re: CFB Advanced Metrics and Computer Polls, Week 12

quote:

---

whereas a system that serves as an aggregate (mine) will do the opposite.

---

What do you mean as an aggregate? Like an aggregate of the composite measures of score differential from an average team (FPI, SRS, SP+, etc.) or at least some composite of continuous variables and their properties (e.g., mean, variance, etc.)? Or are you talking about a composite of the ordinal rankings like the Massey composite? Because ordinal rankings and their aggregation has some considerable problems and weaknesses, both statistically (e.g., the assumptions about the underlying data to make comparisons usually aren't valid; unequal distance between rankings) and practically (what do the rankings mean in terms of the performance of the game itself and winning).

quote:

---

What do you mean as an aggregate?

---

An aggregate is basically the pile of points you acquire during all the stuff you do during the season, usually disregarding when the wins happen (a predictor will understandably place more weight on recent games than on earlier games) or sometimes even where the wins happen (which a predictor should probably take into account).

Power points are the most basic aggregate; most high school associations use this. Louisiana gives you 10 points for a win and 1 point for an opponent's win. Playoff seeds are determined that way. If you beat a team who has four wins, you aggregate 14 points.

My system is an aggregate based on a semi-complex equation of nothing but a team's wins and losses and those of their opponents and opponents' opponents.

ETA: Here's a basic aggregate for CFB I just made for perspective:

wins: A
opponents' wins: B
opponents' opponents' wins: C
maximum number of games played: X

(A*X^2) + BX + C = a team's power ranking


1 Clemson 2623
2 LSU 2561
3 Ohio St 2540
4 Penn St 2429
5 Florida 2416
6 Georgia 2385
7 Cincinnati 2375
8 Notre Dame 2343
9 Oregon 2340
10 SMU 2333
11 Alabama 2328
12 Minnesota 2323
13 Utah 2321
14 Oklahoma 2318
15 Michigan 2315
16 Boise St 2307
17 Baylor 2304
18 Memphis 2300
19 Wisconsin 2298
20 USC 2257
21 Auburn 2247
22 Appalachian St 2245
23 Iowa 2199
24 Air Force 2193
25 Louisiana 2116
26 Florida St 2098
27 Wake Forest 2097
28 Oklahoma St 2077
29 Louisville 2072
30 Hawaii 2069
31 Texas 2065
32 Temple 2060
32 Temple 2060
34 Texas A&M 2056
35 Utah St 2054
36 BYU 2052
37 Marshall 2045
38 W Michigan 2043
39 Virginia 2042
40 UCF 2040
41 San Diego St 2037
42 Iowa St 2033
43 FL Atlantic 2031
44 Washington 2028
45 Navy 2013
46 Miami FL 1994
47 Indiana 1993
48 Louisiana Tech 1981
49 Kansas St 1973
50 Miami OH 1972
51 C Michigan 1956
52 Southern Miss 1948
53 Ga Southern 1939
54 Tulane 1932
55 Virginia Tech 1930
56 Arkansas St 1925
57 Tennessee 1924
58 Illinois 1920
59 Nevada 1887
60 Mississippi St 1885
61 TCU 1881
62 Michigan St 1879
63 Georgia St 1878
64 Wyoming 1871
65 California 1853
66 South Carolina 1851
67 WKU 1828
67 WKU 1828
69 North Carolina 1827
70 Kentucky 1821
71 UCLA 1808
72 Boston College 1807
73 Toledo 1801
74 Arizona St 1794
75 Tulsa 1774
76 Mississippi 1769
77 Missouri 1760
78 Washington St 1753
79 UAB 1752
80 South Florida 1745
81 Purdue 1741
82 Nebraska 1737
83 Buffalo 1735
84 Colorado 1731
85 Stanford 1726
86 West Virginia 1710
87 ULM 1707
88 Army 1703
89 Charlotte 1702
90 Maryland 1699
90 Maryland 1699
92 Arizona 1690
93 Syracuse 1680
93 Syracuse 1680
95 E Michigan 1675
95 E Michigan 1675
97 Ohio 1670
98 Houston 1659
99 Texas Tech 1655
100 San Jose St 1649
101 Troy 1637
102 Liberty 1626
103 Ball St 1625
104 Coastal Car 1608
105 Kent 1601
106 Florida Intl 1597
107 North Texas 1586
108 Colorado St 1559
109 Kansas 1553
110 Texas St 1549
111 N Illinois 1545
112 Rutgers 1528
113 East Carolina 1526
114 UT San Antonio 1524
115 MTSU 1521
116 Bowling Green 1520
117 Georgia Tech 1496
118 Vanderbilt 1465
119 Northwestern 1430
120 Arkansas 1414
121 Connecticut 1391
122 New Mexico 1382
123 Rice 1376
124 UNLV 1349
125 South Alabama 1305
126 Old Dominion 1249
127 New Mexico St 1227
128 UTEP 1199
129 Massachusetts 1069
130 Akron 1001


