College BlogPoll Ballot: Week 8 Draft

by Hooper on Oct 25, 2009 3:30 PM EDT Comment 57 comments

As always, the draft ballot, which is due by Monday morning, is simply the rankings as given by RTT's own computer system. The final ballot, which is due on Wednesday morning, will be a modified version based on the commentary received on the draft. So here are the top 25 in the draft. The full rankings will follow below, along with some brief observations.

Rank	Team	Delta

1	Florida
2	Alabama
3	Texas
4	TCU	2
5	Iowa	3
6	Penn State	7
7	Boise State	2
8	Southern Cal	1
9	Cincinnati	1
10	Oregon	2
11	Virginia Tech	7
12	Georgia Tech	8
13	Arizona	9
14	LSU	3
15	Pittsburgh
16	Utah	2
17	West Virginia	2
18	Miami (Florida)	3
19	Ohio State	8
20	Brigham Young	4
21	Oklahoma	3
22	South Carolina	3
23	Oklahoma State
24	Mississippi
25	Central Michigan
Last week's ballot

Dropped Out: Nebraska (#9), South Florida (#14), Wisconsin (#21), Kansas (#23).

The big table shows all of the numbers in their grand glory. If you wish to experiment, here is the data in Excel 2007
form, and in Excel 2003 form. If you're not one for the numbers, just scroll past them to the comments.

