My very own computer poll for the Big XII

**Judoka** · 10-24-2012, 09:48 AM

I have always considered making my own computer poll and I decided to go ahead and take the plunge.

My methodology is to look at how many points/yards a team holds their opponents to compared to their opponent's average and use that to predict future outcomes. So for example in the Baylor/Iowa State game this weekend:

Baylor's games average 48-48.2. Baylor tends to score 16.2 points more than their opponents usually give up, and Baylor's opponents tend to score 11.02 points more than they usually score. Iowa State games average 21.33-22.33. Iowa State tends to score 1 point less than their opponent's average while holding their opponents to 12.51 points less than they usually score.

So to predict Baylor's score I take what I'd expect from Baylor's offensive average vs Iowa State's defensive average (48-12.51=35.5) and then what I'd expect for Iowa's State's defensive average vs Baylor's offensive average (22.33+16.2=36.5). Averaging this I get 37 (from rounding up) as the predicted score for Baylor. Doing the same for ISU gives me a predicted score of 40 for Iowa State. The algorithm is the same for rushing/passing yards. This leads to the following prediction for the box score:

Code:

+------------+---------+-----+
|   Baylor   | Scoring | 37  |
|            | Rushing | 152 |
|            | Passing | 402 |
+------------+---------+-----+
| Iowa State | Scoring |  40 |
|            | Rushing | 182 |
|            | Passing | 313 |
+------------+---------+-----+

This is a very rough model right now, but I expect it to get better each week. I'm also thinking it might be better to change it from raw points to percentages (i.e. rather than ISU holding teams to 12.51 below their average, they are holding them to 70% of their average) for the calculations.

Each week I'm going to update this thread with my predictions for the following week, and my thoughts on what went good/bad from the week before.

**Judoka** · 10-24-2012, 09:53 AM

Power rankings for this week:

Code:

+-------------------+
|    1. Oklahoma    |
| 2. Kansas State   |
| 3. Texas Tech     |
| 4. Texas          |
| 5. Oklahoma State |
| 6. TCU            |
| 7. Iowa State     |
| 8. Baylor         |
| 9. West Virginia  |
| 10. Kansas        |
+-------------------+

Expected game outcomes:

Code:

+------------+---------+-----+
|   Baylor   | Scoring | 37  |
|            | Rushing | 152 |
|            | Passing | 402 |
+------------+---------+-----+
| Iowa State | Scoring |  40 |
|            | Rushing | 182 |
|            | Passing | 313 |
+------------+---------+-----+

Code:

+--------+---------+-----+
| Texas  | Scoring | 45  |
|        | Rushing | 184 |
|        | Passing | 259 |
+--------+---------+-----+
| Kansas | Scoring |  23 |
|        | Rushing | 226 |
|        | Passing | 197 |
+--------+---------+-----+

Code:

+----------------+---------+-----+
| Oklahoma State | Scoring | 29  |
|                | Rushing | 178 |
|                | Passing | 303 |
+----------------+---------+-----+
| TCU            | Scoring |  28 |
|                | Rushing | 169 |
|                | Passing | 275 |
+----------------+---------+-----+

Code:

+--------------+---------+-----+
| Kansas State | Scoring | 33  |
|              | Rushing | 174 |
|              | Passing | 171 |
+--------------+---------+-----+
| Texas Tech   | Scoring |  29 |
|              | Rushing | 101 |
|              | Passing | 323 |
+--------------+---------+-----+

**Judoka** · 10-24-2012, 09:54 AM

And expected conference finish. This does not totally agree with the power rankings because there have been three games that did not go as the algorithm expects they should have (us over TCU, WVU over Baylor, and K-State over Oklahoma)

Code:

+----------------+------+--------+
|                | Wins | Losses |
+----------------+------+--------+
| Kansas State   |    9 |      0 |
| Oklahoma       |    8 |      1 |
| Texas Tech     |    7 |      2 |
| Texas          |    5 |      4 |
| Oklahoma State |    5 |      4 |
| Iowa State     |    4 |      5 |
| TCU            |    3 |      6 |
| West Virginia  |    3 |      6 |
| Baylor         |    1 |      8 |
| Kansas         |    0 |      9 |
+----------------+------+--------+

**VeloClone** · 10-24-2012, 09:58 AM

I'd take 4-5 in a heartbeat.

**Judoka** · 10-24-2012, 10:05 AM

Originally Posted by VeloClone

I'd take 4-5 in a heartbeat.

Me too. Teams only have a few games so I'd be happy if I'm even 50% correct with the predictions right now. But 4-5 would be awesome. Right now the model calls for us to beat Baylor 40-37, lose to Oklahoma 10-31, lose to Texas 29-34, beat Kansas 26-9, and close the season with a 35-30 win over West Virginia. Depending on how the games go this weekend that prediction could change a lot.

**ianoconnor** · 10-24-2012, 11:01 AM

Very cool.

**chuckd4735** · 10-24-2012, 11:07 AM

Cool stuff. I expect this every week from you :)

**Jerms** · 10-24-2012, 11:08 AM

Originally Posted by Judoka

Me too. Teams only have a few games so I'd be happy if I'm even 50% correct with the predictions right now. But 4-5 would be awesome. Right now the model calls for us to beat Baylor 40-37, lose to Oklahoma 10-31, lose to Texas 29-34, beat Kansas 26-9, and close the season with a 35-30 win over West Virginia. Depending on how the games go this weekend that prediction could change a lot.

I could live with that. Right now it's better than what I'm expecting...

**Judoka** · 10-24-2012, 11:13 AM

Originally Posted by chuckd4735

Cool stuff. I expect this every week from you :)

The hard part was getting it all set up. It shouldn't take more than 10 minutes of work each week now to keep it updated.

**melt** · 10-24-2012, 11:17 AM

Originally Posted by Judoka

This is a very rough model right now, but I expect it to get better each week. I'm also thinking it might be better to change it from raw points to percentages (i.e. rather than ISU holding teams to 12.51 below their average, they are holding them to 70% of their average) for the calculations.

Each week I'm going to update this thread with my predictions for the following week, and my thoughts on what went good/bad from the week before.

I agree, go with percentages. Straight numbers will get you some wonky results for really good and really bad teams, like possibly moving teams into negative points.

**yCy** · 10-24-2012, 11:44 AM

Nice work, Judoka. I look forward to seeing how each week goes. I'd be very happy if we could go 4-5 in conference.

**KillBilly** · 10-24-2012, 11:57 AM

Those of us that closely follow the Tech program knew we would be significantly better this season but 7-2?

That's seems crazy but I would love for you to be right. UT and OSU have had Tech's number the past few years so losing either or both of those games is very possible. OSU is the more likely loss.

**NickTheGreat** · 10-24-2012, 12:13 PM

Well done. I'll be curious to see how this pans out

**Judoka** · 10-24-2012, 12:24 PM

Originally Posted by melt

I agree, go with percentages. Straight numbers will get you some wonky results for really good and really bad teams, like possibly moving teams into negative points.

I made that change and it isn't looking to have a huge impact on outcomes, but the scores shift a bit. Using percentages instead of raw values Iowa State is still predicted to go 4-5 in conference with the following:

37-32 win over Baylor
13-28 loss to Oklahoma
29-31 loss to Texas
27-12 win over Kansas
34-26 win over West Virginia.

Thanks for the recommendation. The algorithm is still a draft so the more I can improve it the better.

**melt** · 10-24-2012, 01:29 PM

I'm doing my own computer poll, but you're doing it to different ends. I'm just ranking all the teams, while you are trying to predict scores, which is cool, I think that's probably more interesting than what I'm doing. Anyway, in case anyone gives a crap, here's some Big 12 rankings. 1st number is rank in Big 12, parentheses is overall rank.

Points For-Points Against:
1. KSU (4)
2. Oklahoma (5)
3. TTU (9)
4. OSU (16)
5. TCU (23)
6. Texas (42)
7. ISU (60)
8. Baylor (63)
9. WVU (71)
10. Kansas (113)

Yards For-Yards Against
1. TTU (3)
2. OSU (4)
3. Okalhoma (8)
4. TCU (15)
5. KSU (21)
6. Baylor (66)
7. WVU (72)
8. Texas (80)
9. ISU (90)
10. Kansas (97)

It makes sense that PF-PA correlates more to wins, but it's kind of amazing how jumbled it gets at YA-YF. My power rankings for the Big 12 are as follows:

1. KSU (3)
2. TTU (6)
3. Oklahoma (7)
4. OSU (22)
5. WVU (25)
6. Texas (28)
7. TCU (38)
8. ISU (47)
9. Baylor (52)
10. Kansas (102 haha)

This is a formula based on wins, PF-PA, YDSF-YDSA, strength of schedule, and conference membership. Lastly, here's how I rank the conferences, the number to the right is based on 1.00 being the average, so above 1.00 is good, below is bad.

Code:

CONFERENCE
FACTOR

BIG12
1.311

SEC
1.171

BIG10
1.124

BIGE
1.106

PAC12
1.081

ACC
1.043

IND
1.000

WAC
0.959

MAC
0.915

SUN BELT
0.886

MWC
0.788

CUSA
0.615