Clemson, Florida, and USC seem a bit high because they've played 11 games as opposed to most everybody else's 10. That would take care of itself in two weeks. Then the teams who play for their conference championship will get an extra boost, as well they should, in my opinion.

quote:

---

Cincinnati's opponents have a better combined record. Cincinnati's opponents' opponents have a better combined record.

That's a pretty reasonable way to judge these things, I'd say, and that's before you take into account that tOSU beat Cincinnati by 38 points more than the points by which LSU beat Bam

---

It’s not that reasonable actually. You’re basically applying an RPI-like system used in basketball, which itself is pretty flawed, but it’s less flawed in basketball due to a larger sample size, larger and better crossover data between and within conferences without the issues with the 2 subdivisions, and there is better dispersion of quality both across conferences and across the spectrum of teams in general.

But those issues that minimize the methods problems, are absent in football. And on top of that, some major conferences require their teams to play 9 conferences games whereas other conferences play only 8 and teams play really poor FBS teams outside the power 5 or FCS teams. And using your method, teams in 9 game conferences would be disproportionately impacted by the fact that they’re trading a likely win for a tougher game, and they’re losing to one another.

quote:

---

It’s not that reasonable actually. You’re basically applying an RPI-like system

---

No, I'm simply comparing two teams in a discussion.

quote:

---

And using your method, teams in 9 game conferences would be disproportionately impacted by the fact that they’re trading a likely win for a tougher game, and they’re losing to one another.

---

My method effectively penalizes capitalism. Playing more conference games is a money grab. Playing FCS teams is a money grab. Both effectively hurt a team's ranking.

quote:

---

With only 4 of 130 teams making it in CFB, the metrics need to be much more precise and ensure that there are no false positives, even though there may be some false negatives unfortunately.

---

Very reasonable take with which I would be more inclined to agree if every conference didn't have a championship game now. 10 teams will all have one extra win over the #1 team in their conference's opposite division to separate themselves from the pack at the end of the regular season. What you call 4 out of 130 is, the way I see it, 4 out of 10+a few. The value of an extra win while the others are sitting at home is, in my opinion, more considerable than the inconvenience of the relatively small playoff field.

quote:

---

There is no magic computer bullet.

---

That's the kick in the nuts of nerds like me everywhere. The more complicated we make our math, the more we realize that we're running in circles. For example: as crazy as it sounds, in my hundreds of thousands of hours of tinkering around with this stuff and comparing techniques with other supernerds, I've concluded that counting margin of victory is no better than not counting margin of victory.

quote:

---

These aren't words I'm leaning on or would recommend you or anyone else lean on. A rating system that serves as an aggregate and/or counts margin of victory likely counts tOSU's win over Cincinnati as more valuable than LSU's win over Alabama, and it isn't unreasonable to do so.

---

You used the word better first

quote:

---

Not only is Cincinnati a quality win, but it's a better win than the Alabama win regardless of the score.

---

Also again I think Cincinnati is a high quality win, but it’s a ridiculous statement to say it’s a better win than Alabama regardless of the score. If you said it was a better win including the score then you’d have a pretty good argument. Neither Alabama or Cincinnati have a top 25 win and both have very weak schedules. So what makes Cincinnati a clearly better win? You seemed to be way too obsessed with numbers and stats

quote:

---

This probably isn't the thread for you.

---

Why? I’m just trying to understand his point of view

quote:

---

If a team wins a conference then how are they undeserving? Who cares how many losses they have? They accomplished something that nobody else in the conference could do, win a championship.

I know people hate the bylaws and 8 I post but those are truly the only and best solutions

---

Not necessarily undeserving for other teams within that conference, but could certainly be considered undeserving for teams in other conferences.