Rank	Team	WL	SOS	PED	RD	3DO	TD	PEO	OPPG	TO	3DD	SCORE
1	Florida	1.892	0.420	2.013	1.105	0.897	2.178	1.484	1.940	1.457	2.374	146.225
2	Alabama	1.892	0.805	1.908	1.792	-0.171	1.985	0.123	1.768	0.610	1.356	135.925
3	Texas	1.892	-0.002	1.477	2.338	0.678	2.076	0.632	1.464	1.117	2.575	134.382
4	TCU	1.892	0.388	1.172	1.320	0.382	1.887	1.612	1.583	0.940	1.667	129.117
5	Iowa	1.892	1.396	1.701	0.464	0.285	1.031	-0.120	1.300	-0.613	0.371	120.868
6	Penn St.	1.377	-0.479	1.751	1.448	1.975	1.995	0.900	2.115	0.937	1.258	116.479
7	Boise St.	1.892	-0.036	1.336	1.009	-0.409	1.287	2.154	1.325	1.071	1.412	111.137
8	Southern Cal.	1.303	1.473	0.881	1.446	-0.549	1.113	0.817	1.246	1.158	0.892	108.690
9	Cincinnati	1.892	-0.451	1.072	0.563	0.508	0.670	1.915	1.445	1.485	0.041	97.762
10	Oregon	1.303	1.501	1.697	0.543	-1.107	1.020	-0.168	1.028	-0.144	0.471	93.220
11	Virginia Tech	0.714	1.970	1.512	-0.185	-0.110	0.672	1.829	0.691	0.070	1.136	92.937
12	Georgia Tech	1.377	1.494	-0.981	0.349	1.716	0.022	1.539	0.034	0.806	-0.277	88.977
13	Arizona	0.714	1.897	0.302	0.950	0.863	0.716	0.352	-0.002	1.248	0.062	87.980
14	LSU	1.303	0.878	1.283	0.355	-0.110	0.903	0.479	1.424	-1.253	-0.363	84.657
15	Pittsburgh	1.377	-0.045	-0.310	0.930	1.116	0.729	1.766	0.850	0.419	0.307	77.913
16	Utah	1.303	0.253	1.491	-0.128	0.197	0.829	0.472	0.890	0.384	1.422	76.955
17	West Virginia	1.303	-0.471	0.708	1.297	1.005	0.584	0.828	0.393	0.806	0.604	74.488
18	Miami (FL)	0.714	1.458	-0.046	0.368	0.837	0.618	1.454	0.077	0.294	-0.054	73.878
19	Ohio St.	0.861	0.229	1.321	1.159	0.048	1.233	-0.021	1.525	-0.412	0.658	71.803
20	BYU	0.861	0.197	-0.742	0.878	2.861	0.135	1.679	-0.070	1.295	-0.780	68.711
21	Oklahoma	0.124	0.264	1.363	1.671	-0.394	1.574	0.382	1.940	0.772	0.652	60.872
22	South Carolina	0.861	0.512	1.113	-0.123	0.122	1.162	0.015	0.919	-0.176	0.281	60.270
23	Oklahoma St.	1.303	-0.856	0.572	1.003	0.387	0.077	1.679	0.572	0.746	0.902	59.974
24	Mississippi	0.714	-0.239	1.794	0.212	0.232	0.971	0.084	1.464	0.288	2.241	58.245
25	Central Mich.	1.377	-1.325	0.499	0.661	1.094	0.448	1.242	1.248	0.423	0.188	56.373
26	Nebraska	0.124	-0.064	1.453	0.926	0.728	1.547	0.091	1.761	0.095	0.695	53.877
27	Wisconsin	0.714	0.657	-0.565	0.642	1.094	0.525	0.159	-0.102	0.337	-0.277	48.701
28	Navy	0.861	-0.473	0.043	0.508	1.468	0.695	0.703	0.468	-0.225	1.393	48.380
29	Notre Dame	0.714	1.265	-0.549	0.355	-0.171	-0.905	1.616	-0.002	1.122	-0.415	46.908
30	Stanford	0.345	1.015	-0.380	0.369	0.744	-0.165	0.970	0.242	1.065	0.066	43.667
31	Clemson	0.124	1.428	1.028	0.282	-0.726	1.084	-0.782	0.790	-0.664	0.593	43.435
32	Tennessee	-0.465	1.163	1.323	0.659	-0.194	1.490	-0.053	0.790	0.165	0.846	41.791
33	California	0.714	0.790	-0.542	0.675	0.017	-0.471	0.369	0.254	0.973	0.495	40.185
34	Oregon St.	0.124	1.212	-0.211	0.606	0.965	-0.418	0.600	-0.220	0.445	0.055	39.223
35	Fresno St.	0.124	0.506	0.205	-0.125	1.252	0.048	0.657	0.295	1.400	1.053	35.313
36	Kansas	0.714	-0.768	-0.146	1.019	0.847	0.202	0.786	0.196	1.781	0.222	35.305
37	Georgia	0.124	1.604	-0.366	0.348	-0.043	-0.153	0.002	-0.498	-0.746	0.955	28.901
38	Idaho	0.861	-0.319	-1.088	-0.042	2.192	-0.765	1.666	-0.920	0.973	-2.721	27.372
39	South Fla.	0.714	-1.205	0.851	0.229	-0.019	0.721	1.065	0.929	0.059	0.386	27.251
40	Michigan St.	-0.170	0.786	-0.337	0.890	0.207	0.500	0.763	0.468	0.493	-0.466	26.907
41	Auburn	0.345	0.685	0.706	-0.733	0.122	-0.205	0.227	-0.383	0.986	0.244	26.202
42	Houston	1.303	-0.790	0.017	-1.775	1.613	-1.444	1.322	-0.280	2.890	-0.736	24.849
43	Arizona St.	0.124	-0.105	0.738	1.363	-1.148	1.314	-0.906	0.790	-0.321	0.846	20.955
44	Boston College	0.345	0.334	0.069	0.615	-0.810	0.420	-0.231	0.485	-0.847	1.314	19.705
45	Air Force	-0.170	-0.441	1.208	0.421	0.268	1.541	-0.709	1.509	-0.604	1.637	19.630
46	Texas Tech	0.345	-0.678	0.246	0.302	0.232	-0.190	1.499	0.052	2.024	-0.511	18.229
47	Kentucky	0.124	1.295	0.699	-0.737	-0.202	0.075	-1.296	0.156	-0.700	-0.777	17.014
48	Connecticut	0.124	-0.188	-0.047	0.355	0.215	0.574	0.366	0.632	0.268	0.005	14.715
49	Troy	0.714	-0.616	-0.514	0.444	0.153	-0.484	0.703	-0.498	1.506	0.321	13.634
50	La.-Monroe	0.124	-0.436	0.127	0.831	0.911	0.077	-0.308	-0.677	0.286	-0.869	9.033
51	Southern Miss.	0.345	-1.736	0.390	0.760	0.512	0.737	0.825	0.485	0.540	0.421	8.983
52	Kansas St.	0.345	-0.509	0.185	0.965	-0.942	0.500	-0.075	0.346	-0.354	0.533	8.976
53	Wake Forest	-0.170	0.384	-0.271	-0.320	0.822	-0.070	0.964	0.190	0.074	-0.914	7.962
54	North Carolina	0.124	-0.516	0.913	0.920	-1.356	1.549	-1.113	1.067	-1.562	1.127	7.881
55	Washington	-0.686	2.597	-1.051	-0.745	1.526	-1.146	-0.343	-0.781	-0.198	-0.070	6.992
56	Michigan	0.345	-0.340	0.133	0.172	-0.087	-0.184	-0.372	0.086	0.511	0.625	5.480
57	UCLA	-0.465	1.978	0.824	-0.396	-1.533	0.290	-1.517	0.215	-1.441	0.278	4.646
58	Missouri	0.124	-0.122	0.156	0.335	-0.394	0.185	0.002	0.116	-0.077	-0.474	4.257
59	Virginia	-0.465	0.882	1.546	-0.512	-0.943	0.702	-1.018	0.453	-1.567	0.989	3.890
60	Iowa St.	0.345	-0.276	-0.186	0.111	0.353	-0.463	-0.700	0.607	0.207	0.293	2.999
61	Florida St.	-0.465	1.518	-1.168	-0.922	1.015	-0.971	1.431	-0.538	0.929	-0.389	1.908
62	Temple	0.714	-1.569	0.060	1.042	-0.550	0.640	-0.341	0.591	-1.127	0.275	1.566
