
mail
PhiTau74
mail
Posted: 11/7/2014 10:19 PM
Reverse the Ball St and N. Illinois rankings and the MAC makes sense.
The Situation
General User
Member Since: 7/13/2010
Location: Columbus, OH
Post Count: 957
mail
The Situation
mail
Posted: 11/8/2014 9:07 AM
PhiTau74 wrote:expand_more
Reverse the Ball St and N. Illinois rankings and the MAC makes sense.
I agree the outcome on Wednesday suggests the two should be switched.

If you strip the strong loss bonuses away from all teams,

-Ball State becomes #78
-NIU becomes #83
-OHIO becomes #99
-Buffalo becomes #119

However, Arkansas falls from #49 to #74. And South Carolina falls from #40 to #66.

It's not that I don't want to please everyone. The sport simply won't allow everyone to be pleased.

At the end of the day, NIU was (and still is) 0-1 against teams with winning records. We just don't know much about them.
The Situation
General User
Member Since: 7/13/2010
Location: Columbus, OH
Post Count: 957
mail
The Situation
mail
Posted: 11/9/2014 11:17 AM
Results:

- OPPA++ 30 of 47 correct (64%)
- Sagarin 33 of 47 correct (70%)
- Sixty-seven percent of our picks matched, which means I missed 7 he got correct and he missed 4 that I got correct

Additional comment:

Two (2) of my "misses" were from "upsets" where the home team underdog won but was ranked one spot behind their opponent. (OHIO beating UB and TCU beating Kansas State)

Every week it seems Jeff is able to eek out a few extra picks. I still can't help but wonder if the part of his methodology that remains secret includes a little manual override before the rankings hit the streets.

P.S.

The unrevised rankings model that started the week was 32 of 37 (68%). Regardless, I feel the most recent modifications made this week are more fair to teams top to bottom.
Last Edited: 11/9/2014 11:21:20 AM by The Situation
PhiTau74
General User
PT74
Member Since: 8/6/2010
Location: Columbia, SC
Post Count: 458
person
mail
PhiTau74
mail
Posted: 11/9/2014 1:46 PM
Looks like manual override sometimes makes sense. Two examples would be coming into this week having Ball State so high at 3-5 and N. Illinois at 6-2 so low. Same is true of Texas AM at 6-3 so low in the SEC at 6-3 then they beat Auburn. Obviously no one would have time to look at all the teams and schedules to pick out strange rankings but those two popped out before their wins this weekend.
Texas AM destroyed South Carolina at South Carolina so I would have definitely had them above USC. Texas A&M only lost to #3, 7 and 12 teams in the country yet this system had them ranked really low in the SEC.
Last Edited: 11/9/2014 2:04:24 PM by PhiTau74
L.C.
General User
LC
Member Since: 9/1/2005
Post Count: 10,584
person
mail
L.C.
mail
Posted: 11/9/2014 1:59 PM
Maybe the adjustment was reasonable, but you need to reduce the magnitude of it a little so that it had less effect in the top half of your ratings. I was a bit surprised when you revealed what you had decided as I expected a lot smaller adjustment given that your ratings were already working pretty well.
JSF
General User
Member Since: 1/29/2005
Location: Houston, TX
Post Count: 6,580
mail
JSF
mail
Posted: 11/9/2014 4:49 PM
PhiTau74 wrote:expand_more
Looks like manual override sometimes makes sense.
Manual overrides completely compromise the integrity of the system. I don't think he does it, though.
The Situation
General User
Member Since: 7/13/2010
Location: Columbus, OH
Post Count: 957
mail
The Situation
mail
Posted: 11/9/2014 8:15 PM
After all was said and done I actually like the second to last revision the best. The difference between the last two was the application of the strong loss bonuses. The final revision had them implemented by the same method top to bottom. However the second to last revision only had SLBs for Top 16 losses and the rankings seemed to be more palatable. I didn't like the Top 16 only SLBs at the time because it operated in a grey area of objectivity (by naturally excluding a majority of teams who won't even play a top 16 team all year). There's work to be done.

Interestingly enough, the magic number this week was 15%. Reducing the SLB in the final revision from 50% to 15% improved the correct pick percentage to 72% (34 out of 49; one more than Sagarin).

No tinkering will predict the Texas and Texas A&M wins this weekend. And manual overrides (for me) are not an option. For the reason JSF mentioned, comprising the integrity of the system invalidates the entire process for me.

The reason I continue to allege jerry rigging in other ranking systems are projections such as Michigan State as a Top 12 team before this weekend. I've looked at the data closely, and I just don't see a way anyone can put all the data through a machine and come out with Michigan State Top 12 (without bending the rules). Conference specific assumptions alone won't catapult MSU without bringing alone their less attractive Big 10 buddies, which would certainly draw a moan from the audience. I suspect these "experts" use team specific assumptions that unfairly skew results in the "eye's" favor.
The Situation
General User
Member Since: 7/13/2010
Location: Columbus, OH
Post Count: 957
mail
The Situation
mail
Posted: 11/10/2014 1:31 PM
L.C. wrote:expand_more
Maybe the adjustment was reasonable, but you need to reduce the magnitude of it a little so that it had less effect in the top half of your ratings. I was a bit surprised when you revealed what you had decided as I expected a lot smaller adjustment given that your ratings were already working pretty well.
So I did some more tinkering.

I think awarding 0-0.85 points (currently 0-0.77 pts) for wins against teams w/ under 0.500 records (to complete the continuum up to WVWTs) is merited. I don't think these points are really shaking up the top quartile, as those teams already have multiple WVWTs. So I'll keep this scoring mechanism in place.

The 50% SLB last week may have been too heavy. In general is the most subjective part of the system. I can tinker with this all day, to manipulate results. For example I can increase this past week's correct pick percentage to 74% by weighting SLBs in favor of Top 4, 8, and 16 losses and giving none for the rest. And of course there is no guarantee that method will improve results for a given week in the future.

The SLB is a necessary evil because when a team loses to another by one possession, most times the outcome can indicate relative strength. The choice is how much credit to give for demonstrating relative strength. I want to preserve the integrity of this system as much as possible while still including the SLB. The method though has to be applied the same, top to bottom. Therefore isolating Top 16 losses (though it may produce better results) is not fair.

I'll reduce the 50% to 25% this week. This gives an edge to teams that actually won vs teams that lost close (and teams like LSU will take a hit as a result despite their three top 4 losses, two by one possession).

I'd like to hear your thoughts L.C. when you get the chance.
L.C.
General User
LC
Member Since: 9/1/2005
Post Count: 10,584
person
mail
L.C.
mail
Posted: 11/10/2014 8:16 PM
It's fine to rig and tune your system all you like, using prior data to see what modifications worked the best....but....once you are done tinkering, then the model must be tested against different data. Thus you can tinker with it using last week's data, then test it using next week's results, and the week after.

As for looking into the model in more detail, I don't have time for it now. I will likely be very busy for at least a few more weeks.

BTW, how does your model work in the home field advantage? Should #111 Ohio beat #96 NIU since the game is at Ohio? Should #114 Miami beat Ohio with the game at Oxford?
Last Edited: 11/10/2014 8:20:21 PM by L.C.
The Situation
General User
Member Since: 7/13/2010
Location: Columbus, OH
Post Count: 957
mail
The Situation
mail
Posted: 11/11/2014 8:24 AM
L.C. wrote:expand_more
BTW, how does your model work in the home field advantage? Should #111 Ohio beat #96 NIU since the game is at Ohio? Should #114 Miami beat Ohio with the game at Oxford?
There is no consideration for home field advantage.

The purpose of this system is to rank teams. I only compare prediction results to Sagarin to establish credibility.

I've found that home team underdogs can typically upset a team within 50 ranking positions of themselves (mostly within 30 spots). The range of typical upsets is too large to switch teams for a given week (especially considering each team's relationship to the other team's they've played). The rankings would be a mess, and mostly a product of circumstance.

The matchups you've described are too close to call. Both teams in both instances are within striking distance of the other. I can tell you right now I personally won't touch a game between teams split within 15 ranking spots of each other. I can also tell you I hit my Pick 3s the last two weeks by picking only away teams.
Showing Messages: 1 - 13 of 13
MAC News Links



extra small (< 576px)
small (>= 576px)
medium (>= 768px)
large (>= 992px)
x-large (>= 1200px)
xx-large (>= 1400px)