King Yao-ing College Football Season Wins

[bevo_stevo]

King Yao's season win article inspired this attempt at creating a season win total for the college football season. However, instead of using my daft handicapping abilities to create power rankings I decided to use the market values for known games to generate rankings and then use those rankings to simulate the entire college football season.

There were a number of simplifications that went into the model and several unknowns that had to be taken into account. For example, all the the I-AA teams were assumed to be huge underdogs (52 points on a neutral field against USC) to most of the I-A teams. Additionally, some of the power rankings for the smaller division I-A schools had to be fudged (like Rice, San Diego State, et al). Therefore if you have reason to believe that this years Weber State team is destined for college football greatness you should back off the numbers that are given against their opponents.

In any case, since the results are fairly lengthy I'll just provide the results for some of the more popular teams, and if you're curious about any others I'll post them later. The predicted spread is given in parentheses.

Notre Dame Season Wins: 6.44
Georgia Tech (-2.5) 0.551
@Penn St (8.5) 0.269
@Michigan (12.0) 0.22
Michigan St (-7.5) 0.698
@Purdue (4.0) 0.391
@UCLA (9.5) 0.266
Boston College (-1.5) 0.566
USC (12.0) 0.22
Navy (-14.0) 0.822
Air Force (-9.0) 0.718
Duke (-28.0) 0.985
@Stanford (-9.5) 0.734


USC Season Wins: 10.238
Idaho (-43.0) 0.99
@Nebraska (-11.0) 0.765
Washington St (-31.5) 0.99
@Washington (-25.5) 0.985
Stanford (-30.5) 0.99
Arizona (-18.5) 0.904
@Notre Dame (-12.0) 0.78
@Oregon (-11.0) 0.765
Oregon St (-14.5) 0.83
@California (-7.0) 0.69
@Arizona St (-11.5) 0.775
UCLA (-11.5) 0.775

Texas Season Wins: 9.703
Arkansas St (-37.0) 0.99
TCU (-17.0) 0.872
@Central Florida (-12.5) 0.804
Rice (-27.0) 0.985
Kansas St (-17.5) 0.882
Oklahoma (0.0) 0.500
@Iowa St (-18.5) 0.904
@Baylor (-29.0) 0.985
Nebraska (-11.0) 0.765
@Oklahoma St (-3.5) 0.6
Texas Tech (-15.0) 0.832
@Texas A&M (-3.0) 0.584

[MyTurn2Raise]

how are you converting spreads into moneylines/win pcts?
They don't correspond from college to pro.
They are also heavily dependent on the expected total.

market values at the beginning of the season are not the greatest estimate for power rankings


power rankings ignore individual matchups and schemes which matter quite a bit in a systems based game like football


ummm...how are you factoring in home field?
some home fields are worth more
home fields matter more when teams are near even or when the home team is slightly worse



I also use something similar for teams


more important though is to look at the possible dispersion of win totals... the mean is not the most relevant for many of these types of bets

[MyTurn2Raise]

also...some of those .99s should be higher than that

we looked into how well 26 point+ favorites did a few times last season over the past decade or so....it's higher than .99 ... in addition, this year will be better for the big favorites due to changes in the rules/game structures... some of those games should just be 1.00s

[MyTurn2Raise]

I don't want to seem critical

this is great work that you've put up...one of the best first two posts ever

I just want to know how you are getting spreads and converting those into win pcts

[MyTurn2Raise]

[ QUOTE ]
also...some of those .99s should be higher than that

we looked into how well 26 point+ favorites did a few times last season over the past decade or so....it's higher than .99 ... in addition, this year will be better for the big favorites due to changes in the rules/game structures... some of those games should just be 1.00s

[/ QUOTE ]

quoted for making me look ridiculously stupid

also, the Big11Ten has lost all claims to being an elite conference

[bevo_stevo]

I think your criticisms are very accurate and important. One of the first things they teach you in college it that truth comes through argument so I'll try to respond to some of the points that you brought up.