63	Texas A&M	0.124	-0.398	-0.505	-0.227	1.401	-0.942	0.565	-1.212	2.033	1.066	-0.925
64	Tulsa	0.124	-1.410	0.402	0.507	-0.363	0.427	1.214	0.632	0.414	0.320	-1.525
65	Northwestern	0.345	-1.834	-0.303	0.566	1.628	0.317	0.162	0.068	0.297	0.164	-2.475
66	East Carolina	0.124	-0.674	-0.324	0.715	0.870	0.077	-0.907	0.136	-0.834	-0.209	-4.510
67	Purdue	-0.686	0.446	0.801	-0.323	0.454	-0.073	-0.095	-0.383	0.448	-0.381	-5.195
68	Rutgers	0.714	-2.275	0.189	0.827	-0.469	0.966	-0.449	1.048	-0.507	0.732	-5.963
69	Marshall	0.345	-0.284	0.382	-0.760	0.072	-0.518	-0.801	0.433	-0.833	-0.896	-7.583
70	Nevada	0.124	-0.160	-2.178	0.662	1.306	-0.973	0.465	-1.014	2.332	-0.415	-8.357
71	Ohio	0.345	-1.198	1.129	-0.132	-0.858	0.227	-0.560	0.433	-1.097	0.661	-9.762
72	San Diego St.	-0.465	0.088	-0.005	-0.188	0.580	0.205	0.360	-0.519	-0.453	-1.287	-10.505
73	Arkansas	-0.465	2.017	-1.165	-0.287	-1.802	-1.079	0.812	-0.459	0.726	-0.054	-11.311
74	Northern Ill.	0.124	-1.021	-0.673	0.609	0.298	0.119	0.229	0.671	-0.579	-0.811	-11.609
75	UCF	0.124	-1.492	-0.470	1.158	0.170	0.366	0.196	0.691	-0.877	-0.246	-12.097
76	Duke	0.124	-1.123	-0.480	0.080	-0.030	0.232	0.880	-0.141	0.376	-0.383	-14.699
77	N.C. St.	-0.465	-0.805	-0.920	1.079	0.781	0.780	0.310	-0.419	0.263	0.044	-15.229
78	Minnesota	-0.170	1.255	-0.312	-0.823	-0.706	-0.872	-0.864	-0.140	-1.502	-2.011	-18.221
79	Mississippi St.	-0.686	0.595	0.272	0.002	-0.814	-0.013	-0.883	-0.070	-0.090	0.743	-19.259
80	Buffalo	-0.686	-0.665	0.235	-0.178	0.834	0.065	0.211	-0.209	0.626	-0.607	-23.632
81	Wyoming	0.124	-0.742	0.336	0.040	-0.980	-0.275	-1.492	0.077	-0.857	0.161	-27.137
82	Middle Tenn.	0.124	-1.059	0.182	-0.515	-1.226	-0.092	-0.409	-0.320	0.628	0.548	-31.601
83	Western Mich.	-0.170	-0.751	-0.106	-1.000	0.626	-1.080	0.011	-0.452	0.538	0.142	-32.919
84	SMU	-0.465	0.576	0.759	-0.757	-2.082	-0.626	-0.533	-1.053	-0.118	-0.777	-33.794
85	Indiana	-0.170	-0.655	-0.560	-0.083	-0.103	-0.615	-0.303	-0.330	-0.291	-0.822	-34.685
86	Bowling Green	-0.686	0.193	-0.112	-1.581	0.363	-0.538	0.020	-0.591	0.484	0.993	-34.959
87	Colorado St.	-0.686	0.893	-1.099	0.360	-1.285	-0.850	0.334	-0.868	-0.478	-2.696	-36.362
88	Baylor	-0.465	-0.179	0.308	-0.975	-0.686	-0.763	-0.250	-0.023	-0.371	-1.231	-39.532
89	Toledo	-0.170	-0.708	-0.021	-1.069	-0.070	-1.146	0.511	-1.960	1.153	-1.605	-41.505
90	Arkansas St.	-0.858	-0.802	-0.812	1.057	-0.412	0.437	-0.456	0.618	-0.579	0.220	-41.937
91	Louisiana Tech	-0.465	-1.312	0.326	-1.084	0.410	-0.062	-0.043	0.275	-0.373	-0.070	-42.901
92	Syracuse	-0.465	-0.316	-1.480	1.363	-2.082	0.033	-0.041	-0.439	-1.068	0.508	-44.832
93	Louisville	-1.055	1.094	-1.694	-0.456	-0.069	-0.317	-0.387	-0.775	-0.255	-0.431	-44.976
94	Colorado	-1.055	0.702	-0.725	-0.171	-0.229	-0.388	-1.649	-0.617	-1.338	0.902	-48.791
95	Kent St.	-0.170	-1.537	0.374	0.328	-2.181	0.047	-0.802	0.381	-1.009	0.329	-50.446
96	Hawaii	-1.055	0.152	-0.966	-1.322	1.104	-1.003	0.338	-1.251	0.857	-1.533	-53.727
97	Maryland	-1.202	0.253	-0.845	0.022	-0.250	-0.385	-0.392	-1.042	-0.910	0.307	-54.755
98	Vanderbilt	-1.202	-0.526	0.995	-0.505	-0.899	0.628	-2.335	0.884	-0.957	0.751	-56.466
99	La.-Lafayette	0.124	-1.530	-1.009	-0.879	0.572	-0.978	-0.639	-1.153	-0.561	-1.627	-58.304
100	Utah St.	-1.055	-0.075	-0.209	-1.276	-1.385	-0.976	0.450	-0.974	0.919	-0.216	-66.869
101	Fla. Atlantic	-0.858	-0.276	-1.085	-1.852	0.948	-1.540	0.319	-1.579	1.301	-0.893	-66.921
102	UNLV	-0.686	-0.742	-1.791	-1.107	1.899	-1.322	-0.211	-1.215	0.104	-2.139	-67.319
103	Army	-0.686	-1.796	0.326	0.615	-1.626	1.134	-2.242	0.207	-1.745	-0.030	-67.808
104	UTEP	-0.465	-0.079	-0.716	-1.659	-0.230	-2.178	-1.307	-1.272	-0.497	-1.056	-73.594
105	Illinois	-1.644	0.563	-0.636	-0.995	-0.236	-1.015	-1.349	-0.498	-0.412	-1.226	-79.462
106	Memphis	-1.055	-0.976	-0.356	-1.028	-0.621	-0.829	-0.534	-0.717	-0.252	0.625	-82.037
107	Miami (OH)	-2.233	1.178	-0.403	-0.806	-0.578	-0.233	-1.172	-1.423	-1.083	-0.869	-82.532
108	New Mexico St.	-0.686	-1.226	0.051	-1.026	-0.708	-0.220	-2.466	-0.538	-2.196	-0.630	-84.610
109	North Texas	-1.644	-0.486	0.012	-0.929	-0.125	-1.331	-0.363	-1.866	0.494	-0.246	-87.943
110	Akron	-1.644	-0.366	-0.377	-0.621	-1.001	0.082	-1.164	-0.320	-1.732	-1.104	-89.889
111	Tulane	-1.055	-0.522	-1.950	-1.233	0.646	-0.873	-0.449	-1.746	-0.877	-2.081	-90.922
112	UAB	-1.055	-0.218	-1.699	-0.089	-1.063	-1.769	-1.186	-1.291	-0.136	-1.554	-92.982
113	San Jose St.	-1.546	1.736	-0.802	-2.720	-1.822	-1.720	-0.714	-1.301	-1.615	-0.545	-93.703
114	Washington St.	-1.644	1.805	-1.630	-1.692	-1.916	-2.450	-1.112	-1.787	-1.485	0.525	-103.344
115	FIU	-1.644	-0.025	-0.806	-2.245	-0.774	-2.284	-0.732	-1.430	-0.988	0.124	-118.048
116	Ball St.	-1.718	-1.517	-1.437	-0.357	-0.649	-0.580	-2.094	-0.815	-1.059	-0.019	-129.098
117	Eastern Mich.	-2.233	-0.871	0.243	-3.071	-0.191	-0.733	-1.830	-1.330	-1.997	-0.707	-141.802
118	New Mexico	-2.233	-0.573	-1.471	-0.803	-2.190	-1.245	-1.006	-1.827	-1.462	-1.664	-150.612
119	Rice	-2.233	0.002	-2.459	-1.561	-1.166	-1.937	-1.874	-2.966	-1.387	-0.758	-162.957
120	Western Ky.	-2.233	-0.526	-3.123	-2.870	-1.575	-2.614	-1.441	-2.461	-1.560	-2.484	-199.613

