Last week was a brutal one. 0-5. Sorry if you took my advice. For the season we are 11-13-1. The one thing I have learned in the last 30 years of football prognostication (yes, I was predicting football outcomes when I was 10 years old!) is that if you can’t put in the time to do the proper research, best not to share your predictions or make predictions at all! Last weekend was heavily occupied with family events and of course family comes first (even before football!). Well this weekend I am in Chicago for work and a bit of pleasure with no family around. That means I have more time to dedicate to football research.
We made some revisions to our algorithm and recently in my graduate program in data science at Northwestern (which explains why I am in Chicago this weekend), you should always be revising your model when new data is available. Well we have an extra week of NFL data which is 25% more than the in-season data we used last week. We are also at the point in the season where we are relying less on pre-season expectations and leaning more heavily on what is happening this season. Any fluky events should have worked themselves out and we are now in the “what you see is what you get” point in the NFL season. Historically, the middle of October has been good to us with more reliable data at our disposal so we hope that trend continues.
Still there is a good deal of variance we will experience with the NFL season so we should take note of that. In order to account for that variance, our revised model uses three techniques for finding the best value for the NFL games this weekend based on the point spread. Here are explanations of those three techniques and we will try to keep them succinct (even though it is hard to hold back our NFL geekdom!).
- We just started to use regression technqiues to predict the point differential for each game and compare that differential to the actual point spread. This was done by comparing the point differentials for all 78 games played so far in the NFL and to the the differntial in DVOA from FootballOutsiders.com or FO. For those not familiar with DVOA, it stands for Defense-adjusted Value Over Average and it is a really good metric for NFL team evalution. Here are a couple of links if you wish to seek a thorough explanation of DVOA.
We started Week 1 by computing the differntial in DVOA between both teams in each game using the pre-season expectations on FO and correlating it to the point differentialg those games. Then in Week 2, we adjusted the DVOA ratings by using 1/8 of the current DVOA rating combined with 7/8 of the pre-season DVOA rating. As each week passes by, we increase the weight of the current DVOA rating by 1/8 and decrease the pre-season DVOA rating by 1/8. For example, we have 5 weeks of current season data so the weight for the current DVOA is 5/8 and pre-season DVOA is 3/8. By the time we get to the midway point of the NFL season, we will exclusively be using the current DVOA rating. We also removed any influential data points so that our linear regression model is more robust. Any point differentials that were over 30 points were discared. This has helped establish a linear model that more closely mirrors what we should expect from most games. The plan is to compare our model to the actual point spread and select the games where our model diverges the most from the point spread. Of course we will choose the team where the divergence works to our advantage.
The 2nd technique is to rank the points spreads for each game and to also rank the differntial in DVOA differntial for each game. For example, the Minnesota Vikings are favored by 10.5 points this weekend so we should expect a high differential in DVOA between them and the Arizona Cardinals. We will focus on games that have a huge discrepancy in those two rankings.
Finally, we will attempt to normalize the point spreads and the DVOA differentials and compute a z-score for both of those values. Then we take the difference in the z-scores to see which games experience the great difference.
Based on those three techniques and NFL point spreads published by Westgate, here is what we found to be the 5 best teams to roll with this weekend.
- LA Chargers -1
- Miami +3 (Sorry Bears fans! Even being in Chicago won’t make me change my mind about this game.)
- LA Rams -6.5
- Baltimore -2.5
- Kansas City +3.5
Now we will try to explain our choices with some reasoning (like the NFL experts do!).
The Chargers have been a funky team this year. Many expected them to be a bona fide Super Bowl contender. 3-2 is not bad but they are not as hot as the two remaining undefeated teams: LA Rams and Kansas City. Interestingly, the Chargers only two losses were to those two teams so perhaps we should not be so hard on them. They barely beat the 49ers but did cover the spread against Buffalo and Oakland. This week they travel to Cleveland and the Browns seemed to have somewhat shaken their losing ways with a couple of wins so far this season. Still, this is the Browns and all we need is for the Chargers to win with a -1 spread. The Browns defense is good but the Chargers have plenty of offensive firepower to counteract whatever defense schemes the Browns throw their way. The Chargers have sorely missed pass rusher Joey Bosa who is still out until at least early November but expect running back Melvin Gordon to continue to dominate the ground game which should help the Chargers control the time of possession. This game will be close but not as close the oddmakers think. Prediction: LA Chargers 27 - Cleveland 20.