The spread-to-moneyline conversions are based roughly on the Pinny conversion numbers from last year. They are not perfectly fit, I didn't create a table, I just created a function that estimated the moneyline based on several known data points.

You're probably right that some of the percentages should be greater than .99 for extremely high spreads. The important thing to realize is that going from .99 to .999 will probably not affect the season win totals in any important way.

For home field, I just used a 3 point home team advantage. I think the number should actually be a bit higher than that if I remember correctly (closer to 4) - I did that work last year but forgot what the actual number was. An additional point is that some of the neutral site games were not accurately adjusted for, only the OU/Texas UCLA/USC games were taken into account - so inaccuracies may have been introduced there as well.

I decided to use the market values to adjust the spreads simply due to the fact that there are so many teams in college football - most of which I know nothing about. I completely agree with your point that certain types of matchups will exaggerate point spreads and that this model does not take that into account.

I tend to disagree about the mean being an accurate way to predict totals, this really goes back to King Yao's work and I think he did an excellent job arguing the merits of the system.

In any case, after posting the Syracuse totals yesterday I realized that there was a bug in my graph traversal algorithm, and I had to make some minor adjustments to my code. This change affected all the posted totals (the Syracuse one more than any others), so I'll repost the new results.

USC Season Wins: 10.447
Idaho (-43.0) 0.99
@Nebraska (-10.0) 0.746
Washington St (-22.0) 0.949
@Washington (-29.5) 0.985
Stanford (-31.5) 0.99
Arizona (-20.0) 0.914
@Notre Dame (-12.0) 0.78
@Oregon (-12.5) 0.804
Oregon St (-21.0) 0.939
@California (-8.5) 0.731
@Arizona St (-11.5) 0.775
UCLA (-16.0) 0.845

Notre Dame Season Wins: 6.475
Georgia Tech (-3.0) 0.584
@Penn St (8.0) 0.293
@Michigan (11.0) 0.235
Michigan St (-7.5) 0.698
@Purdue (4.0) 0.391
@UCLA (9.5) 0.266
Boston College (-2.0) 0.528
USC (12.0) 0.22
Navy (-14.0) 0.822
Air Force (-9.0) 0.718
Duke (-28.0) 0.985
@Stanford (-9.5) 0.734

Texas Season Wins: 9.807
Arkansas St (-37.0) 0.99
TCU (-17.0) 0.872
@Central Florida (-12.5) 0.804
Rice (-26.0) 0.985
Kansas St (-21.5) 0.95
Oklahoma (-0.0) 0.500
@Iowa St (-18.0) 0.888
@Baylor (-29.0) 0.985
Nebraska (-9.0) 0.718
@Oklahoma St (-7.5) 0.698
Texas Tech (-15.0) 0.832
@Texas A&M (-3.0) 0.584

[bevo_stevo]

Heres how some of the perennial SEC powers are expected to perform:

Florida Season Wins: 8.605
Western Kentucky (-28.5) 0.985
Troy (-34.0) 0.99
Tennessee (-6.0) 0.646
@Mississippi (-12.5) 0.804
Auburn (-7.0) 0.69
@LSU (7.0) 0.31
@Kentucky (-3.0) 0.584
@Georgia (2.0) 0.472
Vanderbilt (-20.5) 0.923
@South Carolina (-3.0) 0.584
Florida Atlantic (-36.0) 0.99
Florida St (-5.0) 0.626

Arkansas Season Wins: 8.084
Troy (-26.0) 0.985
@Alabama (2.0) 0.472
Kentucky (-3.0) 0.584
North Texas (-32.5) 0.99
Chattanooga (-37.5) 0.99
Auburn (1.0) 0.488
@Mississippi (-6.5) 0.661
Florida International (-36.5) 0.99
South Carolina (-3.0) 0.584
@Tennessee (6.0) 0.354
Mississippi St (-12.0) 0.78
@LSU (13.0) 0.206


