2010 NFL Point Spread Picks Week 13
Pick 1: New England -3 
Pick 2: St. Louis -3
Pick 3: Atlanta -3
What the hell is going on? This weekend I spent hours trying to see if I was doing a mistake reading in the data, in the analysis, and basically in every line of code, nothing. The model simply has sucked this year. As you can tell, I'm very frustrated and disappointed, but I'm keeping my cool.
This week I like the experienced Patriots over the Jets, is it Monday yet? Atlanta and Roddy White scoring touchdowns on Tampa Bay's strong D. Finally, at the top of the ATS Standings, the Rams beating down on a smiling Derek Anderson.
The last few weeks (except for the last which we usually don't bet) and the playoffs are historically my best weeks. Assuming this, then my probability of winning is higher. Therefore, according to the formula, the weekly risk increases a bit more. Now taking it at 25% giving each game this week a wager of $120. Good luck, it wouldn't hurt.
Other notable picks include: Chicago, Oakland, and Houston.
How to read the table:
Pick 2: St. Louis -3
Pick 3: Atlanta -3
What the hell is going on? This weekend I spent hours trying to see if I was doing a mistake reading in the data, in the analysis, and basically in every line of code, nothing. The model simply has sucked this year. As you can tell, I'm very frustrated and disappointed, but I'm keeping my cool.
This week I like the experienced Patriots over the Jets, is it Monday yet? Atlanta and Roddy White scoring touchdowns on Tampa Bay's strong D. Finally, at the top of the ATS Standings, the Rams beating down on a smiling Derek Anderson.
The last few weeks (except for the last which we usually don't bet) and the playoffs are historically my best weeks. Assuming this, then my probability of winning is higher. Therefore, according to the formula, the weekly risk increases a bit more. Now taking it at 25% giving each game this week a wager of $120. Good luck, it wouldn't hurt.
Other notable picks include: Chicago, Oakland, and Houston.
| Game | Vegas Line | Estimate | Prediction-Vegas | Confidence |
|---|---|---|---|---|
| NY JETS @ NEW ENGLAND | -3 | -10.0 | -6.5 | 64.3% |
| ST LOUIS @ ARIZONA | 3 | 6.3 | 3.8 | 61.7% |
| ATLANTA @ TAMPA BAY | 3 | 12.1 | 9.1 | 61.5% |
| CHICAGO @ DETROIT | 3.5 | 13.4 | 9.9 | 58.5% |
| OAKLAND @ SAN DIEGO | -13 | -10 | 3 | 58.3% |
| HOUSTON @ PHILADEPHIA | -9 | -3 | 6 | 55.4% |
| WASHINGTON @ NY GIANTS | -7 | -2 | 5 | 54.8% |
| CLEVELAND @ MIAMI | -4.5 | -5.5 | 1.0 | 52.6% |
| JACKSONVILLE @ TENNESSEE | -3 | -6.1 | -3.1 | 52.1% |
| PITTSBURGH @ BALTIMORE | -3 | 1 | 4 | 50.7% |
| BUFFALO @ MINNESOTA | -6.5 | 1.1 | 7.6 | 50.0% |
| DALLAS @ INDIANAPOLIS | -5.5 | -3 | -2.5 | 50.0% |
| DENVER @ KANSAS CITY | -9 | -10 | -1 | 49.7% |
| SAN FRANCISCO @ GREEN BAY | -10 | -16 | -6 | 49.2% |
| NEW ORLEANS @ CINCINNATI | 7 | 5.4 | 1.6 | 48.0% |
| CAROLINA @ SEATTLE | -6 | -5 | 1 | 48.0% |
How to read the table:
- Vegas Line: A NEGATIVE number implies the spread favors the HOME team
- Estimate: NFL pickles spread estimate
- Pred-Vegas: Subtraction of the previous two. POSITIVE implies VISITING team will cover spread.
- Confidence: The probability that the spread pick is on the correct side.
Comments
2-2 last week. 15-9-1 (62.5%) on the season.
1. you don't like giving up too many points but your preference for small favorites exposes you to trap games like Philly -3 @ Chicago or Indy -1 vs San Diego.
2. You pick teams you think will lose by less than the spread. The spread affected several games in week 12, but was a non-factor in the first 11 weeks.
Also, can you clarify what you mean by #2?
If I could clarify it does seem you pick a lot of road teams, which is good since covers.com reports road teams cover 53% of the time.
I guess at the core I don't understand the confidence ranking, or why you feel better about Skins Giants than Saints Bengals.
Also, I realized that I made a mistake in running my code, and the Broncos picks was a mistake. I'm still counting it in my record, but the updated picks are here.
342 Bills/Vikings UNDER 44.5
301 Texans +9 [Thu]
341 Bills +5.5
I wanted to get on the Lions +4, but it's off the board for reasons which I'm sure would be much more clear if I had access to decent sports TV.
I have noticed that my model tends to prefer road teams. I'm not sure why. The home field advantage is not static, i.e. it is not 3 for every game. Nor is it specific for each team (too many parameters to estimate, implies less accuracy). Now you have me thinking, and I'll probably run some stats to see why road teams tend to have higher probability ratings.
I have a similar skew with my model toward road teams. In my backtesting, not only did my model prefer road teams, but the road team picks also did better.
Also, why do you have picks with confidence levels under 50%? Is this because you are adjusting for a known overconfidence in your model? Seems like if one side were under 50%, you would switch the pick (unless the probability of a push was large enough to make the other side even lower; this doesn't seem to be the case though).
Nick,
According to my data, there is not such a large road team advantage against the spread over the last decade (I see about 51%). Perhaps covers was referring to a shorter, recent period of time, because a 53% advantage for road teams over a long period would really surprise me.
Why do you keep changing the confidence threshold that you use to make the picks? This week it is back to 60 from your usual 57. While models obviously require adjustment, if you are 57 percent certain about a game (say Chicago or Oakland) then it IS a good bet, no?
In any event, my picks for this week came out to be:
Buffalo +5.5
Cleveland +5
Oakland +13
Giants Win
Saints Win
Chicago - Detriot over 44
Saints - Bengals over 45.5
Rams - Cardinals over 43
I'm not really a fan of many of those, but who knows - the season has certainly not gone as expected thus far.
Ben
Benjamin, I'm sticking to 60% until the end of the season or until I have time to dig deeper in analysis of this threshold. The problem is that if the threshold is set to high, only a few games show, and too low well the results would probably be no good. What is the optima threshold, I don't know, yet.
Congrats on the great week!
Ben