Predicting Win Totals (part 1)
I like playing with predictive models. Keep in mind, I'm no Nate Silver. I don't have the time, intelligence, resources, patience... you get it. But I find them interesting. In the interest of getting one out here I whipped one up pretty hastily. I recognize that it has a lot of flaws and needs work, but I'm using it as an excuse to talk about the upcoming season. In coming weeks I will put more effort into really building one I can be proud of (and even use to place a bet or two) but right now, this is just for fun.
When I say it's flawed, I mean that I used a pretty basic prediction model (Random Forest Regression) and a pretty basic feature set. Basically, I trained it on how well the team did in the previous season, and I used data going back to 2020. Here are the features I used for this run:
full_features = [
'total_losses', 'home_wins', 'home_losses', 'away_wins', 'away_losses',
'spread_beats', 'games_favored', 'games_underdog', 'favored_wins',
'favored_losses', 'underdog_wins', 'underdog_losses',
'pts_scored', 'pts_allowed', 'point_diff', 'sos_avg_opp_wins'
]
Basically, using how well you did last year to predict how well you'll do this year. of course, there are a lot of factors not taken into account: player turnover being chief among them. I am working on a player grading system that I plan on incorporating into this model... but that is for the future.
So... with all that said, here are my results.
Model vs. Vegas Win Totals (2025)
Comparison Table
| team | predicted_wins | vegas_win_total | delta |
|---|---|---|---|
| MIN | 13.46 | 8.5 | 4.96 |
| DET | 14.55 | 10.5 | 4.05 |
| KC | 14.62 | 11.5 | 3.12 |
| WAS | 11.66 | 9.5 | 2.16 |
| PHI | 13.62 | 11.5 | 2.12 |
| PIT | 9.97 | 8.5 | 1.47 |
| SEA | 9.92 | 8.5 | 1.42 |
| BUF | 12.91 | 11.5 | 1.41 |
| LAC | 10.53 | 9.5 | 1.03 |
| GB | 10.50 | 9.5 | 1.00 |
| HOU | 10.00 | 9.5 | 0.50 |
| TB | 9.98 | 9.5 | 0.48 |
| LA | 9.97 | 9.5 | 0.47 |
| NO | 4.92 | 4.5 | 0.42 |
| DEN | 9.88 | 9.5 | 0.38 |
| IND | 7.73 | 7.5 | 0.23 |
| BAL | 11.71 | 11.5 | 0.21 |
| ATL | 7.71 | 7.5 | 0.21 |
| NYJ | 4.96 | 5.5 | -0.54 |
| DAL | 6.88 | 7.5 | -0.62 |
| CIN | 8.86 | 9.5 | -0.64 |
| MIA | 7.80 | 8.5 | -0.70 |
| ARI | 7.58 | 8.5 | -0.92 |
| CLE | 2.95 | 4.5 | -1.55 |
| CAR | 4.84 | 6.5 | -1.66 |
| LV | 3.97 | 6.5 | -2.53 |
| TEN | 2.92 | 5.5 | -2.58 |
| NYG | 2.77 | 5.5 | -2.73 |
| CHI | 5.02 | 8.5 | -3.48 |
| JAX | 3.98 | 7.5 | -3.52 |
| SF | 5.87 | 10.5 | -4.63 |
| NE | 3.85 | 8.5 | -4.65 |
Immediately, the teams that stand out to me are Minnesota and San Francisco (No, not New England. I'll get to you, New England...). I am predicting Minnesota to win 5 games more than Vegas is, and I'm predicting San Fran to win nearly 5 games less. This is obviously because of last season's results and not taking into account the QB change in Minnesota and the...something something change in San Francisco.
Wait, can someone remind me why we think San Francisco is going to win over 10 games this season? Is it solely because of Christian McCaffrey and his completely healed (and in no way lingering) bilateral Achilles tendonitis? Is it because in Kyle Shanahan we trust? And, no, I'm not a Purdy hater. I like him but... the team around him has taken a beating and the injuries are already piling up. But then there is...
| Team | Strength of Schedule |
|---|---|
| SF | 32 |
| NO | 31 |
| NE | 30 |
| ATL | 29 |
| CAR | 28 |
| TB | 27 |
| TEN | 26 |
| JAX | 25 |
| BUF | 24 |
| ARZ | 23 |
| MIA | 22 |
| SEA | 21 |
| IND | 20 |
| NYJ | 19 |
| LA | 18 |
| HST | 17 |
| LV | 16 |
| LAC | 15 |
| DEN | 14 |
| CIN | 13 |
| BLT | 12 |
| PIT | 11 |
| KC | 10 |
| DAL | 9 |
| WAS | 8 |
| GB | 7 |
| CLV | 6 |
| CHI | 5 |
| PHI | 4 |
| MIN | 3 |
| DET | 2 |
| NYG | 1 |
The 49ers have the easiest schedule in 2025 according to PFF. That certainly matters. Perhaps they don't need everyone firing on all cylinders to bang out double digit wins...
Minnesota, on the other hand, has the third hardest schedule and they are relying entirely on the unknown x-factor that is J.J. McCarthy. My prediction is obviously not putting a lot of stock into the SOS, and none into the QB situation because I haven't implemented that yet. So I certainly don't trust that the Vikings will win 13 or 14 games, but do I trust them to win more than 8? Yeah, I kinda do.
Now, as for the aforementioned Patriots... my model has them around 4 wins (like last season) but they have brought in Mike Vrabel as head coach and Josh McDaniels as OC... both moves I really like for them. Other than Drake Maye and a few players on defense, they don't really have names on their roster that give opposing teams pause, but that could change after a few weeks of solid coaching and play calling. They also have the third easiest schedule, according to PFF. So is that enough for them to have a winning season this year? I guess it depends on how you feel about their division.
The Bills are still the Bills, and the Jets are still the Jets(?).... that leaves the Dolphins. Are they the 2023 version or the 2022 version. My model is leaning toward 2022 (because it's programmed that way) but so is Vegas and just about every talking head in the NFL-o-sphere. They all have the Dolphins buried in third or last place in the division. And honestly, it's hard to blame them. The Dolphins have a history of having too much faith in an O-Line that underperforms every season. And don't even get me started on that secondary... woof. Frankly, unless there is a massive surprise this season, GM Chris Grier should lose his job for purposefully fielding such a weak unit. Or should I say units because I would throw tight end in there too. Now, if Tua stays healthy (that's a big if these days) and of Tyreek can get his head out of his ass (and even bigger if) then their offense can be powerful enough to out-shoot opponents... but man... right now I blame no one for taking the under on them.
As for the rest of the list. My model's win total for Detroit is too high, but I still think they'll be good. KC, Buffalo and Philly are spot on. I hammer the over on KC every year and I'm never disappointed. Baltimore will probably sail past 11 wins so I like the over there. Pittsburgh and Seattle stick out to me as candidates for regression. In Washington's case, I just want to see if Kingsbury can keep it up, as he never could in Arizona. I still think they'll be good though. Pittsburgh and Seattle have new QB's, both with question marks, so it's hard to confidently hit the over on them.
Then there's the 9.5 club. The Commanders, Chargers, Texans, Packers, Buccaneers, Rams, Broncos, Bengals. Man, I want to go over on all these teams. Is that statistically possible? The Commanders are candidates for regression, but I think they'll still be good. I think right now I'm going to go Over, Under, Over, Over, Over, Under, Over, Over... The Chargers are going to struggle in their division I think... but shouldn't that mean I should take the under on the Broncos too? No... Their defense is stellar and I just trust Sean Payton with the offense. The Rams I also have going under, and mainly because of Stafford's back. I have doubts he's going to play the entire season... but hopefully those doubts are unfounded.
As for the rest... Jacksonville and Dallas will probably win 8 or more games... In fact I'm strangely confident Dallas will. Less confident about Jacksonville. Anyone confident in Jacksonville is projecting for Lawrence. That's not necessarily a wrong thing to do, but let's be honest... he hasn't shown it yet. Yet.
Giants under. Browns under (They are going to fuckin tank). Saints over??? I don't know why but I think they're not going to be the dumpster fire we're all convinced they will be. Colts under. Bears over (now who's projecting?).
Okay, that's enough of that. I'm currently working on V2 of my model so I'll revisit this when it's done.