Tuesday, October 4, 2011

Week 4 Recap

I’m a fucking swaddling master. All the nurses compliment my swaddles. Get your swaddle game up you dads out there.

-Channing Crowder

Week 4 came and went and along with it, our second week of attempting to predict who will cover versus the NFL lines.  This week we introduced two models to accomplish this goal.  The first model was the same one used as last week.  As I’m sure everyone recalls, last week this model was correct on 62.5% of the games and would have ended up winning $370 (if you had bet $100 on every game).  Introduced this week, was a 2nd model that mined the data to produce the highest adjusted R square.  The two models ended up disagreeing on three games: Browns-Titans, Dolphins-Chargers, and Jets-Ravens.  Below is how the two models did:







GameSpreadM1 PredictionM2 PredictionMike_ZWojayKyle KellyActual Result
Panthers AT BearsBears by 6PanthersPanthersPanthersBearsBearsPanthers
Bills AT BengalsBills by 3.5BillsBillsBillsBillsBillsBengals
Titans AT BrownsPick 'emBrownsTitansBrownsBrownsTitansTitans
Lions AT CowboysCowboys by 1.5LionsLionsLionsLionsLionsLions
Steelers AT TexansTexans by 4TexansTexansSteelersTexansSteelersTexans
Saints AT JaguarsSaints by 7SaintsSaintsSaintsSaintsSaintsSaints
Vikings AT ChiefsVikings by 2VikingsVikingsVikingsVikingsChiefsChiefs
49ers AT EaglesEagles by 949ers49ers49ersEaglesEagles49ers
Redskins AT RamsRedskins by 2RedskinsRedskinsRamsRamsRedskinsRedskins
Giants AT CardinalsGiants by 1GiantsGiantsGiantsGiantsGiantsGiants
Falcons AT SeahawksFalcons by 4.5FalconsFalconsFalconsFalconsFalconsSeahwaks
Broncos AT PackersPackers by 12.5PackersPackersBroncosPackersPackersPackers
Patriots AT RaidersPatriots by 4PatriotsPatriotsPatriotsPatriotsRaidersPatriots
Dolphins AT ChargersChargers by 7DolphinsChargersDolphinsDolphinsChargersChargers
Jets AT RavensRavens by 3.5JetsRavensJetsRavensJetsRavens
Colts AT BuccaneersBuccaneers by 10ColtsColtsColtsColtsBuccaneersColts


As you can see both models did very well.  The first model ended up correctly predicting 10 out of 16 games to do exactly as well as last week (62.5%) and winning back another $370 dollars.  Through two weeks this model has ended up making over $700 dollars.  However, the 2nd model exceeded all expectations by correctly predicting 13 out of 16 games (81.3%) and winning $881.83.  That is a phenomenal return.  However, we can’t assume that this model will continue being this accurate.  What may be even more encouraging is that the 1st model maintained it’s accuracy through two weeks.

Percent Right Season Results
Mike Z  0.4375 0.4375
Wojay  0.5 0.5
Kyle Kelly 0.5 0.5
Model 1 0.625 0.625
Model 2 0.8125 0.8125


As you can see in the season results, everyone basically repeated their performance.  Kyle Kelly and Model 2 were added this week.  Meanwhile, me, Wojay and the 1st model did exactly the same.  Through two weeks, I’ve correctly predicted 14 out of 32 games, Wojay 16, and Model 1 has predicted 20. 

Model 1 Model 2
Starting 1600 1600
Week 3 1973.33              .
Week 4 2346.66   2481.83


As you can see, both models have made money through 2 weeks.  Let’s hope this pattern continues!

No comments:

Post a Comment