2020 NFL Game Predictions: Week 12

Welcome!  I intend this to be an ongoing project of predicting NFL game outcomes, point spreads, and final scores.  My hope is that my models will improve over time as they become more sophisticated and use better data.  I will try to regularly publish predictions to keep myself accountable, and will make observations along the way about what is working and what isn't.  See below for latest updates.  Enjoy!

Previous 2020 Predictions:

• Week 11: 6-8 (RF); 4-10 (LR)
• Week 10: 8-6 (RF); 8-6 (LR)
• Week 9: 10-4 (RF); 12-2 (LR)
• Week 8: 9-5 (RF); 9-5 (LR)
• Week 7: 9-5 (RF); 10-4 (LR)
• Week 6: 8-6 (RF); 8-6 (LR)
• Week 5: 9-5 (RF); 11-3 (LR)
• Week 4: 11-4 (RF); 11-4 (LR)
• Week 3: 7-8-1 (LR)
• Week 2: 8-8 (LR)
• Week 1: Kickoff

________________________________________________________

Recap:

12-4 (LR) with 10.21 expected, and 13-3 (RF) with 9.297 expected.  I personally go 13-3.  A very good week for predictions, but a weird week with Thanksgiving, changing game schedules, and Covid 19 illness impacts.

• Washington beats Cowboys as I thought they would.  RF gets this but LR misses.
• Ravens lose to Steelers.  Both models miss this as I thought they would.  The game was originally scheduled for Thanksgiving, which is why it shows up that way in my data.  Due to Covid 19 illnesses in both teams but primarily in the Ravens, it was moved to Sunday, then to Tuesday, then to Wednesday.
• Falcons crush Raiders 43-6, to my surprise.  Both models get this right.
• Saints beat Broncos as I thought they would and as my models picked.  Saints rookie QB played instead of their starter, but Broncos didn't even have a QB as all four Bronco QBs were disqualified due to Covid 19 restrictions.  Their QB was actually a practice squad wide receiver that was promoted a couple hours before the game.  Not a surprising end result given the circumstances.
• Giants beat the Bengals as I thought they would and as my models predicted.
• Colts lose to Titans, and Rams lose to 49ers in an exciting back and forth game with a field goal for the win.  Misses for both models and myself.  

On to next week which starts today!  Time to get the next round of predictions ready.

________________________________________________________

Week 12:

Happy Thanksgiving! A pretty bad round of predictions last week, but hoping it was an unlucky probabilistic fluke.  No changes in the models except to retrain them with the latest data. The models disagree with each other on one game this week, and disagree with 538 on three games (QB adjusted) and 2 (traditional).

538's QB adjusted predictions (default) factor in a scoring of QBs, which impacts the overall traditional prediction.  As quoted here, " Our quarterback-adjusted Elo model incorporates news reports to project likely starters for every upcoming game and uses our quarterback Elo ratings to adjust win probabilities for those games. A team’s current quarterback adjustment is based on possible starters in its next game and how much better or worse that QB is than the team’s top starter. "

As you can see based on the below predictions, most of the disagreement between my models and 538's QB adjusted predictions come down to whether the team's best QB is starting or injured and how good the QBs are.  My models factor this in as well, but they don't appear to have as much influence as they do in 538's models.

16 games this week as no team has a bye.

Oddities and Bold Predictions:

• Cowboys vs. Washington
• LR says Cowboys but RF says Washington.  538 says Washington (QB adjusted) and Cowboys (traditional).  Both are 3-7 and Cowboys play at home.  Cowboys are without their best QB but Washington has theirs. I'd go with Washington.
• Ravens vs. Steelers
• Both models surprisingly predict a Ravens victory, while 538 predicts Steelers (QB adjusted and traditional).  Steelers are 10-0 and Ravens are 6-4.  Steelers play at home.  Injured QBs are not a factor here.  Not sure what my models see that I don't, but going with the Steelers.
• Falcons vs. Raiders
• Both models pick Falcons while 538 picks Raiders (QB adjusted) and Falcons (traditional).  Falcons are 3-7 and Raiders are 6-4.  Falcons play at home.  Both top QBs are playing.  I think 538's model practically says that the Falcons are a better overall team but the Raiders have an even better QB, which is why the model changes when factoring in the QB. Personally going with Raiders.
• Saints vs. Broncos
• Both models pick Saints while 538 picks Broncos (QB adjusted) and Saints (traditional).  Saints are 8-2 and Broncos are 4-6.  Broncos play at home.  The Saints starting QB is injured, which, after looking at the stats, is the only reason that makes sense to me why one would pick the Broncos to win.  Not sure why 538 picks Broncos other than the replacement Saints QB is a rookie (and so has a low starting QB score).  Sticking with the Saints.
• Giants vs. Bengals
• Both models pick Giants while 538 picks Giants (QB adjusted) and Bengals (traditional).  Bengals are 2-7-1 and play at home.  Giants are 3-7.  Bengals best QB is injured.  Giants have their starting QB and are coming off a bye.  Sticking with Giants.

