## Game Probabilities Week 5

Game probabilities for week 5 are up at the New York Times. This week I discuss the preeminence of offense over defense.

### 12 Responses to “Game Probabilities Week 5”

1. bytebodger says:

What in the heck is going on with them sticking the probabilities in that tiny, almost unreadable image???

2. ASG says:

Am I reading this correctly that Houston and Detroit are over 50% to win despite being around 2.5-1+ underdogs in Vegas?

3. Anonymous says:

These predictions are based on Brian's model. If you compare the probabilities he generates with the probabilities that Vegas comes up with, you'll see disparities every week. I don't track them, but in the past the Vegas spread has been very good done better than numerical models. (I don't know about Brian's in particular.)

4. Anonymous says:

Brian,
Wanted to know what your thoughts are on the Prediction Machine.(Predictionmachine.com.)Their model plays every game 50,000 times.They claim that their system is the most advanced sports system.Their system takes into account healthy,hurt players.Coaches,officials,home field advantage,and weather.They are 44-19 on the year picking games straight up.Cbs sports.com lists their weekly picks.Your system and their system are in disagreement with 5 of the following games. For the record;PM-Sea,You-Indy,PM-Gb,You-Det,PM-New Eng,You-Cincy,PM-Car,You-Ari,PM-Sf,You-Hou.

5. Anonymous says:

Extending ASG's comment about Houston and Detroit, at least Detroit being favored coincides with the offensive and defensive EPA numbers for the Lions and Packers (they seem close from the Team Stat Visualizations chart, and Detroit is at home). Meanwhile, San Francisco's offensive and defensive EPA numbers are both better than Houston's and San Francisco is at home. What other potential major input am I missing? Special teams? ST can't be that impactful. Injuries? Willis and Anthony Davis might not play, but ditto for Cushing. Just curious what else I should be considering.

6. j says:

JAC .43? seems way too high for me, more like 25% at best

BAL .33? MIA has TO issues, I'd say the game is more 50-50

ARI .53? CAR is vastly underestimated

7. Michael Beuoy says:

Anonymous: If you picked the betting favorite each game, you would have gone 43-20. So, the prediction machine is one game better than the market, which is actually good for a model, but it's only four weeks in.

I'm not convinced a simulation model is any better than a more straightforward regression. Your simulation is only as good as your ability to determine true talent level, and there are just fundamental sample size issues that prevent us from knowing that much.

8. James says:

EPA numbers represent what *has* happened, while the game predictions are based on what *will* happen. Many things that have happened are not likely to continue happening (such as an interception returned for a TD), and so the game predictions factor those out.

9. Anonymous says:

They need to make the probabilities table larger or make a clickable link to it like there was last week.

10. mitch says:

I agree offense beats defense in the regular season, but not so sure about teams winning the Super Bowl.
If your team goal is to have a great regular season, a la the 2007, 2011 or 2012 Pats, the 2cd, 4th and 7th highest scoring teams in history then offense it is. But if your goal is winning SB's then looking back there's more defensive oriented teams with ave to below ave defenses that win the the other way around.

The 90 Gaints, 2001 Pats, 2003 Pats, 2002 Bucs, come to mind while the only offensive team with a below ave defense is 2009 Saints but even then they were rank 3rd in defensive passer rating.

The 2006 Colts were weak defensively in the regular season but were way better defensively then offensively in the playoffs.

While some are busy looking at bell curves I look at the actual results of games --- the 2007 Giants scored the fewest points of any SB winner in the 16 game schedule era while beating the 2cd highest scoring team in NFL history.

The 2001 Pats scored the fewest points of any SB winner at the time while beating the greatest show on earth Rams, one of the highest scoring teams in history.

The 90 Gaints beat the high flying Bills scoring the fewest points in SB history in the 16 game schedule at the time and still ranks the 2cd fewest ever, and the Bills were one of the greatest teams in history scoring in points scored per plays run.

There's 3 defensive team that needed the fewest points scored in history of the 16 game schedule to beat the greatest offensive teams of all-time.

11. mitch says:

I could add to that, the 2001 Pats offense scored what, 3 TD's in the entire postseason. But they had an incredible stat line of having one of the worst defensive ave per plays while in their opp territory but once the opp cross-mid-field they were one of the best defenses in the league.

The 2003 Pats defense had an incredible stat line of giving up a very high ratio of fields goals attempted to TD's, one of the best in history.

I forgot the 2000 Ravens, a monster defensive team.

When we look at monster offensive teams that won SB's like the 1991 Skins, 89 and 94 49ers they had far better than ave defenses, they were all very strong defensive teams.

12. Jero GB says:

Correct one of the Anons above, DET is @ Lambeau, not at home. Haven't won in WI since 1991. That is 21 years of losing in WI, they broke their 0-21 hex in D.C. two weeks ago, 21 might be an interesting number again this week... I fear.

I am also wondering why the tiny image of probabilities on the NYT blog?