If the game were in Chicago, count on me buying a ticket to Solider Field and watching my first game at that historic venue. Well, the Bears take their talents to South Beach after having a week off. The research on how teams perform after a bye week is inconclusive so we won't go there with our analysis of this game. The Dolphins have lost their last two games but those games were on the road against good teams like New England and Cincinnati. The Dolphins are home and happen to be a home dog. Yes, the Bears have played well but our model seems to indicate that the Dolphins will eek out a victory at home. At 3-2, Miami is not that bad of a team and should not be a home dog. If anything, this game should be a pick ‘em. Prediction: Miami 23 – Chicago 21.
There is no stopping this LA Rams team and we cannot wait for their game against Kansas City on November 19 on Monday Night Football. Both teams could be 10-0 when this game takes place in Mexico City! Kudos to the NFL schedule makers for having the foresight to schedule such a great game on national TV at an international location and right before the Thanksgivng holiday. For now, the Rams will need to keep up their winning ways and that should happen without any problems in Denver. Yes, the elevation can be tough but this Rams team is built Ford tough (actually wrong car company!). You know what I mean! This is a very, very good football team and while the Broncos have shown some resliency this season, they are no match to the well-roundedness of the LA Rams. Where the Broncos are weak is stopping the run which plays right into the strength of the Rams. Expect running back Todd Gurley to have a field day at Mile High. Prediction: LA Rams 30 - Denver 20.
If we use the transitive property in mathematics, then Baltimore should beat Tennessee without any issues. Think about it ... in Week 1 the Ravens beat the Bills 47 - 3 in Buffalo and last week the Titans lost to the Bills 13 - 12 also in Buffalo. I wish the NFL worked that simply which is why you have the phrase, "On any given Sunday, anything can happen in the NFL." Still, Baltimore is a better a team than Tennessee. Yes, both teams are 3-2 but Baltimore is a legitimate playoff contender whereas Tennessee will probably be an 8-8 team. A spread of -2.5 is not asking much and our model has the Ravens winning by 5 points. This is a veteran Ravens team who knows how to win on the road by beating Pittsburgh by 12 in Pittburgh 2 weeks ago. What is more telling is that the Ravens have the 4th highest defensive DVOA rating while the Titans have the 5th worst offensive DVOA rating. We should have a typical low scoring Ravens game here where defense rules the day. Prediction: Baltimore 20 - Tennessee 13.
We end our blog post with perahps the best game so far in this young NFL season: Kansas City at New England. Again, the NFL schedule makers have done a great job with the schedule and I am eagerly anticipating this matchup. It won't be easy for the Chiefs to stay undefeated when they face the Rams next month since New England has given Kansas City some trouble during the past few years. Playing on the road and on national TV with all of Tom Brady's weapons back will pose a significant challenge for Patrick Mahomes and company. However, we don't need the Chiefs to win to cover. In fact, the Chiefs can lose by a field goal and we cover. While the Patriots have played better recently, that was against mediocre competition. The Chiefs are a completely different story. There is ridiculous talent and speed on this Chiefs team and the Patriots defense simply do not have the defensive talent to slow down this offensive juggernaut. We know Bill Belichick is the master at stopping the other team's best offensive weapon but who do you stop? Karim Hunt? Tyreke Hill? Travis Kelce? Patrick Mahomes? You double or triple one guy, any one of the other three guys will burn you. We know the Chiefs defense is not great so expect a lot of scoring on Sunday night. How the over-under line is not higher than 60 is beyond me. This game will probably come down to whoever has the ball last so we feel good with a +3.5 spread. Actually, we like the Chiefs to win this one so if you are feeling lucky, take the Chiefs on the moneyline. Prediction: Kansas City 33 - New England 31.
OK, there you have it! We went 0-5 this week so we will go 5-0 this week! Good luck and enjoy the games!