So staying with my models for 2 and disagreeing on 2 when they disagree with 538.  Good luck and hope you have a great holiday weekend!

Here are the predictions for Week 12:

2020 NFL Game Predictions: Week 11

 Welcome!  I intend this to be an ongoing project of predicting NFL game outcomes, point spreads, and final scores.  My hope is that my models will improve over time as they become more sophisticated and use better data.  I will try to regularly publish predictions to keep myself accountable, and will make observations along the way about what is working and what isn't.  See below for latest updates.  Enjoy!

Previous 2020 Predictions:

• Week 10: 8-6 (RF); 8-6 (LR)
• Week 9: 10-4 (RF); 12-2 (LR)
• Week 8: 9-5 (RF); 9-5 (LR)
• Week 7: 9-5 (RF); 10-4 (LR)
• Week 6: 8-6 (RF); 8-6 (LR)
• Week 5: 9-5 (RF); 11-3 (LR)
• Week 4: 11-4 (RF); 11-4 (LR)
• Week 3: 7-8-1 (LR)
• Week 2: 8-8 (LR)
• Week 1: Kickoff

________________________________________________________

Recap:

4-10 (LR) with 8.583 expected, and 6-8 (RF) with 8.36 expected.  Terrible week, but it's bound to happen every now and then.  In fact, running a simulation of 100,000 weeks with the probabilities given in the LR model, getting a result of 4 games or fewer will occur about 1% of the time, and 6 games or fewer will occur about 12% of the time.  So very rare, but not impossible.

As the models have gotten really lucky in the past, this week they were really unlucky, despite being in agreement largely with each other and with 538.  The disagreements between LR and RF (Bengals vs. Washington and Eagles vs. Browns) both went for RF this week, which is why it did two games better.  This is the first time RF did better than LR, although it is still behind LR overall.

• Seahawks beat Cardinals, thankfully, against my models and my personal prediction.  It was a back and forth game and could have easily gone the other way.
• Bengals lose to Washington, and Eagles lose to Browns.  RF picks up both predictions and LR misses both.
• Dolphins lose to Broncos, Lions get shut out by Panthers (ouch), Packers lose to Colts in OT, Patriots lose to Texans, Ravens lose to Titans, Vikings lose to Cowboys (really?), and as I thought might happen, Buccaneers lose to Rams.

Jets are without a win at 0-10.  Steelers are still undefeated at 10-0.  NFC East has no one over .500 as each team has only won 3 games.  AFC North and NFC West both have 3 teams over .500.

On to next week!

________________________________________________________

Week 11:

No changes in the models except to retrain them with the latest data. The models disagree with each other on two games this week, and disagree with 538 on one.  So predictions are pretty consistent.

14 games this week. Bills, Bears, Giants, and 49ers have a bye.