Georgia Season Wins: 8.416
Oklahoma St (-7.0) 0.69
South Carolina (-8.0) 0.707
Western Carolina (-41.0) 0.99
@Alabama (-2.0) 0.528
Mississippi (-17.5) 0.882
@Tennessee (2.5) 0.449
@Vanderbilt (-12.0) 0.78
Florida (-2.0) 0.528
Troy (-30.0) 0.99
Auburn (-5.0) 0.626
Kentucky (-6.5) 0.661
@Georgia Tech (-3.0) 0.584

LSU Season Wins: 9.721
@Mississippi St (-16.0) 0.845
Virginia Tech (-7.0) 0.69
Middle Tenn State (-43.0) 0.99
South Carolina (-13.0) 0.794
@Tulane (-28.5) 0.985
Florida (-7.0) 0.69
@Kentucky (-7.0) 0.69
Auburn (-10.0) 0.746
@Alabama (-6.0) 0.646
Louisiana Tech (-32.5) 0.99
@Mississippi (-16.5) 0.861
Arkansas (-13.0) 0.794

Auburn Season Wins: 7.903
Kansas St (-13.0) 0.794
South Florida (-2.5) 0.551
Mississippi St (-15.0) 0.832
New Mexico State (-23.5) 0.972
@Florida (7.0) 0.31
Vanderbilt (-15.5) 0.859
@Arkansas (-1.0) 0.512
@LSU (10.0) 0.254
Mississippi (-14.5) 0.83
Tennessee Tech (-38.5) 0.99
@Georgia (5.0) 0.374
Alabama (-5.0) 0.626

Tennessee Season Wins: 8.063
@California (4.0) 0.391
Southern Miss (-17.5) 0.882
@Florida (6.0) 0.354
Arkansas St (-31.5) 0.99
Georgia (-2.5) 0.551
@Mississippi St (-9.0) 0.718
@Alabama (-1.0) 0.512
South Carolina (-6.0) 0.646
Louisiana-Lafayette (-32.5) 0.99
Arkansas (-6.0) 0.646
Vanderbilt (-17.5) 0.882
@Kentucky (-0.0) 0.5

[Thremp]

[ QUOTE ]
You're probably right that some of the percentages should be greater than .99 for extremely high spreads. The important thing to realize is that going from .99 to .999 will probably not affect the season win totals in any important way.

[/ QUOTE ]

They should just be 1.

[GoldenBears]

Nice... I love the idea of these bets, but I've always hated tying my money up for the entire year. I only take one when I think I've got a huge edge.

Where did you get the lines for these games? A lot of them are on GOTYs but did you just guess at the rest or is there somewhere that lists every single game as a GOTY? I don't mind the spread to moneyline conversion like a couple of other guys did, I think that fussing about it is worrying about rounding a rough estimate to three decimal places.


Let me congratulate you - these are some pretty sharp measurements, you must be pretty close to what the houses are using.

going with the 5dimes and greek lines:
Notre Dame under 7.5 at -150 (greek) is the only one that differs by more than 1.0 games, and even then has a ton of juice with it.
florida under 9.5 at -155 (dimes) or under 10 at -320 (greek) and Tennessee under 9 -140 (greek) are off by almost a game, but again have 40-50 points of juice.

LSU and Auburn are off by a bit but again, the juice evens it out.


Can you do a computation for Cal? Just for my curiosity? I want to see how biased I am... these were numbers that I estimated and pulled out of my ass:

09/01 Tennessee .61
09/08 at Colo St .86
09/15 La Tech 6:30 .98
09/22 Arizona TBD .79
09/29 at Oregon .50
10/13 Oregon St .76
10/20 at UCLA .55
10/27 at Arizona St .58
11/03 Wash St .85
11/10 USC .39
11/17 at Washington .66
12/01 at Stanford .91

8.44 wins. (although honestly I think I am selling us short on a few of those games)

[GoldenBears]

Also, can you do:

Oregon (over 7 -105)
Pittsburgh (over 6.5 +125)
WVU (over 10.5 +120)
Arizona State (over 7.5 -110)
Arizona (over 6.5 -130)



Thanks man