As many know, I run a College Football analytics site called PerfectFall. On the site I use a model I created to measure every teams on-field performance and gives them an Overall, Offensive, and Defensive 's-values'. It also measures team talent, offensive tendencies, and strength of schedule/opponent. *Offensive tendencies is used to rate a team appropriately. A run heavy team shouldn't be penalized as much for having a poor passing game, but if they then have a poor rushing game, it should punish them more. And vice versa.
New this year, I'm using data created by my site to train a few machine learning models which can then predict future results (this feature is normally only available to Patreon members, but I'll post just the Purdue prediction here in this column each week).
Things like these rely on data, data, and more data. That means 2 things. For ratings, I do factor in preseason values for the first 4 weeks, but my entire models intention is to rate how teams are performing right now, this season and not overly rely on the past, so rankings early in the season may look somewhat weird for the first 3-4 weeks. And second, even more so on the reliance on data, I dont expect predictions to be any sort of reliable until at least week 5. So while I will post the prediction, if you use it to bet, it's your money, not mine #ThisIsNotFinancialAdvice
Statistical Analysis...
Unfortunately don't have time to do the full spiel this week, so just gonna include the prediction, update the table and call it good for this week
Purdue:
Minny:
Machine learning prediction:
Pretty similar prediction to last week. Expects another low scoring affair. It was pretty close on the Illinois score last week, but expected more Boiler points. Let's hope this one is just as close on the opponent side again but also close on the good guys side as well this time... Model is 3-0 picking the winner so far this season at least. This last week was the furthest off its been from the actual spread.
Last section I'll include is our scheduled opponents, where they currently sit in the ratings, and the predicted outcomes vs actual results:
Again, expect rankings and predictions to improve as the season goes as with ANY mathematical rating model, not just my own Though pretty happy in general with how they are looking right now! Let me know what y'all think!
New this year, I'm using data created by my site to train a few machine learning models which can then predict future results (this feature is normally only available to Patreon members, but I'll post just the Purdue prediction here in this column each week).
Things like these rely on data, data, and more data. That means 2 things. For ratings, I do factor in preseason values for the first 4 weeks, but my entire models intention is to rate how teams are performing right now, this season and not overly rely on the past, so rankings early in the season may look somewhat weird for the first 3-4 weeks. And second, even more so on the reliance on data, I dont expect predictions to be any sort of reliable until at least week 5. So while I will post the prediction, if you use it to bet, it's your money, not mine #ThisIsNotFinancialAdvice
Statistical Analysis...
Unfortunately don't have time to do the full spiel this week, so just gonna include the prediction, update the table and call it good for this week
Purdue:
Minny:
Machine learning prediction:
Pretty similar prediction to last week. Expects another low scoring affair. It was pretty close on the Illinois score last week, but expected more Boiler points. Let's hope this one is just as close on the opponent side again but also close on the good guys side as well this time... Model is 3-0 picking the winner so far this season at least. This last week was the furthest off its been from the actual spread.
Last section I'll include is our scheduled opponents, where they currently sit in the ratings, and the predicted outcomes vs actual results:
| Week | Opponent | Current Ranking (Movement) | Predicted Outcome (Spread) | Actual Outcome (Spread) |
|---|---|---|---|---|
| 1 | Oregon State | 35 (up 23) | N/A | |
| 2 | UConn | 130 (still last lol no movement) | 43-3 Purdue (-40) | 49-0 (-49) |
| 3 | Notre Dame | 6 (down 1) | 27 - 45 Notre Dame (+18) | 13-27 (+14) |
| 4 | Illinois | 94 (down 6) | 32-15 (-17) | 13-9 (-4) |
| 5 | Minnesota | 72 (down 27) | 29-14 (-15) | TBD |
| 6 | Iowa | 33 (down 11) | TBD | TBD |
| 7 | Wisconsin | 42 (down 9) | TBD | TBD |
| 8 | Nebraska | 54 (down 5) | TBD | TBD |
| 9 | Michigan Sate | 7 (up 14) | TBD | TBD |
| 10 | Ohio State | 21 (down 14) | TBD | TBD |
| 11 | Northwestern | 83 (down 11) | TBD | TBD |
| 12 | Indiana | 77 (down 20) | TBD | TBD |
Again, expect rankings and predictions to improve as the season goes as with ANY mathematical rating model, not just my own
Though pretty happy in general with how they are looking right now! Let me know what y'all think!
Last edited:
