Model Review: NFL Week 9
Something that I have noticed about the NFL when compared to other sports is the complete lack of consistency when it comes to outcomes. I think that this has a lot to do with the fact that teams have an entire week to prepare and strategize for their games when in other leagues its more day to day. Because of that scheduling, NFL teams have an opportunity to rebrand every week while implementing change in other sports takes longer. This week, for the first time all season, I couldn't stop watching the Cowboys game. Somehow, someway the Cowboys nearly pulled off an upset on the undefeated Steelers with a 4th string quarterback. This was a game that I looked at early and thought the Steelers would cover 14.5 points with ease but instead the game came down to the last drive and frankly, the Cowboys should have won if it weren't for a few questionable calls. Queue Skip Bayless for the Cowboys fan reaction:
He's angry but he's not wrong. Anyways, below are the views from yesterday attempting to predict outcomes, I am going to attempt to see how poorly it predicted. I ran each game outcome twice by team in order to have more context around each prediction.
Game 1: Bears 17 @ Titans 24 (Titans cover 6.5)
On score: Clear miss. The machine stated that Tennessee would lose this game
On points: Another miss. The model predicted between 49 and 51 points and to take the over.
Game 2: Broncos 27 @ Falcons 34 (Falcons cover 4)
On score: Another miss. The model is expecting Atlanta to lose which means taking the Bears and the points would be secure- not so much
On points: The points prediction came between 46 and 50. While this is clearly off, one of those predictions was under the point bet and the other was over the point bet so I would have laid off the bet completely.
Game 3: Dolphins 34 @ Cardinals 31 (Cardinals did not cover 6)
On score: The model predicted that the Dolphins were 3 points better, but it also stated that both teams would win and therefore I think you would want to lay off the bet.
On points: It got the over right, it did not predict 65 points but it got the over correct.
Through 3 games: WIN-LOSS-LAYOFF
Game 4: Giants 23 @ Football Team 20 (Washington did not cover 3)
On score: The prediction worked, it stated that New York would win and they were getting points.
On points: Slight miss, the prediction called for the over and the game missed the over by .5.
Game 5: Lions 20 @ Vikings 34 (Minnesota covered 4)
On score: Inconclusive. The model said both teams would win.
On points: Wrong. Predicted the under.
Game 6: Green Bay 34 @ San Francisco 17 (Packers covered 6)
On score: Wrong. Stated 49ers were going to win and were getting points.
On points: Inconclusive. The model predicted the Over/Under number and I would have laid off of it.
Through 6 games: WIN-LOSS-LAYOFF
Game 7: Panthers 31 @ Chiefs 33 (Chiefs did not cover 10)
On score: The model was off but not enough. It would have lead me to take the Panthers and the points.
On points: Wrong. Model was calling for the under.
Game 8: Raiders 31 @ Chargers 26 (Raiders covered 1)
On score: Inconclusive, model stated that the Raiders were better by a point but also predicted both teams would lose
On points: Wrong. Called for the under.
Game 9: Ravens 24 @ Colts 10 (Ravens covered 1.5)
On score: Wrong. The model stated to take the Colts and the points.
On points: Correctly suggested the under.
Through 9 games: WIN-LOSS-LAYOFF
Game 10: Saints @ 38 Bucs 3 (Bucs did not cover 4)
On score: Correct, not accurate. The model predicted the Saints would win but not by a blowout.
On points: Inconclusive. The model predicted 1 over and 1 under.
Game 11: Seahawks 34 @ Bills 44 (Seahawks did not cover 3)
On score: Correct, model suggested taking the Bills and the points.
On points: Inconclusive. The model was wrong but it predicted 1 over and 1 under so I would have laid off.
Game 12: Steelers 24 @ Cowboys 19 (Steelers did not cover 14.5)
On score: The model predicted a Cowboys blowout which is incredibly weird so I need to figure out why that occurred. It did say to take the Cowboys though so that is a win.
On points: Inconclusive. The model predicted over 43 points on both instances but the Steelers instance was too close to the actual line to take.
Game 13: Texans 27 @ Jaguars 25 (Houston did not cover 7)
On score: Inconclusive. The model stated both teams would win.
On points: Completely accurate.
Final score: WIN-LOSS-LAYOFF
Conclusion: Not great. I was hoping towards the beginning that the model would continue to fail, that way I have confidence betting against it. Instead, I come out about even with the spreads and somewhat negative with the points. Next week I am going to re-run this same algorithm and bet for the spread and against the points outputs - watch the results flip.
Data provided by www.pro-football-reference.com
Lines provided by www.bovada.lv