We use team statistics to create models that predict outcomes of NFL games. Our models generate either a win probability for the home team or a point spread, with each model generating a unique prediction. Our models consider many different metrics, such as turnovers, first downs, points scored, pace of play, offensive and defensive efficiency, as well as a few metrics that we have created ourselves. Beyond these statistics, some models consider factors such as “did a team make the playoffs last year?” and “how far is the away team traveling?” See this article for a more detailed breakdown of each model. So far this postseason, our average probabilities have picked 7/8 winners correctly, and Models 2 and 3 have been the most accurate – also going 7 for 8.
Round 3 Matchups
Packers @ Falcons (-5)
While there is not complete agreement among the models, on average we give the Falcons a 57% chance of winning, and our spread model has ATL winning by 3 points. Conversely, Model 1 (47%) and Model 6 (20%) both have GB winning. We pointed out last week that GB has not played well in their last 10 games as a road underdog outside of their division: 3-7 straight up, 4-6 against the spread (ATS). Interestingly, 7 of those 10 games went over the Vegas total, and all 9 ATL home games this year also went over the total. This point is not lost on book makers, as this game has one of the largest totals in NFL history (61 points). Safe to say there should be plenty of points in this one.
Green Bay is seeing support from the models due to their performance vs. playoff teams, previous playoff success, rushing defense, and passing success on 3rd down. The models that favor the Falcons are driven by their advantages in passing offense efficiency, rushing yards per game, yards per play, point differential, more frequent trips to the red zone, and a better red zone defense. The Falcons are the better team on paper, and Matt Ryan should have no trouble picking apart the depleted Packer secondary, but this game could easily come down to whichever team has the ball last.
The table below gives our predicted spread for this matchup, as well as win probabilities for the home team from six different models. The Model 1 value of 0.47 means that this model gives Atlanta an 47% chance to win (and Green Bay a 53% chance)
| Home | Away | Our Spread | Mod 1 | Mod 2 | Mod 3 | Mod 4 | Mod 5 | Mod 6 | Avg |
| ATL | GB | ATL by 3 | 0.47 | 0.78 | 0.75 | 0.67 | 0.53 | 0.20 | 0.57 |
Prediction: ATL 33, GB 30
Steelers @ Patriots (-6)
Three of our models have the Patriots winning, three have the Steelers winning, and our spread model has the Patriots winning by 7. The average probability comes to 0.45, giving PIT a 55% chance of winning (for what it’s worth, Models 2 and 3 have been the most accurate this year, and they both have NE winning). New England’s largest advantages over the Steelers are a more efficient passing offense, more points scored, a better 3rd down defense, and more rushing attempts per game. The point spread model (which likes NE) is largely driven by points scored and passing offensive efficiency – two areas where the Patriots have outperformed the Steelers. The models like Pittsburgh’s defensive rushing efficiency, yards per play, and the fact that their offense rarely goes three and out. In addition, this game has a high total in Vegas (51 points), which may not appear to give one team an advantage, but the numbers show that this actually favors Pittsburgh. Home teams have won 56% of games the last two years – however, that number dips to 51% when the total is at least 48 points and rises to 59% when the total is 42 or lower (i.e., a higher total is good for the road team). The models are also penalizing NE for their weak schedule, their home loss to Seattle, and their low sack and interception totals.
As we highlighted last week, the Patriots are 20-1 (15-5-1 ATS) as home favorites outside of their division since 2013, while the Steelers are 12-10 (12-9-1 ATS) in road games outside of their division over the same span. Digging a bit deeper, the Patriots have played in 10 AFC Championship games under Belichick, going 5-5 overall, but 4-1 when they are at home (4-6 ATS overall, 2-3 ATS at home). Of those 5 home contests, only 1 went over the Vegas total.
| Home | Away | Our Spread | Mod 1 | Mod 2 | Mod 3 | Mod 4 | Mod 5 | Mod 6 | Avg |
| NE | PIT | NE by 7 | 0.13 | 0.61 | 0.59 | 0.39 | 0.55 | 0.41 | 0.45 |
Prediction: PIT 24, NE 23
Model Performance in Rounds 1-2
| MODEL | RD 1 | RD 2 | RD 3 | RD 4 | TOT |
|---|---|---|---|---|---|
| Model 1 | 3-1 | 3-1 | – | – | 6-2 |
| Model 2 | 4-0 | 3-1 | – | – | 7-1 |
| Model 3 | 4-0 | 3-1 | – | – | 7-1 |
| Model 4 | 4-0 | 2-2 | – | – | 6-2 |
| Model 5 | 4-0 | 2-2 | – | – | 6-2 |
| Model 6 | 3-1 | 2-2 | – | – | 5-3 |
| Avg. | 4-0 | 3-1 | – | – | 7-1 |
| Spread (vs. Vegas) | 1-3 | 2-2 | – | – | 3-5 |
| Spread (Winner Correct) | 3-1 | 2-2 | – | – | 5-3 |
