NFL 2015 Week 17 Predictions – Zoltar likes Vegas Favorites Colts, Texans, Broncos, Chiefs

Predicting the results of NFL football games is a fascinating problem. I wrote a computer program called Zoltar to do just that. Unlike many prediction programs that use classical statistics techniques, Zoltar uses an information-theoretic approach.

Zoltar’s point spread predictions for week 17 of the 2015 NFL season are:

Zoltar favorite =     falcons  by    6  underdog =      saints
Zoltar favorite =       bills  by    1  underdog =        jets
Zoltar favorite =       lions  by    0  underdog =       bears
Zoltar favorite =     bengals  by   10  underdog =      ravens
Zoltar favorite =    steelers  by    9  underdog =      browns
Zoltar favorite =       colts  by   11  underdog =      titans
Zoltar favorite =    redskins  by    0  underdog =     cowboys
Zoltar favorite =      texans  by   11  underdog =     jaguars
Zoltar favorite =    patriots  by    9  underdog =    dolphins
Zoltar favorite =      eagles  by    0  underdog =      giants
Zoltar favorite =    panthers  by   12  underdog =  buccaneers
Zoltar favorite =   cardinals  by    6  underdog =    seahawks
Zoltar favorite =     broncos  by   11  underdog =    chargers
Zoltar favorite =      chiefs  by   10  underdog =     raiders
Zoltar favorite =        rams  by    0  underdog = fortyniners
Zoltar favorite =     packers  by    6  underdog =     vikings

Notice that Zoltar predicts a tie in four games this week. When this happens, just in order to calculate a head-to-head winning percentage, Zoltar predicts the team with the higher internal rating (Zoltar computes ratings for each team) will win.

Zoltar’s predictions are usually very close to the Las Vegas point spreads, but there are six games where Zoltar sees a theoretical betting opportunity:

* Zoltar says the Bills will beat the Jets by 1 point but Vegas says the Jets will win by 3.0 points. so Zoltar would recommend betting on the Vegas underdog Bills.

* Zoltar says the Colts will beat the Titans by 11 points but Vegas says the Colts will win only by 6.0 points, so Zoltar would recommend betting on the Vegas favorite Colts.

* Zoltar says the Texans will beat the Jaguars by 11 points but Vegas says the Texans will win only by 6.0 points, so Zoltar would recommend betting on the Vegas favorite Texans.

* Zoltar says the Broncos will beat the Chargers by 11 points but Vegas says the Broncos will win only by 7.5 points, so Zoltar would recommend betting on the Vegas favorite Broncos.

* Zoltar says the Chiefs will beat the Raiders by 10 points but Vegas says the Chiefs will win only by 6.5 points, so Zoltar would recommend betting on the Vegas favorite Chiefs.

* Zoltar says the Rams and the 49ers are evenly matched, but Vegas says the Rams will win by 3.5 points, so Zoltar would recommend betting on the Vegas underdog 49ers.

ZoltarVsBingThruWeek16

For the season, through week 16, in head-to-head predictions, meaning just predicting the winner of a game (which isn’t particularly useful except for parlay betting), Zoltar has gone 174-66 for 73% accuracy. Bing Predictions is 152-88 for 63% accuracy. By the way, if you simply picked the home team to win, you’d only get about 55% accuracy. And if you just picked the Las Vegas favorite to win, through week 16, you’d be 146-82 for 64.0% accuracy (note that some games, called pick games, don’t have a Vegas favorite).

Update, Monday January 4, 2016: For week 17, in head-to-head predictions (which isn’t terribly useful except for parlay bets), Zoltar went a mediocre 11-5 (69% accuracy). Bing was weak 9-7 (56% accuracy). Against the Vegas point spreads, Zoltar ended up an OK but not great 3-2 (60%) against the spread. Note that the Colts vs. Titans game was a push because the Colts won by exactly 6 points. Zoltar was oh so close to a perfect 5-0 against the Vegas point spread, but both the Chiefs and Broncos missed covering the spread by 1 point each. Arg.

In head-to-head predictions, Zoltar finished the 2015 regular season with a record of 185-71 for 72% accuracy. Bing Predictions finished 161-95 for 63% accuracy.

This entry was posted in Machine Learning. Bookmark the permalink.