Oddities and Bold Predictions:

• Bengals vs. Washington
• Bengals predicted to win by LR but lose by RF.  Bengals play away.  538 picks Washington.  Washington is 2-7 and Bengals are 2-6-1.  I was sort of leaning to Bengals victory as they appear to have kept games closer than Washington.  But Alex Smith is back as QB for Washington and has kept the past two games close.  I don't know.
• Eagles vs. Browns
• Eagles predicted to win by LR but lose by RF. Eagles play away. 538 picks Browns.  Browns are 6-3 and Eagles are 3-5-1.  Eagles have played better teams but have been very inconsistent.  Going with the Browns.
• Cardinals vs. Seahawks
• Both models predict a Cardinals victory, but 538 predicts a Seahawks.  I hope 538 is correct.  Cardinals play away.  Both are 6-3.  Cardinals won the last round in overtime and have been playing well.  Seahawks have not been looking good on both offense and especially defense.  I'm inclined to predict Cardinals victory.
• Buccaneers vs. Rams
• The models and 538 agree on a Buccaneers victory.  Buccaneers play at home. Not inclined to disagree but a Rams victory wouldn't surprise me.

Good luck!

Here are the predictions for Week 11:

2020 NFL Game Predictions: Week 10

Welcome!  I intend this to be an ongoing project of predicting NFL game outcomes, point spreads, and final scores.  My hope is that my models will improve over time as they become more sophisticated and use better data.  I will try to regularly publish predictions to keep myself accountable, and will make observations along the way about what is working and what isn't.  See below for latest updates.  Enjoy!

Previous 2020 Predictions:

• Week 9: 10-4 (RF); 12-2 (LR)
• Week 8: 9-5 (RF); 9-5 (LR)
• Week 7: 9-5 (RF); 10-4 (LR)
• Week 6: 8-6 (RF); 8-6 (LR)
• Week 5: 9-5 (RF); 11-3 (LR)
• Week 4: 11-4 (RF); 11-4 (LR)
• Week 3: 7-8-1 (LR)
• Week 2: 8-8 (LR)
• Week 1: Kickoff

________________________________________________________

Recap:

8-6 (LR) with 9.309 expected and 8-6 (RF) with 8.527 expected.  Not a great week as I would have hoped for 9-5, but it balances out with last week.  538 gets 9-5.  My personal inclinations get 10-4.

• Titans lose to Colts.  Both models miss.
• Bills lose to Cardinals.  Both models miss.
• Broncos lose to Raiders. LR missed this but RF got this.  I personally thought this would happen.
• Lions beat Washington.  LR got this correct but RF missed this.  My personal pick of Lions was correct.
• Eagles lose to Giants.
• Ravens lose to Patriots.
• Seahawks lose to Rams as I personally suspected they would.  Both models missed this.
• Vikings beat the Bears as the models and I personally thought they would.

Jets have yet to win a game and Steelers are still undefeated.

On to next week!

________________________________________________________

Week 10:

Good showing last week with both models.  No changes in the models except to retrain them with the latest data. I fixed a minor bug in code that projected the starters based on current injuries.  The models disagree with each other on two games this week, and disagree with 538 on two others.

14 games this week. Chiefs, Cowboys, Falcons, and Jets have a bye.

Oddities and Bold Predictions:

• Broncos vs. Raiders
• LR picks Broncos but RF picks Raiders.  538 picks Raiders at 64%.  Raiders play at home.  Raiders are 5-3 and Broncos are 3-5.  Raiders have beaten the Saints, Chiefs, and Browns.  Broncos have only won against moderate or weak teams.  Definitely picking the Raiders.
• Lions vs. Washington
• LR picks Lions but RF picks Washington. 538 picks Lions at 53%.  Lions are 3-5 and Washington is 2-6.  Lions play at home.  Picking the Lions.
• Seahawks vs. Rams
• Both models pick Seahawks but 538 picks Rams at 50% (essentially a toss up).  Rams are 5-3 and Seahawks are 6-2.  Rams play at home.  Looks like an offense (Seahawks) vs. defense (Rams) matchup.  Seahawks are pretty banged up and Rams had a bye last week.  Seahawks have definitely had a tougher schedule so far, but they didn't look very good last week.  I hope I am wrong but I am sort of leaning towards the Rams given the bye week and the current condition of the Seahawks team.
• Vikings vs. Bears
• Both models pick Vikings but 538 picks Bears at 56%.  Bears play at home.  Bears are 5-4 and Vikings are 3-5.  It seems like a pretty good matchup.  Bears have lost the last couple while Vikings have won the last couple and they look healthier.  I'm kind of leaning towards the Vikings.

Good luck!

Here are the predictions for Week 10:

2020 NFL Game Predictions: Week 9

    Welcome!  I intend this to be an ongoing project of predicting NFL game outcomes, point spreads, and final scores.  My hope is that my models will improve over time as they become more sophisticated and use better data.  I will try to regularly publish predictions to keep myself accountable, and will make observations along the way about what is working and what isn't.  See below for latest updates.  Enjoy!

Previous 2020 Predictions:

• Week 8: 9-5 (RF); 9-5 (LR)
• Week 7: 9-5 (RF); 10-4 (LR)
• Week 6: 8-6 (RF); 8-6 (LR)
• Week 5: 9-5 (RF); 11-3 (LR)
• Week 4: 11-4 (RF); 11-4 (LR)
• Week 3: 7-8-1 (LR)
• Week 2: 8-8 (LR)
• Week 1: Kickoff

________________________________________________________

Recap:

12-2 (LR) with 9.084 expected.  10-4 (RF) with 7.972 expected.  The models did much better than expected this week.  Definitely got lucky.

• Bills beat the Seahawks (correct for LR but incorrect for RF).  
• Chargers lose to Raiders (incorrect for LR but correct for RF). 
• Falcons beat the Broncos (correct for LR but incorrect for RF).  
• Titans beat the Bears (correct for LR but incorrect for RF).  
• Dolphins beat Cardinals (correct for both).    
• Saints beat Buccaneers (correct for both).  
• Washington loses to Giants (incorrect for both).

538 misses 3 this week.  I personally missed 3 (counting indecision as 0.5).  The surprise this week is that there weren't any real surprises, as the clear majority of games came out as expected and losses were from relatively evenly matched teams.  A couple games looked like they were going to be surprises/upsets as the favorites were behind, but they virtually all caught up in the end to win.

The linear regression model continues to outperform the random forest.

On to next week!

________________________________________________________

Week 9:

A busy and exciting week with the election and other priorities.  No changes in the models except to retrain them with the latest data.  The models disagree on four games this week, so it should be a real test of which model predicts better as they have been mostly in agreement to this point.  Also, lots of disagreement with 538, so another opportunity for distinguishing my models from theirs (for good or ill).  14 games this week. Bengals, Browns, Eagles, and Rams have a bye.

Oddities and Bold Predictions:

• Bills vs. Seahawks
• The LR model predicts a Bills victory while RF predicts Seahawks.  538 chooses Seahawks.  Seahawks are 6-1 and Bills are 6-2.  Seahawks have had a pretty challenging schedule, but so have the Bills.  I can see why the models disagree.  No clear rational choice to me, but as a fan, personally choosing to support the Seahawks. Go hawks!
• Chargers vs. Raiders
• LR chooses Chargers while RF chooses Raiders.  538 chooses Raiders.  Chargers are 2-5 and Raiders are 4-3.  The teams played are fairly similar, and those in common the Raiders have won more of.  I'd go with the Raiders.
• Falcons vs. Broncos
• LR chooses Falcons while RF chooses Broncos.  538 chooses Falcons.  Falcons are 2-6 and Broncos are 3-4.  Falcons have played and lost to a lot of good teams, but so have Broncos.  I really don't know.
• Titans vs. Bears
• LR chooses Titans while RF chooses Bears.  538 chooses Titans.  Titans are 5-2 while Bears are 5-3.  Both teams have kept games pretty close and have played roughly the same kind of teams.  It's another that I don't really have any strong feeling for.
• Dolphins vs. Cardinals
• Both models choose Dolphins but 538 chooses Cardinals.  Dolphins are 4-3 and Cardinals are 5-2.   Cardinals are coming off a bye week, but Dolphins have played better teams.  I think I will stay with Dolphins.
• Saints vs. Buccaneers
• Both models choose Saints but 538 chooses Buccaneers.  Buccaneers are 6-2 and Saints are 5-2.  Saints beat Buccaneers in the first week but haven't done well against mediocre teams since.  Buccaneers have a better record against slightly better teams, and it looks like they have done better against them as well from a points perspective.  I am inclined to go with Buccaneers.

Six of the 14 games are odd or bold this week, so lots of room for upsets and missed predictions.  Good matchups this week.  Good luck!

Here are the predictions for Week 9: