2020 NFL Game Predictions: Week 17, Playoffs, Best Team

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 16: 9-7 (RF); 10-6 (LR)
• Week 15: 11-5 (RF); 10-6 (LR)
• Week 14: Playoffs
• Week 14: 7-9 (RF); 9-7 (LR)
• Week 13: 10-5 (RF); 11-4 (LR)
• Week 12: 13-3 (RF); 12-4 (LR)
• 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: 0-0 (RF); 7-8-1 (LR)
• Week 2: 0-0 (RF); 8-8 (LR)
• Week 1: Kickoff

________________________________________________________

Recap:

13-3 (LR) with 10.887 expected; 13-3 (RF) with 10.418 expected.  The models did better than expected. Using the LR probabilities, there was about a 19% chance of getting 13 or better games correctly predicted.  538 QB missed 4 while 538 traditional missed 5.  Personally I missed 3.  

• Raiders did beat Broncos.  My models and myself correct; 538 incorrect.  Broncos lose by 1 in a back and forth game.
• Rams beat the Cardinals.  My models and 538 traditional correct; 538 QB and myself incorrect.  Cardinals got the first score but never scored again as they miss the playoffs.  Cardinals top QB was injured early in the game, missing a lot of the rest of it and certainly not playing as well as he would have if uninjured.
• Steelers lose to Browns.  Myself and 538 QB correct; my models and 538 traditional incorrect.  Steelers did not start their top QB to rest him and others for the playoffs.  Browns had a huge lead early in 4th quarter and held on to win despite a comeback by the Steelers.  Browns make it into the playoffs and will be playing the Steelers in the Wild Card round.
• Washington beats the Eagles.  My models, 538 QB, and myself correct; 538 traditional incorrect.  Washington makes it into the playoffs.
• Chiefs lose to Chargers.  Everybody incorrect.  Chiefs sat their top QB among other starters in order to rest them for the playoffs, so a loss is not really surprising.
• Cowboys lose to Giants.  Everybody incorrect.  Giants took the lead early and never gave it up.  Either team potentially had a shot at the playoffs if Washington lost, but Washington won, so both teams miss out.

On to the playoffs!!

________________________________________________________

Week 17:

Happy New Year! This is the last week before the playoffs.  No changes in the models except to retrain them with the latest data. 

The models disagree with each other on 0 games this week, that is, they are in complete agreement.

With 538 QB, there are 3 disagreements, and with 538 traditional, there are 2 disagreements.  16 games this week as no one has a bye.

Oddities and Bold Predictions:

• Raiders vs. Broncos
• My models pick the Raiders but 538 (QB and traditional) picks the Broncos.  Broncos play at home and are 5-10.  Raiders are 7-8.  Sticking with the Raiders.
• Rams vs. Cardinals
• My models pick the Rams, 538 QB picks the Cardinals, but 538 traditional picks the Rams.  Rams play at home and are 9-6.  Cardinals are 8-7.  Rams are in the playoffs regardless, but Cardinals need to win to get into playoffs.  Rams starting QB is out.  Picking the Cardinals.
• Steelers vs. Browns
• My models pick the Steelers, 538 QB picks Browns, but 538 traditional picks Steelers.  Browns play at home and are 10-5.  Steelers are 12-3.  Browns need to win to stay in playoffs.  Browns are pretty beat up but they have more on the line.  Picking the Browns to win.
• Washington vs. Eagles
• My models pick Washington, 538 QB picks Washington, but 538 traditional picks the Eagles.  Eagles are 4-10-1 and play at home while Washington is 6-9.  Looks like the Eagles best QB will be playing, and so will the best Washington QB.  Washington is playing to stay in the playoffs while the Eagles don't have any playoff chances.  Sticking with Washington.  

An exciting final week as a lot is still at stake for which teams will get into the playoffs.  Good luck!

Here are the predictions for Week 17:

Playoffs:

Here are the latest predictions using the LR model for who will get to the playoffs. 

The AFC top seed remains with the Chiefs.  Bills stay at #2, but Steelers move from #4 to # 3 with last week's win over the Colts.  Colts drop from #3 to #6 with their loss to the Steelers.  Titans move from #6 to #4 as the South Champ, having a better division win % than the Colts.  Ravens are back in the playoffs at #5 with head to head advantage over Colts.  Dolphins remain at #7 with conference win % advantage over Browns, meaning that Browns drop out of the playoffs.

The NFC top seed remains with the Packers.  Saints remain at #2 with conference win % over Seahawks, and Seahawks remain at #3.  Washington stays at #4 with head to head advantage over Cowboys.  Buccaneers move up from #6 to #5 and Rams move down from #5 to #6.  Bears jump back into playoffs at #7 with win % in common games advantage over Cardinals, meaning that the Cardinals are out of the playoffs.  

• AFC
1. Chiefs: West Champ
2. Bills: East Champ, head to head win over Steelers
3. Steelers: North Champ
4. Titans: South Champ, division win % over Colts
5. Ravens: Wild Card #1, head to head over Colts
6. Colts: Wild Card #2
7. Dolphins: Wild Card #3, conference win % over Browns
• NFC
1. Packers: North Champ
2. Saints: South Champ, conference win % over Seahawks
3. Seahawks: West Champ
4. Washington: East Champ, head to head over Cowboys
5. Buccaneers: Wild Card #1
6. Rams: Wild Card #2
7. Bears: Wild Card #3, win % in common games over Cardinals

If the foregoing is true (and I did the tiebreaking correctly), how will the playoffs playout?  Using the LR model for one-time head-to-head matchup predictions:

• Wildcard:
• AFC
• Chiefs (1)-bye
• Bills (2) vs. Dolphins (7) -> Bills (2) - (0.542)
• Steelers (3) vs. Colts (6) -> Steelers (3) - (0.502)
• Titans (4) vs. Ravens (5) -> Titans (4) - (0.594)
• NFC
• Packers (1)- bye
• Saints (2) vs. Bears (7) -> Saints (2) - (0.705)
• Seahawks (3) vs. Rams (6) -> Seahawks (3) - (0.592)
• Washington (4) vs. Buccaneers (5) -> Buccaneers (5) - (0.527)
• Division:
• AFC
• Chiefs (1) vs. Titans (4) -> Chiefs (1) - (0.643)
• Bills (2) vs. Steelers (3) ->  Bills (2)- (0.55)
• NFC
• Packers (1) vs. Buccaneers (5) -> Packers (1) - (0.59)
• Saints (2) vs. Seahawks (3) -> Seahawks (3) - (0.511)
• Conference:
• AFC: Chiefs (1) vs. Bills (2) -> Chiefs (1) - (0.568)
• NFC: Packers (1) vs. Seahawks (3) -> Packers (1) - (0.566)
• Super Bowl:
• Chiefs (1) vs. Packers (1) -> Chiefs (1) - (0.596)

The LR model still predicts that the Chiefs will go all the way and win the Super Bowl against the Packers.

If we run this as a Monte Carlo simulation with the seeding fixed and 20,000 runs, we get the following result:

The Chiefs still win most often at 24% of the time (down from 25%), followed by the Packers at 20% (up from 18%) and the Bills at 9%.  

Importance of Getting the #1 Seed

How important is getting the #1 seed in the playoffs?  The Bills (see below) are predicted to be a slightly better team than the Packers, and yet, the Packers have an 11% Super Bowl win percentage-point advantage over them.  If I switch the Bills to the #1 seed and the Chiefs to #2 seed, the Bills' probability of a Super Bowl victory increases from 9% to 17% (8 percentage point increase), while the Chiefs' probability decreases from 24% to 14% (10 percentage point decrease).  Likewise, switching the Saints to the #1 seed and the Packers to the #2 seed increases the Saints from 8% to 13% (5 percentage point increase) and decreases the Packers from 20% to 12% (8 percentage point decrease).  If I swap the Bills and the Packers so that the Bills are the #1 seed in the NFC and the Packers are the #2 seed in the AFC, then the Bills have a 20% chance of Super Bowl victory (11 percentage point increase) while the Packers have only an 8% chance (12 percentage point decrease).  That is, they essentially swap percentages!

Getting the #1 seed and a first round bye appears to increase a team's chance of winning the Super Bowl by 8 to 10 percentage points, which amounts to roughly doubling the chances of a Super Bowl victory for that team.  Together, the number #1 seeds have a 44% chance of one of them winning the Super Bowl.  Clearly, getting the #1 seed makes it much more likely that the #1 seeded team will win the Super Bowl.

Perhaps this should not surprise us mathematically.  If we assume each team has a 50% chance of winning any particular game, then a #1 seed has to win 3 games to win the Super Bowl, which gives it a 50% *50% *50% = 12.5% chance of winning.  Seeds 2-7 have to win 4 games, which gives them each a 50% *50% *50%*50% = 6.25% chance.  That is, all other things being equal, a #1 seed has double the chance of winning the Super Bowl that a 2-7 seed has, or a 6.25 percentage point advantage.

This advantage is increased by how good the #1 seed is with respect to the other teams.  For example, we saw that the Packers gain much more from being the #1 seed than the Saints do.  This is because the Packers have an average probability against all teams of 69% (according to LR), whereas the Saints have a 64% average probability.  Thus, for each game played, the Packers get to keep more of their probability towards winning the Super Bowl than the Saints do (i.e., they keep 69% of their likelihood, while the Saints only keep 64%).  With only 3 games, the Packers would have a 0.69^3 = 33% chance of winning the Super Bowl.  The Saints would comparatively have a 0.64^3 = 26% chance.  But the Saints also have to play one additional game, meaning that they only have a 0.64^4 = 17% chance of winning the Super Bowl.  They lose 36% of their likelihood because they have to play an additional game, and would lose even more if they were a worse team.

To recap, getting the #1 seed increases a team's chances of winning the Super Bowl by at least 6.25 percentage points.  Also, the better the #1 seed is compared to the other playoff teams, the greater this advantage is.  Since the #1 seed is guaranteed to be the "best" team in the conference from a wins/loss perspective, the advantage is likely to be significant.  Together, having one fewer game to play combined with being the "best" team in the conference means that the #1 seed has a much greater likelihood of winning the Super Bowl than would occur in a randomized head-to-head matchup playoff system with no team having a bye.  Not to mention that the #1 seed always has home field advantage and always gets to play the lowest ranked seed (i.e., the supposedly "worst" team) available, which is another advantage. 

The playoff system is designed to give the "best" two teams every advantage to get to and win the Super Bowl.  Perhaps this is for the best as we want the "best" team to win the Super Bowl.  But it does also feel like it is perhaps too biased in the #1 seeds' favor as we also want the playoffs to be competitive and to give every playoff team a fair chance to win the Super Bowl.  As shown below, the "best" teams aren't that much better than other teams, and the advantage given to them in the playoffs can feel a bit disproportionate.  "For whoever has, to him more shall be given; and whoever does not have, even what he has shall be taken away from him." (Mark 4:25).  This most certainly is true in the current NFL playoff system, but maybe it is the best compromise of our conflicting desires to have the "best" team win the Super Bowl while still having a competitive playoff system.

Best Team:

What is the "best" team?  If each team were to play every other team once, which team would have the most wins?

By summing up the probabilities of each game matchup, the 32 game season would have a final top 5 rank of:

1. Chiefs
2. Bills
3. Packers
4. Saints
5. Seahawks

Chiefs remain on top. Bills move from 6 to 2.  Packers move from 2 to 3.  Seahawks move from 3 to 5.  Saints move up from 7 to 4.  Titans move from 4 to 6.  Colts get pushed from 5 to 7.

By summing the outcome of each game matchup (1 point per win), the final top 5 rank would be:

1. Chiefs
2. Bills
3. Packers
4. (4) Steelers
5. (4) Dolphins

Chiefs stay on top.  Bills move from 6 to 2.  Packers move from 2 to 3.  Steelers move from 11 to 4.  Dolphins move from 7 to 4.  Seahawks move from 3 to 6, and Colts move from 4 to 6.

The difference between these two lists is striking this week.  While the top 3 are the same, 4-7 are very different.  The Saints are judged to have the 4th highest expected value of games won, but are 9th in total wins.  This means that the Saints are predicted to win less often, but when they are predicted to win, they are much more likely to win.  This suggests that they match up really well against some teams, but are mediocre against others.  This is in contrast with the Steelers and Dolphins who are ranked lower in expected value, but higher in total wins.  These teams are predicted to win a lot of games, but only barely.  

An extreme of this paradoxical situation would be if a team was predicted to win each game it played with 50.1% probability.  Then in the list below, it would have an expected value of 16.032 games won, placing it in the middle of that list, but a total wins value of 32, placing it at the top of that list.  A team with this sort of probability is obviously not going to win every single game.  Thus, the expected value is probably a much better judge of how good a team is.

For the top three teams, the expected value and total win ranks are in agreement.  In short, the Chiefs still remain the "best" team.  The Bills move up to second and the Packers move down to third.

2020 NFL Game Predictions: Week 16, Playoffs, Best Team

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 15: 11-5 (RF); 10-6 (LR)
• Week 14: Playoffs
• Week 14: 7-9 (RF); 9-7 (LR)
• Week 13: 10-5 (RF); 11-4 (LR)
• Week 12: 13-3 (RF); 12-4 (LR)
• 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: 0-0 (RF); 7-8-1 (LR)
• Week 2: 0-0 (RF); 8-8 (LR)
• Week 1: Kickoff

________________________________________________________

Recap:

 10-6 (LR) with 10.519 expected, and 9-7 (RF) with 9.937 expected.  So the models performed as expected.  I personally go 12-4.  538 QB goes 11-5, 538 traditional goes 12-4.

• Seahawks beat the Rams.  LR, myself, and 538 get this correct; RF misses this.  Rams got the first score, but Seahawks were able to tie things up by the half and then took the lead without losing it again in the second half.
• Chargers beat the Broncos.  538 and myself correct; LR and RF incorrect.  Chargers took the lead early, but Broncos came back to tie it up near the end.  Chargers finished the game with a field goal to win.
• Steelers win against Colts.  538 traditional and myself correct; 538 QB, LR, RF incorrect.  Steelers were down 21-7 at the half.  Steelers got 3 touchdowns in second half to put them over the top 28-24.  
• 49ers beat Cardinals.  Everyone misses.  Cardinals got the first score, but 49ers quickly took the lead and never let go.
• Jets upset the Browns by winning their second game.  Everyone misses. Browns scored first but Jets took the lead and then never gave it up.
• Bengals beat the Texans.  Everyone misses.  Bengals took the lead first but Texans kept tying it up.  Texans even took the lead halfway through the 4th quarter, but Bengals came back with a TD to take the lead and a field goal to cap it off.
• Panthers beat Washington.  Everyone misses.  Panthers got up 20-3 in the first half, and though they didn't score anymore, they held on to win 20-13.  

On to the last week!

________________________________________________________

Week 16:

Merry Christmas! Two weeks left before the playoffs!  No changes in the models except to retrain them with the latest data.  However, another code update was required to handle some new bad data that was giving incorrect results for starter information.

The models disagree with each other on one game, with 538 on another, and 538 models split on a third.  So pretty consistent this week.

16 games this week as no one has a bye.

Oddities and Bold Predictions:

• Seahawks vs. Rams
• LR predicts a Seahawks victory while RF predicts a Rams victory.  538 (QB and traditional) pick Seahawks.  Seahawks play at home and are 10-4, while Rams are 9-5.  Rams won the last matchup in LA.  Rams lost to Jets last week.  Picking the Seahawks.
• Broncos vs. Chargers
• LR and RF barely pick Broncos, but 538 (QB and traditional) barely picks Chargers.  Chargers play at home.  Both teams are 5-9.  Broncos won the last matchup by 1 point. Looking at the statistics, Chargers seem to be the better team on paper.  Picking the Chargers.
• Colts vs. Steelers
• LR and RF pick Colts.  538 QB adjusted picks Colts as well while 538 traditional picks Steelers.  Steelers play at home and are 11-3.  Colts are 10-4.  Steelers have lost 3 in a row, including last week to the Bengals.  Colts have won the last 3 games.  Looking at the statistics, Colts appear to be the better team, but they also haven't played as strong a schedule as the Steelers.  Hmmm. I could go either way, but I feel like I should pick the Steelers.  I don't think they will lose four in a row...

Good luck!

Here are the predictions for Week 16:

Playoffs:

Here are the latest predictions using the LR model for who will get to the playoffs. 

Changes in AFC include Steelers moving to seed 4 (lost to Bills, predicted loss to Colts), and Bills take second seed.  Colts jump to third seed with a predicted win over Steelers.  Browns remain at seed 5 with head to head win over Titans.  Titans move down to seed 6.  Dolphins remain seed 7 with conference win % advantage over Ravens.

In the NFC, Packers remain on top.  The Saints stay at the second seed with conference win % advantage over Seahawks.  Seahawks jump to third seed.  Washington remains fourth seed.  Rams go to 5th seed with head to head record over Buccaneers.  Buccaneers are 6th seed, and Cardinals remain as 7th seed.

• AFC
1. Chiefs: West Champ
2. Bills: East Champ, head to head win over Steelers
3. Colts: South Champ, predicted win over Steelers
4. Steelers: North Champ
5. Browns: Wild Card #1, head to head over Titans
6. Titans: Wild Card #2
7. Dolphins: Wild Card #3
• NFC
1. Packers: North Champ
2. Saints: South Champ, conference win% over Seahawks
3. Seahawks: West Champ
4. Washington: East Champ
5. Rams: Wild Card #1, head to head record over Buccaneers
6. Buccaneers: Wild Card #2
7. Cardinals: Wild Card #3

If the foregoing is true (and I did the tiebreaking correctly), how will the playoffs playout?  Using the LR model for one-time head-to-head matchup predictions:

• Wildcard:
• AFC
• Chiefs (1)-bye
• Bills (2) vs. Dolphins (7) -> Bills (2) - (0.516)
• Colts (3) vs. Titans (6) ->Colts (3) - (0.512)
• Steelers (4) vs. Browns (5) -> Browns (5) - (0.526)
• NFC
• Packers (1)- bye
• Saints (2) vs. Cardinals (7) -> Saints (2) - (0.573)
• Seahawks (3) vs. Buccaneers (6) -> Seahawks (3) - (0.633)
• Washington (4) vs. Rams (5)-> Washington (4) - (0.510)
• Division:
• AFC
• Chiefs (1) vs. Browns (5)-> Chiefs (1) - (0.632)
• Bills (2) vs. Colts (3) ->  Colts (3) - (0.513)
• NFC
• Packers (1) vs. Washington (4) -> Packers (1) - (0.589)
• Saints (2) vs. Seahawks (3) -> Seahawks (3) - (0.589)
• Conference:
• AFC: Chiefs (1) vs. Colts (3) -> Chiefs (1) - (0.601)
• NFC: Packers (1) vs. Seahawks (3) -> Packers (1) - (0.513)
• Super Bowl:
• Chiefs (1) vs. Packers (1) -> Chiefs (1) - (0.590)

The LR model still predicts that the Chiefs will go all the way and win the Super Bowl, now against the Packers.

If we run this as a Monte Carlo simulation with the seeding fixed and 20,000 runs, we get the following result:

The Chiefs still win most often at 25% of the time, followed by Packers (18%) and Seahawks (10%).

Best Team:

What is the "best" team?  If each team were to play every other team once, which team would have the most wins?

By summing up the probabilities of each game matchup, the 32 game season would have a final top 5 rank of:

1. Chiefs
2. Packers
3. Seahawks
4. Titans
5. Colts

Chiefs remain on top.  Packers move up to 2 from 3, Seahawks move from 5 to 3, Saints move down from 2 to 7, Titans stay at 4, and Colts move to 5.

By summing the outcome of each game matchup (1 point per win), the final top 5 rank would be:

1. Chiefs
2. Packers
3. Seahawks
4. Colts
5. Titans

Chiefs stay on top.  Packers move from 3 to 2.  Saints move from 2 to 8.  Seahawks move up to 3.  Colts move to 4 and Titans move to 5 from 4.  Washington moves to 12.  

In short, Chiefs still remain the "best" team, with Packers taking second and Seahawks taking third.

2020 NFL Game Predictions: Week 15, Playoffs, Best Team

 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 14: Playoffs
• Week 14: 7-9 (RF); 9-7 (LR)
• Week 13: 10-5 (RF); 11-4 (LR)
• Week 12: 13-3 (RF); 12-4 (LR)
• 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:

10-6 (LR) with 10.759 expected; 11-5 (RF) with 10.516 expected.  So the models performed as expected, and RF did slightly better than LR.  Personally I went 11-5 as I agreed with RF on each prediction.

• Cowboys beat the 49ers.  LR gets this correct; RF, myself, and 538 miss this.  Cowboys either had the lead or were tied the entire game.
• Falcons lose to Buccaneers after giving up a 17 point lead.  LR misses this, RF, myself, and 538 get this correct.  It could have easily gone the other way.
• Washington loses to the Seahawks.  Seahawks had the lead the whole game, but Washington came on strong in the 4th quarter and nearly came back to win.  Again, it could have easily gone the other way.   LR misses this, RF, myself, and 538 get this correct. 
• Chargers beat the Raiders in a back and forth game decided in OT.  Everybody misses.
• Jets beat the Rams for the biggest upset of the week, thereby ending their winless season so far.  Rams started playing the second half but the Jets' lead was too much to overcome.  Everybody misses.
• Bears beat the Vikings.  Bears had the lead or were tied for the entire game.  Everybody misses.
• Bengals beat the Steelers in another huge upset.  Bengals always had the lead.  Steelers have now lost 3 in a row after being undefeated.  Everybody misses.

On to week 16!

________________________________________________________

Week 15:

Three weeks left before the playoffs!  No changes in the models except to retrain them with the latest data. However, I did update the code to make playoff and matchup predictions, and I also added a little code to handle some bad data that was resulting in a not-so-insignificant issue in the features being created.  Playoff and matchup predictions are below the Week 15 predictions.

The models disagree with each other on three games this week.  RF is in complete agreement with 538 QB and traditional, which are also the same this week.  So much more consistency with only LR disagreeing in three games.  LR has outperformed RF so far.  Will it have some correct surprises this week?

16 games this week as no one has a bye.

Oddities and Bold Predictions:

• Cowboys vs. 49ers
• LR chooses Cowboys and RF chooses 49ers.  538 (QB and traditional) chooses 49ers. Cowboys are 4-9 and play at home.  49ers are 5-8.  Picking the 49ers.
• Falcons vs. Buccaneers
• LR chooses Falcons and RF chooses Buccaneers.  538 (QB and traditional) chooses Buccaneers.  Falcons are 4-9 and play at home.  Buccaneers are 8-5.  Picking the Buccaneers.
• Washington vs. Seahawks
• LR chooses Washington but RF chooses Seahawks.  538 (QB and traditional) chooses Seahawks.  Washington is 6-7 and plays at home.  Seahawks are 9-4.  Picking the Seahawks.

Not sure what my LR model is thinking, but I am siding with RF and 538 on each one.  Good luck!

Here are the predictions for Week 15:

Playoffs:

Here are the latest predictions using the LR model for who will get to the playoffs.  No changes in the AFC.  In the NFC, Packers take the place of the Saints as #1 seed while the Saints go to the #2 seed.  Washington takes the place of the Giants as the #4 seed and so the Giants are out of the playoffs.  Cardinals replace the Vikings as the Wild Card #3, so the Vikings are out.  NFC West currently has 3 of its 4 teams in the playoffs.

• AFC
1. Chiefs: West Champ
2. Steelers: North Champ
3. Bills: East Champ
4. Titans: South Champ
5. Browns: Wild Card #1
6. Colts: Wild Card #2
7. Dolphins: Wild Card #3
• NFC
1. Packers: North Champ
2. Saints: South Champ
3. Rams: West Champ
4. Washington: East Champ
5. Seahawks: Wild Card #1
6. Buccaneers: Wild Card #2
7. Cardinals: Wild Card #3

If the foregoing is true, how will the playoffs playout?  Using the LR model for one-time head-to-head matchup predictions:

• Wildcard:
• AFC
• Chiefs (1)-bye
• Steelers (2) vs. Dolphins (7) -> Dolphins (7) - (0.518)
• Bills (3) vs. Colts (6) ->Colts (6) - (0.503)
• Titans (4) vs. Browns (5) -> Titans (4) - (0.551)
• NFC
• Packers (1)- bye
• Saints (2) vs. Cardinals (7) -> Saints (2) - (0.643)
• Rams (3) vs. Buccaneers (6) -> Rams (3) - (0.509)
• Washington (4) vs. Seahawks (5)-> Washington (4) - (0.514)
• Division:
• AFC
• Chiefs (1) vs. Dolphins (7)-> Chiefs (1) - (0.602)
• Titans (4) vs. Colts (6) -> Titans (4) - (0.55)
• NFC
• Packers (1) vs. Washington (4) -> Packers (1) - (0.524)
• Saints (2) vs. Rams (3) -> Saints (2) - (0.657)
• Conference:
• AFC: Chiefs (1) vs. Titans (4) -> Chiefs (1) - (0.613)
• NFC: Packers (1) vs. Saints (2) -> Saints (2) - (0.503)
• Super Bowl:
• Chiefs (1) vs. Saints (2) -> Chiefs (1) - (0.592)

The LR model predicts that the Chiefs will go all the way and win the Super Bowl.  Definitely some surprises in here along the way, but the final result is not surprising.

If we run this as a Monte Carlo simulation with the seeding fixed and 10,000 runs, we get the following result:

The Chiefs win most often, but its only about a quarter of the time.  Though sometimes small, every team has a chance.

Best Team:

What is the "best" team?  If each team were to play every other team once, which team would have the most wins?

By summing up the probabilities of each game matchup, the 32 game season would have a final top 5 rank of:

1. Chiefs
2. Saints
3. Packers
4. Titans
5. Seahawks

By summing the outcome of each game matchup (1 point per win), the final top 5 rank would be:

1. Chiefs
2. Saints
3. Packers
4. Titans
5. Washington

So the Chiefs are the top pick in both ranks and would be considered to be the "best" team.  It is surprising to me that Washington is ranked so highly by my model and that the Steelers are ranked so low.  But generally speaking, this ranking seems to be more correct than not.  Here are the full results.

2020 NFL Game Predictions: Playoff Predictions (Week 14)

 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 14: 
• Week 13: 10-5 (RF); 11-4 (LR)
• Week 12: 13-3 (RF); 12-4 (LR)
• 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

________________________________________________________

Week 14 Playoff Predictions:

I have managed to resurrect my code and Excel files for predicting who will get to the playoffs.  I am still working on the code for what happens after that, that is, who will win in the playoffs and ultimately win the Super Bowl.

Using my linear regression model to predict the outcome for the games remaining (weeks 14-17) , adding those results to the current results, and implementing (I hope correctly) the tie breaking procedures and other rules for picking the playoff seeds, we have the below playoff spot predictions.  Note that there are 7 spots this year per conference.  Seeds 2-7 in each conference play 3 games in the Wild Card round while seed 1 has a bye.  The 3 winners along with seed 1 play two games in the divisional round.  The two winners play each other in the conference round.  The winners from each conference play each other in the Super Bowl.

Here are the predictions for who will get into the playoffs:

• AFC
1. Chiefs: West Champ
2. Steelers: North Champ
3. Bills: East Champ
4. Titans: South Champ
5. Browns: Wild Card #1
6. Colts: Wild Card #2
7. Dolphins: Wild Card #3
• NFC
1. Saints: South Champ
2. Packers: North Champ
3. Rams: West Champ
4. Giants: East Champ
5. Seahawks: Wild Card #1
6. Buccaneers: Wild Card #2
7. Vikings: Wild Card #3

2020 NFL Game Predictions: Week 14

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 13: 10-5 (RF); 11-4 (LR)
• Week 12: 13-3 (RF); 12-4 (LR)
• 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:

9-7 (LR) with 10.633 expected, and 7-9 (RF) with 9.836 expected.  As I anticipated, not a great week as there was too much disagreement in the models for any to do well.  Personally I went 9-7.  538 goes 11-5 (QB) and 10-6 (traditional).

• Rams beat the Patriots.  LR and myself correct; RF incorrect.
• Bills beat the Steelers. LR correct; RF and myself incorrect.
• Cowboys beat the Bengals.  LR correct; RF and myself incorrect.
• Cardinals beat the Giants.  LR and myself incorrect; RF correct.
• Washington beats the 49ers.  LR and myself correct; RF incorrect.
• Ravens beat the Browns.  LR incorrect; RF and myself correct.
• Bears beat the Texans.  LR, RF, and myself incorrect.
• Buccaneers beat the Vikings.  LR and RF incorrect; myself correct.
• Chargers beat the Falcons.  LR, RF, and myself incorrect.
• Broncos beat the Panthers.  LR, RF, and myself incorrect.
• Eagles beat the Saints.  LR, RF, and myself incorrect.

The models and myself missed three "uncontentious games" in addition to generally splitting on the contentious games.  Given the above, sort of surprised it wasn't worse.

Moving on to next week.

________________________________________________________

Week 14:

Four weeks left before the playoffs!  No changes in the models except to retrain them with the latest data. The models disagree with each other on six games this week, and disagree with 538 on two others.  So half of the games are in contention.  Lots of disagreement in the models so doubtful that it is going to be a good week for any model (unless one model gets really lucky).  16 games this week as no one has a bye.

Oddities and Bold Predictions:

• Rams vs. Patriots
• LR chooses Rams but RF chooses Patriots.  538 chooses Rams (QB and traditional).  Rams play at home and are 8-6.  Patriots are 6-6.  Going with the Rams.
• Bills vs. Steelers
• LR chooses Bills but RF chooses Steelers.  538 chooses Bills (QB and traditional).  Bills are 9-3 and play at home.  Steelers are 11-1.  Steelers have had an easier schedule than Bills.  Will Steelers lose two in a row after being undefeated? I don't see a good reason for betting against them.  Picking the Steelers.
• Cowboys vs. Bengals
• LR chooses Cowboys and RF chooses Bengals.  538 chooses Cowboys (QB) and Bengals (traditional).  Bengals play at home and are 2-9-1.  Cowboys are 3-9.  Both top QBs are out.  Meh.  Picking the Bengals.
• Giants vs. Cardinals
• LR chooses Giants and RF chooses Cardinals.  538 chooses Cardinals (QB) and Giants (traditional).  Giants are 5-7 and play at home.  Cardinals are 6-6.   Giants just beat the Seahawks and have won the past 4 games they have played.  They have had a decently tough schedule.  Cardinals have lost the last 3 and have also had a pretty tough schedule.  Surprised I am saying this, but choosing the Giants.
• Washington vs. 49ers
• LR chooses Washington and RF chooses 49ers.  538 chooses 49ers (QB and traditional).  49ers play at home and are 5-7.  Washington is 5-7 as well.  Washington has won past 3 games, including against the Steelers.  Their schedule has been sort of easy recently except for Steelers.  49ers have been struggling.  Leaning towards Washington (another surprising statement).
• Browns vs. Ravens
• LR chooses Browns and RF chooses Ravens.  538 chooses Ravens (QB and traditional).  Browns play at home and are 9-3.  Ravens are 7-5.  Interesting that Browns have fewer points scored per game than points allowed per game, and yet they are 9-3.  I think a lot of that comes from the first game in which they got blown out by the Ravens 38-6, and another game against the Steelers 38-7.  A lot of their wins are against bad teams.  I think the Ravens have had a slightly tougher schedule.  I think I am going with the Ravens.
• Texans vs. Bears
•  Both models choose Texans but 538 chooses Bears (QB and traditional).  Bears play at home and are 5-7.  Texans are 4-8. Texans have had a tough schedule.  So have Bears.  Bears are on losing streak of 6.  I'm sticking with the Texans.
• Vikings vs. Buccaneers
• Both models choose Vikings but 538 (QB and traditional) chooses Buccaneers.  Buccaneers play at home and are 7-5.  Vikings are 6-6.  Apart from a win over the Packers, I am not very impressed with the Vikings.  Buccaneers beat the Packers as well.  Definitely going with the Buccaneers.

Too many games to weigh in on personally.  Splitting myself between the two models, and siding with 538 on one of the two conflicts with my models.  So I guess I am somewhere in the middle of this week's mess.  Hopefully next week is more straightforward.

Good luck!

Here are the predictions for Week 14:

2020 NFL Game Predictions: Week 13

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 12: 13-3 (RF); 12-4 (LR)
• 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:

11-4 (LR) with expected 9.642, and 10-5 (RF) with 9.677 expected.  The models performed as expected.  Interesting that the RF had a higher expected value than LR, as it has previously always been more conservative in its estimates.  Personally, I went 11-4.

• Lions beat the Bears.  Bears had the lead all the way until late in the 4th quarter when the Bears QB was stripped of the ball, which was recovered by the Lions, leading to a touchdown.  Tough loss.  LR correct; RF and myself incorrect.
• Saints beat Falcons as Saints keep the lead through the whole game.  LR, RF, and myself correct.
• Colts beat Texans.  LR and RF incorrect; myself correct.
• Bills beat 49ers.  LR, RF, and myself correct.
• Seahawks lose to Giants.  Terrible.  From a probability perspective, this was the biggest upset of the week as the models gave an average of 62% likelihood of Seahawks win.  LR, RF, and myself incorrect.
• Titans lose to Browns.  Browns took early lead and never gave it up.  LR, RF, and myself incorrect.  
• Steelers lose to Washington, ending their undefeated season.  Steelers led until halfway through 4th quarter when Washington tied.  Washington took the lead at 2 minutes with a field goal.  An interception led to another Washington field goal in final seconds before the end of the game. Models gave an average of 59% probability of Steelers victory.  LR, RF, and myself incorrect.

Steelers are no longer undefeated, but Jets are still winless.  

On to week 14!

________________________________________________________

Week 13:

Happy Friday!  Due to the Steelers/Ravens game being postponed until Wednesday this past week, the data for the model was not completely available until Friday morning.  Fortunately, as there was no Thursday game (the Ravens were supposed to play the Cowboys but that has been pushed to Tuesday), my predictions will still be made before any game has been played for the week.

No changes in the models except to retrain them with the latest data.  I am starting to think about the playoffs and will hopefully start making some playoff predictions in the next few weeks.

The models disagree with each other on one game this week, and disagree with 538 on two games (QB adjusted) and one (traditional).

15 games this week as Panthers and Buccaneers both have a bye.

Oddities and Bold Predictions:

• Bears vs. Lions
• LR picks Lions but RF picks Bears. 538 picks Bears (QB adjusted and traditional).  Bears play at home and are 5-6 while Lions are 4-7.  Bears QB is injured but currently listed as "probable" (though he was "doubtful") and Lions QB is injured and listed as "questionable" (but he played last week).  The model data interprets this based on each team's injury history that Bears will be without their starting QB but Lions will have theirs.  LR relies on the QB presence more heavily than RF, which is why I think it swings to favoring the Lions in the absence of the Bears' top QB whereas RF isn't as easily influenced by this and sticks with the Bears.  Bears won the last matchup but are on a losing streak.  Bears have kept their losses closer than Lions have.  It may depend on which QBs show up to play, but I'd pick the Bears to win this if both top QBs are in the game.  However, latest reports say backup Bears QB is starting while Lions top QB is starting.  I have to pick one so.... going with the Bears.
• Saints vs. Falcons
• Both models choose Saints.  538 QB adjusted picks Falcons but traditional picks Saints. Falcons play at home and are 4-7 while Saints are 9-2.  Saints starting QB is injured but Falcons QB is healthy.  I'm still picking the Saints.
• Texans vs. Colts
• Both models choose Texans.  538 QB adjusted picks Colts but traditional picks Texans.  Texans are at home and are 4-7 while Colts are 7-4.  Colts QB is "questionable"  Apart from beating the Packers, the Colts haven't beaten any really good teams.  They did beat Titans once but lost last week to them so that doesn't help.  Texans haven't beaten any really good teams.  If Colts starting QB plays I'd probably pick them, and the latest reports say he will.  Picking the Colts.
• Bills vs. 49ers
• Both models choose Bills.  538 QB adjusted picks Bills as well but traditional picks 49ers.  49ers play at home and are 5-6 while Bills are 8-3. 49ers former starting QB is on injured reserve, but current QB starter is not injured and is scheduled to play.  Bills starting QB is "probable".  49ers lineup is pretty injured.  Bills have done better against a comparatively similar schedule.  Sticking with the Bills.

 Good luck!

Here are the predictions for Week 13: