Big Plays (17-23-2 YTD, 132.5-114.5-8 lifetime)
- Green Bay +9 vs. Minnesota [MP Line: GB +5.9]
- Cincinnati +3.5/+4 vs. Detroit [MP Line: CIN -2.1]
- NY Jets +7 vs. LA Chargers [MP Line: NYJ +4.5]
- Tennessee +6.5/+7 vs. LA Rams [MP Line: TEN +3.5]
- New England -11 vs. Buffalo [MP Line: NE -15.5]
- NY Giants +4 at Arizona [MP Line: NYG +0.9]
Other Plays (21.5-6-0.5 YTD, 172.5-130.5-7 lifetime)
- Jacksonville -4 at San Francisco [MP Line: JAC -7.7]
- Seattle +4.5/+5 at Dallas [MP Line: SEA +2.6]
Break-Even or Better (unofficial leans) (13.5-8-1.5 YTD, 69.5-78.5-9 lifetime)
- Baltimore -13.5 vs. Indianapolis [MP Line: BAL -15.8]
- Tampa Bay +10 at Carolina [MP Line: TB +8.9]
- Miami +10.5 at Kansas City [MP Line: MIA +9.0]
- Houston +8.5/+9 vs. Pittsburgh [MP Line: HOU +7.3]
Our Lines on Remaining Games
- Chicago -7.5 vs. Cleveland
- New Orleans -4.3 vs. Atlanta
- Washington -4.5 vs. Denver
- Philadelphia -9.7 vs. Oakland
Some of you have asked about how I determine whether something is a big play, other play, or lean. Why, you might ask, is Houston +8.5 be a lean, when the MP line is Houston +7.3, and the ‘8’ is a dead number? The MP line is a mean, an average prediction; it’s linear and does not account for the distribution. So the difference between MP lines of -3.5 and -2.5 is no bigger than -2.5 and -1.5. Because the actual distribution of score differentials is not smooth — points are (usually) scored 3 or 7 points at a time — we need to translate the MP line into a distribution of possible outcomes, which I do using what’s called an ordered logistic regression. This gives me the probability of a team covering any spread I might see, based on the MP line. In order to reflect the true predictive power of my line relative to the market, I regress the probability of a team covering the market line (based on the MP line, using the ordered logit) toward the market’s probability of a team covering (50% assuming even juice). While I don’t know the true predictive power of the MP line relative to the midweek line, I can estimate the optimal regression toward the closing line historically, which comes out to about 60%, though that’s a little higher early in the season. Since I beat the closing line more often than not, I assume a little less regression, and at this point in the season am weighting the market 50% and MP 50%.