Observations:

The TCU Horned Frogs like the top 5, thank you very much. I'm not saying I can't be persuaded again, but I'll need a pretty good argument this time. That was one very dominant performance against a BYU team that is actually quite good. Going forward, they only have Utah to worry about, as the rest of their schedule is against the bottom half of the MWC.
Penn State Nittany Lions at it again. I'm pretty sure everybody here will want PSU knocked down manually. Just be sure to include reasons for it - preferably (but not mandatorily) reasons that relate to the numbers.
Cincinnati Bearcats. I want 'em up there, too, especially after a solid game without Tony Pike. It's still a strength of schedule issue (which, by the way, is slowly eroding the Boise State Bronco's grip). With the Connecticut Huskies, West Virginia Mountaineers, and Pittsburgh Panthers on the schedule, Cincinnati is set to make a major push late.
USC Trojans. A passing efficiency of 0.817 is very good - especially for having a freshman quarterback. Somebody tell Charlie Weis that a pro-style offense can indeed be implemented with a new quarterback. Oh, right, they already did.
Dominance Redacted. The Texas Longhorns finally played a solid game or two, and now they are virtually neck and neck with the Alabama Crimson Tide. Alabama's the better team, but they've been emphasizing fundamentals over stats, and it's making a difference in the numerics.
The Arizona Wildcats are quietly having a great season. They've played a very respectable slate so far and are looking solid in the process It's been so long since anybody cared about the Wildcats that we're not used to thinking of them, but there they are. This is why I like numbers; I probably wouldn't give them the time of day if I was using my own intuition because I haven't heard a thing about them from the press wires. I wonder if the human voters act the same way.
The Florida Gator's Third Down Offense (3DO) has dropped like a stone. Last week, their score was 1.586. Now, they're 0.897. For most of the season, they've been well above 2. Since Tim Tebow's concussion, their third down conversion rate is nowhere near what it used to be. I'm guessing that Meyer is still holding Tebow back for the moment in anticipation of Atlanta and the bowl game. But he's rolling the dice.

Conference Rankings

Here are how the conferences stack up with a straight average of the teams:

The SEC is gaining a little more separation from the rest. With few nonconference games left, this will probably hold. The PAC-10 is a very solid second because (a) they have few bad teams, and (b) they have some surprisingly good teams, like Arizona, that the media chooses to completely ignore. The rest of the BCS conferences are pick-ems, and the Mountain West suffers with several genuinely bad teams dragging down the efforts of TCU, BYU, Utah, and Air Force. Don't get worked up about the 1-A independents; that's only Navy, Notre Dame, and Army, and Navy is the best of the three. I just include them for completeness.

Let us not speak of the Sun Belt, ok?

0 recs | Comment 57 comments | Add comment | Share

Story-email Email Printer Print

Comments

Head to head police:

Georgia Tech over Virginia Tech at 11/12.

And that’s a great point about ’Zona – I forgot one of their losses was to Iowa.

Eight weeks in, I really like the way these rankings have taken shape.

Rocky Top Talk

by Will on Oct 25, 2009 3:36 PM EDT reply actions 0 recs

Thanks for picking up on GaTech / VaTech.

In some cases, head to head matchups can’t justify switches, but it certainly can there. I’ll make sure and change that for the final.

by Hooper on Oct 25, 2009 3:44 PM EDT up reply actions 0 recs

I like TCU at 4

I think they’re legit

by golfballs03 on Oct 25, 2009 3:48 PM EDT reply actions 0 recs

Oh, for reference:

That 1 extra game that Penn State has played in comparison to Southern Cal is making a .074 difference in win/loss which, all else being equal, makes a 1.85 difference in their final scores. That’s not enough on its own to either move Penn State or USC, though it would put USC within about a point of Boise State.

Southern Cal’s tendency to not play games early has always done this to computer polls. Since the UCLA game is in championship week, they take two bye weeks compared to everybody else’s one. For that, I have no problem letting that 2-pointish penalty sit on them throughout the season.

by Hooper on Oct 25, 2009 4:04 PM EDT reply actions 0 recs

Hooper, your numbers are broken!

I’m seeing ERROR509 instead of each team’s overall score. That said, I can’t think of a whole lot of changes I would make. Your top 12 is identical to my MummePoll ballot. The only difference is in the top 5 tier, where I have Cincy instead of Iowa. A couple particular comments:

*Arizona has quietly been good, but I don’t think they’ve been #13 good. They still have 2 losses (which makes them the 2nd highest two-loss team in your poll), and their signature win is still Central Michigan. Contrast this to Miami, who has a win over a top 12 team and two losses to top 35 teams. I’d slide ’Zona in the 17/18 range.

*I told you that Tennessee would jump after the loss, although it was only one spot. Still, with a pair of top 25 teams left on the schedule, there’s room to push a top 25 spot if we can pull a nice little winning streak. Just for fun, I want to run the table without W/L and see how high we are. I bet it’s top 15.

*TCU is the #4 team. I don’t think anyone else has an argument at this point.

*I’d like a manual override to bump Cincy up to #5. They just look more impressive than Iowa and Penn State.

*On the subject of Penn State, they finally have an impressive win over a team with a pulse. The schedule strength is still pretty bad, but how much more impressive have Southern Cal, Boise, and Oregon been? My intuition still has PSU as #12, but they can be successfully defended as high as #7/8.

by Incipient_Senescence on Oct 25, 2009 4:28 PM EDT reply actions 0 recs

Huh.

It works just fine for me. I’m assuming you’re talking about downloading the files I linked. Let me email you directly via your SBN signup email and see if it works.

by Hooper on Oct 25, 2009 4:50 PM EDT up reply actions 0 recs

You were right on UT. I was overestimating the effect of the loss.

The thing keeping Arizona afloat is that massive SOS value. The numbers are Sagarin’s, so I’m not exactly sure how that’s buoying them so well compared to Miami, but it’s the biggest difference. (.4ish * 15 = 6ish).

The one major change I’d like to make to SOS is to find a way to apply it to the metrics directly rather than as an additive. For example, rather than just calling it the ‘10th hardest schedule’ or whatever, I’d like to adjust the offense/defense numbers for each game based on the strength of the opponent. Teams that run well against Alabama would get huge props on their offensive numbers, for example, while teams who rush well against Western Kentucky get their gains minimized.

I don’t know how to do that conveniently, so it’ll probably become an offseason project. But I think that would help some of the remaining squirreliness like Cincy and Arizona.

by Hooper on Oct 25, 2009 4:58 PM EDT up reply actions 0 recs

My bad, Hooper

Wasn’t talking about downloading the file but reading the chart. I’d accidentally used 2007 instead of 2003. The whole chart worked except for the overall score calculation.

I was thinking the same thing on applying SOS to metrics rather than it being its own category. If you can find a way, go for it. I really think it’d help the poll, but it’d be quite a task. And the poll is good as is, with just a couple exceptions (’Zona and Cincy, as mentioned).

BTW, I ran the numbers without W/L. Taken purely on the strength of the overall stats, regardless of who wins, Tennessee is ranked #11. I think that reflects accurately how strong this team could be if they learned to pull out the close games.

by Incipient_Senescence on Oct 25, 2009 5:14 PM EDT up reply actions 0 recs

heh.

in sync…

by Hooper on Oct 25, 2009 5:16 PM EDT up reply actions 0 recs

Ah, yeah

It might’ve been in sync. I’d thought I had you by a hair, but there is a slight delay, isn’t there? Whatever the case, we’ve done this a lot lately.

by Incipient_Senescence on Oct 25, 2009 5:18 PM EDT up reply actions 0 recs

Yeah. I'm not worked up about the timing that much.

But I do like the way you looked at that. And even without the benefit of undefeated records, the top few teams are still way up above everybody else, which says that their records match their performance.

by Hooper on Oct 25, 2009 5:19 PM EDT up reply actions 0 recs

Yeah

It really gives a good indicator of who’s results are far better than performance, those who’s results are worse than performance, and those who are well-matched. The first category are teams that are either lucky or just know how to win in the end, which notably include Boise, Iowa, and Cincy. But the computer hates Cincy, so that’s an anomaly. Iowa matches our intuitions. The second category is headlined by Tennessee to a crazy high degree. That tells us that UT is good but needs to learn to win. That takes a lot more time than I’d like. The last category includes most of the top 10, including basically all the teams who are just undeniably good.

by Incipient_Senescence on Oct 25, 2009 5:26 PM EDT up reply actions 0 recs

BTW

sub “whose” for “who’s” wherever you see it in that last comment. Dumb mistakes drive me crazy.

by Incipient_Senescence on Oct 25, 2009 5:27 PM EDT up reply actions 0 recs

I'd have to go back through the databases all over again to see how easy it is to pull that data.

The lovely thing about this program is, once I get it running, it’s convenient.

by Hooper on Oct 25, 2009 5:23 PM EDT up reply actions 0 recs

That's the holy grail right there - something like what Joel did for Tennessee-Bama?

I don’t know of a way to do it that isn’t brute force, though. Say hello to a 120×120x(variable)x(number of games) matrix. Ruh-roh. You’ll have to run the file in MATLAB or something.

Eric Berry for sending the guy who wins the Heisman spinning 720 degrees in the air at the podium - or for intercepting it and returning it to where it rightfully belongs

by Graysnail on Oct 25, 2009 5:48 PM EDT up reply actions 0 recs

I'd be willing to do that, too.

It would be dicey once I leave school – maybe have to break down and actually buy the silly thing – but I’m all for inverted matrices.

by Hooper on Oct 25, 2009 6:02 PM EDT up reply actions 0 recs

I had this whole idea on how that'd be the best method of measuring true output a couple of years ago.

How much does a copy of MATLAB run, anyway?

Eric Berry for sending the guy who wins the Heisman spinning 720 degrees in the air at the podium - or for intercepting it and returning it to where it rightfully belongs

by Graysnail on Oct 25, 2009 6:18 PM EDT up reply actions 0 recs

I was always afraid of mentioning it because I amost feel like obligating myself to the task.

But oh well.

MATLAB is $100ish for students. Not bad, really. I’m almost guaranteed to get one before I leave, but I get it for free right now so I’m waiting as long as I can to get the most up-to-date version possible.

by Hooper on Oct 25, 2009 6:23 PM EDT up reply actions 0 recs

Nah, I had thought about going down that road before, too.

Far as I can tell, unless there’s a simplified way to do it I think you need the power of it – heck, since (as far as I know) I was the first person to even think of the idea, I kind of think MATLAB would be needed. The best part of this is now you’re not reliant on someone else’s SOS schedule.

Two issues with the brute force method we’d need to address at some point:
1 – what to do with 1-AA teams?
2 – how to deal with the first few weeks of the season, when teams aren’t “linked”, in essence. i don’t think the second one is an issue, but the first one is.

Eric Berry for sending the guy who wins the Heisman spinning 720 degrees in the air at the podium - or for intercepting it and returning it to where it rightfully belongs

by Graysnail on Oct 25, 2009 6:27 PM EDT up reply actions 0 recs

The best part of this is now you’re not reliant on someone else’s SOS schedule.

And that would be a huge burden removed.

by Hooper on Oct 25, 2009 6:29 PM EDT up reply actions 0 recs

Heck, I'd even go so far as to say it's a better SOS metric.

It’s applied granularly, not uniformly – that’s more of what we care about for something like this.

Eric Berry for sending the guy who wins the Heisman spinning 720 degrees in the air at the podium - or for intercepting it and returning it to where it rightfully belongs

by Graysnail on Oct 25, 2009 6:32 PM EDT up reply actions 0 recs

I think so too.

I just need the time.

Remind me of this after basketball season is over. It would make a great offseason project.

by Hooper on Oct 25, 2009 6:33 PM EDT up reply actions 0 recs

Something interesting about that 0 for W/L trick of yours:

Ignoring wins and losses, UT ranks #11. The only other team with negative W/L in the top 30 is Michigan State at #29, and they are at .500. The second highest team with a losing record is Washington at #31.

(W/L is biased down a bit because of the overall winning record of 1-A teams against 1-AA teams. Being .500 is actually a little bit below a true average for 1-A teams.)

by Hooper on Oct 25, 2009 5:14 PM EDT up reply actions 0 recs

Haha

I just keep beating you by a hair. But I didn’t look back at other teams with similar records. That’s really an amazing difference, and just goes to show how big the disparity is between our overall play and our ability to win at the end. The first has largely arrived. The latter, not even close.

by Incipient_Senescence on Oct 25, 2009 5:17 PM EDT up reply actions 0 recs

*unless you're in the ACC

Eric Berry for sending the guy who wins the Heisman spinning 720 degrees in the air at the podium - or for intercepting it and returning it to where it rightfully belongs

by Graysnail on Oct 25, 2009 5:49 PM EDT up reply actions 1 recs

+1

by Incipient_Senescence on Oct 25, 2009 5:55 PM EDT up reply actions 0 recs

I have to say

that I’m really enjoying you mathy guys talking about this stuff. I mean that.

That “removing the W/L” thing is really interesting.

Carry on. I never called, so I’m not hanging up, but I am going to listen.

Rocky Top Talk

by Joel on Oct 25, 2009 6:01 PM EDT reply actions 0 recs

One idea that I'd LOVE to see answered (and I think we can, if we can incorporate SOS into the other metrics):

How much of an effect does SEC defense have on offenses, and will accounting for that depress the SEC’s overall numbers? This is basically being able to answer the question – empirically, with a hopefully reasonable sample size – if it really is all due to the defense.

Also, do we know how each of these stats correlates to winning in each season? Does it vary significantly?

Eric Berry for sending the guy who wins the Heisman spinning 720 degrees in the air at the podium - or for intercepting it and returning it to where it rightfully belongs

by Graysnail on Oct 25, 2009 6:23 PM EDT reply actions 0 recs

It varies

I have been investigating this question a little. The answers have not been terribly satisfying thus far.

________________________________
Eric Berry is better at football than you.

by kidbourbon on Oct 25, 2009 6:30 PM EDT up reply actions 0 recs

Good questions, those.

IIRC, the defensive stats tend to correlate more consistently, but I still feel like the relative proportion of weights shouldn’t be quite as extreme as they are at the moment. (E.g. 9:1 PED to 3DD – maybe more like 5:1?)

by Hooper on Oct 25, 2009 6:30 PM EDT up reply actions 0 recs

I don't know / remember the relative weight distribution at the moment.

However, generally speaking my thought would be the weights should be approximately proportional to the correlation.

Eric Berry for sending the guy who wins the Heisman spinning 720 degrees in the air at the podium - or for intercepting it and returning it to where it rightfully belongs

by Graysnail on Oct 25, 2009 6:34 PM EDT up reply actions 0 recs

Yeah, but it gets tricky.

There’s a lot of autocorrelation going on between those metrics, and you run the risk of overemphasizing things. I’d like to figure out how to weight them such that the lesser weights pick up the details that the higher weights don’t catch, but I haven’t settled in my mind if it makes more sense to weight proportional to correlation, or proportional to difference of correlation, or something in between.

by Hooper on Oct 25, 2009 6:36 PM EDT up reply actions 0 recs

autocorrelation

This is true. The best you are ever going to get out of the correlation numbers is a rough idea.

________________________________
Eric Berry is better at football than you.

by kidbourbon on Oct 25, 2009 6:48 PM EDT up reply actions 0 recs

Joel can attest:

I freaked out about the autocorrelation when he first proposed this system. But so far, it really hasn’t caused a problem.

by Hooper on Oct 25, 2009 6:50 PM EDT up reply actions 0 recs

I do hereby so attest

Although “freaking out” for Hooper is raising both eyebrows instead of one.

Rocky Top Talk

by Joel on Oct 25, 2009 7:34 PM EDT up reply actions 0 recs

O.O

by Hooper on Oct 25, 2009 7:37 PM EDT up reply actions 0 recs

Yes! That's it!

Rocky Top Talk

by Joel on Oct 25, 2009 7:39 PM EDT up reply actions 0 recs

Taking out W/L

I actually think this makes sense. I do my best thinking in the gym, and it occured to me while I was raining left hooks on the heavy bag that factoring in wins and losses as essentially a binary digit (i.e. just the win or the loss, not the margin of victory) has what I would perceive to be a flaw in that…well, it doesn’t factor in MOV. I realize this is how the BCS works (no MOV), but I also think the BCS is stupid for doing it that way. And people much smarter and better at statistics than myself (e.g. Jeff Sagarin) would agree with me.

I just think that adding in wins and losses sans MOV is just sort of misleading to the computers. And my epiphany inspring showering of punishing left hooks on the defenseless heavy bag occurred well before I knew taking W/L out of the equation would bump my Vols up to #11. So, even better.

________________________________
Eric Berry is better at football than you.

by kidbourbon on Oct 25, 2009 6:29 PM EDT reply actions 0 recs

It does that.

The binary effect gets washed out by the end of the season to a certain degree. And points to get accounted for within the other metrics (esp. OPPG, but also in TO). Score does have an effect – just not in a direct manner as the BCS hates so much.

Not that I care about the BCS’s limitations. The one that really bugs me is that the score the human polls based on the vote points but the computers by their ordinals. Whaaaaa?

by Hooper on Oct 25, 2009 6:33 PM EDT up reply actions 0 recs

Upshot:

The BCS is constructed in such a way that the computers are there to confirm the voters’ opinions, not the other way around. That’s why I don’t think about it too much; they’ve been so hamstrung that it really doesn’t matter.

But it still makes more sense than the Harris.

Eric Berry for sending the guy who wins the Heisman spinning 720 degrees in the air at the podium - or for intercepting it and returning it to where it rightfully belongs

by Graysnail on Oct 25, 2009 6:35 PM EDT up reply actions 0 recs

At least the Harris was entertaining.. ;-)

by Hooper on Oct 25, 2009 6:36 PM EDT up reply actions 0 recs

The MOV points would certainly be somewhat cumulative in nature...

…but we already have a lot of overlapping stats in the formula as it is.

One question I have. Why factor in third down offense and third down defense? Might it be better to factor in number of first downs? In other words, picking up first down on first down or second down is just as valuable a skill for a team as picking up first down on third down, I would think.

I can rationalize this more on the defensive side, where you have to stop third down to prevent the opposing team from getting another first down. But, again, it would seem as if a first down differential statistic would sort of wrap its brain around this idea in one fell swoop.

Just thinking aloud.

________________________________
Eric Berry is better at football than you.

by kidbourbon on Oct 25, 2009 6:43 PM EDT up reply actions 0 recs

It was a part of SMQ's correlation work.

A part of it is that 3rd down stats are easier to judge than 1st or 2nd down because they actually get recorded separately rather than tucked away in play-by-play. But a part of it is the concept of ‘success’.

On first, success is usually met at 40% of the way to the next first down (where 1st and goal means the end zone is a ‘first’). Second down success is usually considered 60% of whatever is needed at that point. 3rd down means 100%, which lines up perfectly with the 3rd down conversion rate. So even though it’s an incomplete measure, it does serve as some measure of a team’s success at moving the ball (or stopping it on defense). I would love to go all Football Outsiders on it and go by play-by-play to measure success on every snap, but for 120 teams, that’s a mountain of work. Not to mention that online PbP is often rife with errors and changes in grammatical structure. I’m not sure it’s realistic to pursue.

by Hooper on Oct 25, 2009 6:48 PM EDT up reply actions 0 recs

Fair enough

I can see data accumulation becoming a problem….in any system, really.

________________________________
Eric Berry is better at football than you.

by kidbourbon on Oct 25, 2009 6:51 PM EDT up reply actions 0 recs

Yeah, but boy I wish.

It’s not the volume of data that worries me. It’s the error-correcting. 60ish games a week, and potential for NaN garbage in, say, 5 plays apiece. No thanks.

by Hooper on Oct 25, 2009 6:54 PM EDT up reply actions 0 recs

CFBstats is your friend if you want to go down that route.

I have to think they have something that auto-uploads and updates that if you want to get to that level of detail – maybe contact him?

Eric Berry for sending the guy who wins the Heisman spinning 720 degrees in the air at the podium - or for intercepting it and returning it to where it rightfully belongs

by Graysnail on Oct 25, 2009 7:09 PM EDT up reply actions 0 recs

Good thought.

by Hooper on Oct 25, 2009 7:11 PM EDT up reply actions 0 recs

Nice call

I was not familiar with that site

________________________________
Eric Berry is better at football than you.

by kidbourbon on Oct 25, 2009 8:24 PM EDT up reply actions 0 recs

That site is like college football stat heaven.

I love that place.

Eric Berry for sending the guy who wins the Heisman spinning 720 degrees in the air at the podium - or for intercepting it and returning it to where it rightfully belongs

by Graysnail on Oct 25, 2009 8:30 PM EDT up reply actions 0 recs

Finally, something I can contribute to this conversation!

…I can’t imagine writing without that site.

Rocky Top Talk

by Will on Oct 25, 2009 8:51 PM EDT up reply actions 0 recs

The sentence with the word ordinals

I don’t think I am following

________________________________
Eric Berry is better at football than you.

by kidbourbon on Oct 25, 2009 6:45 PM EDT up reply actions 0 recs

Ordinals

For example, in my blogpoll draft, the ordinals are Florida #1, Alabama #2, etc. The actual scores are in the table. Ordinals are the numeric order, which washes out the actual difference between the teams.

When they measure the BCS, they count the human vote scores and divide by the maximum. They don’t use the #1, #2, etc. But they use those ordinals for the computers prior to averaging them. It ends up making a small difference.

by Hooper on Oct 25, 2009 6:49 PM EDT up reply actions 0 recs

^ the maximum possible, that is

by Hooper on Oct 25, 2009 6:51 PM EDT up reply actions 0 recs

Thanks. Understood.

________________________________
Eric Berry is better at football than you.

by kidbourbon on Oct 25, 2009 6:51 PM EDT up reply actions 0 recs

I wonder

Where Tennessee from last year would be, running these numbers.

Tennessee Fans: We win at teh Internet!

by bobo_the_vol on Oct 26, 2009 2:16 AM EDT reply actions 0 recs

Here's their line:

W/L: -0.5
SOS: 0.766
PED: 1.406
RD: 1.159
3DO: -0.801
TD: 1.73
PEO: -1.307
OPPG: 1.254
TO: -1.535
3DD: 0.3494

Offensive numbers are italicized… sigh…
With W/L, they’re # 45. Without W/L, they’re #22. But again, with defense getting more value than offense, the full effect of the #(!@fense might not be truly reflected there.

by Hooper on Oct 26, 2009 9:06 AM EDT up reply actions 0 recs