Updated Daily Fantasy Football Projections: NFL Wild Card Round


Dr. J.J. McKinley holds a Ph.D. in mathematics and tackles sports data to give you a competitive edge in your fantasy games.

On Wednesday I introduced my daily fantasy football projections that uses a machine learning algorithm to predict individual player performance. The model by and large does a good job at creating reasonable projections, but does have a few shortcomings. I’ve addressed one shortcoming to create more accurate projections for this weekend’s games, and added an additional feature to look at the variation of these projections. This way, we can attempt to use the model to differentiate cash-game plays from GPP-only plays.

Prediction Adjustments

One of the limitations of the model that I mentioned is an inability to properly adjust for injuries. The model projects Le’Veon Bell for 23.7 points (using DraftKings scoring), unaware that he suffered a knee injury this past weekend against Cincinnati. Similarly, Josh Harris and Dri Archer’s projections are probably too low for their potential usage. A few other projections also needed updating, most notably the Colts tight ends (with Dwayne Allen set to play) and DeAngelo Williams expected to return in a complementary role to Jonathan Stewart. Through the remainder of the playoffs, I will have to manually adjust some of the inputs to account for this limitation, but expect to have an algorithmic-based version for next season. At the time of publication, Le’Veon’s status isn’t known, so the table immediately below summarizes the Pittsburgh running back projections based on Bell’s status.

Le'Veon Bell23.67513.4020.00
Dri Archer0.9404.6728.07
Josh Harris0.7180.8811.26

Another minor update – I mistakenly left off Dan Herron and Steve Smith from the projections on Wednesday. This was a simple data merging error and their projections are correctly listed below.

Projecting Variation

While creating projections is a useful exercise, they can only tell a portion of the story. It is also important to determine what kind of variation a player will have around his projection. For cash-games we’d like players with high floors (relative to their price), and for GPPs we are looking for upside. Thus, players projected for a high points per thousand dollars (P/$) with low variation will be ideal cash game plays, and players with the potential for big games relative to their value make great GPP plays. To determine variation I used the projected points from each individual decision tree that made up the random forest. I then took the standard deviation of the 2000 individual decision tree projections and normalized it to the mean to get the coefficient of variation (CV). While this doesn’t make a difference in absolute points (a standard deviation of three points is three points regardless of player) it does make a difference in value (three points on a $3000 player is much more variation in P/$ than three points on a $9000 player).

Projection Picks

Based on the variation results, I still like Antonio Brown and Greg Olsen as a must-play in all formats, but the model says we can add Cole Beasley to that list (I previously had him as a strong GPP play). Two GPP plays I really like are Andy Dalton and Giovani Bernard. Not only do I expect fairly low ownership of them, but they both have significant upside. Dalton is projected by this model to have the highest P/$ but the 2nd highest variation among starting QBs. This upside makes sense since the Bengals are likely to be playing from behind on the road. Enter Bernard. Although the model projects him 6th among running backs in P/$, he sits 9th in variation out of the top 12 projected RBs. If the Bengals do indeed play from behind, look for Dalton to check down to Bernard several times. Considering Indianapolis is 31st in DVOA against RBs in the passing game, this makes the Bengal duo a high-risk, high-reward stack in GPPs.

PlayerPosDK SalaryProj PtsP/$CVCV RankPos CV Rank
Antonio BrownWR890027.443.08414.1%42
Andrew LuckQB880021.842.48214.5%73
Dez BryantWR850020.772.44412.8%21
DeMarco MurrayRB880020.692.35112.4%11
Calvin JohnsonWR880020.192.29414.4%63
Matthew StaffordQB720019.842.75614.0%31
Tony RomoQB750018.702.49315.6%84
Ben RoethlisbergerQB730018.232.49715.8%95
T.Y. HiltonWR780018.022.31018.2%146
A.J. GreenWR800018.012.25117.0%114
Andy DaltonQB610017.942.94118.0%137
Joe FlaccoQB670017.852.66414.2%52
Cam NewtonQB760016.212.13316.1%106
Jeremy HillRB620016.062.59018.7%162
Greg OlsenTE550013.272.41418.4%151
Cole BeasleyWR370013.103.53917.6%125
Torrey SmithWR540012.902.38923.3%2610
Michael FloydWR520012.882.47620.8%198
Joique BellRB550012.492.27122.0%245
Kelvin BenjaminWR590012.322.08821.2%209
Jason WittenTE450012.302.73321.4%223
Jonathan StewartRB530012.222.30521.9%234
Giovani BernardRB560012.072.15522.5%256
Dan HerronRB470011.872.52518.8%173
Steve SmithWR500011.632.32719.7%187
Justin ForsettRB610011.151.82824.0%287
Heath MillerTE390010.592.71621.3%212
Jermaine GreshamTE37009.822.65526.5%324
Golden TateWR55009.601.74623.9%2711
Martavis BryantWR41009.292.26625.7%3013
Dwayne AllenTE38009.162.41027.9%355
Ryan LindleyQB52008.931.71727.8%348
Kamar AikenWR30008.402.80125.7%2912
Donte MoncriefWR36008.132.25828.1%3614
Dri ArcherRB45008.071.79326.2%318
Reggie BushRB48007.171.49427.5%339
Markus WheatonWR31007.122.29528.1%3715
Philly BrownWR30006.962.32029.0%3816
Theo RiddickRB32006.922.16434.2%5011
Terrance WilliamsWR34006.771.99231.1%4521
John BrownWR41006.741.64530.0%4017
Jerricho CotcheryWR34006.711.97231.0%4319
Coby FleenerTE50006.481.29729.4%396
Owen DanielsTE33006.441.95132.2%487
Marlon BrownWR30006.422.14130.8%4218
Jeremy RossWR30006.322.10634.1%4924
Zurlon TiptonRB31006.222.00844.0%7416
Hakeem NicksWR33006.001.82031.8%4723
Kerwynn WilliamsRB37005.831.57630.3%4110
Mohamed SanuWR42005.581.32931.6%4622
Larry FitzgeraldWR47005.331.13431.0%4420
Ryan HewittTE30005.251.75234.6%528
Ed DicksonTE30005.191.73035.2%539
Joseph RandleRB30005.101.70136.7%5512
Stepfan TaylorRB30004.941.64639.0%5813
Reggie WayneWR43004.811.11934.5%5125
Jaron BrownWR31004.781.54340.8%6328
Eric EbronTE30004.721.57335.9%5410
Trent RichardsonRB34004.481.31842.9%7214
Darren FellsTE30004.471.49239.5%6013
Crockett GillmoreTE30004.181.39439.3%5912
Derek AndersonQB50004.130.82554.4%8710
Justin BrownWR30004.011.33641.2%6731
Darrius Heyward-BeyWR30004.011.33641.2%6832
Michael PalmerTE30003.981.32840.6%6114
John CarlsonTE30003.981.32841.2%6615
Brandon TateWR30003.881.29340.9%6429
Fozzy WhittakerRB30003.731.24448.0%8117
Jacoby JonesWR30003.631.20942.8%7134
Kevin BrockTE30003.571.19137.8%5611
DeAngelo WilliamsRB40003.540.88643.2%7315
James WrightWR30003.431.14441.0%6530
Logan ThomasQB50003.420.68350.0%859
Lance MooreWR30003.341.11340.8%6227
Brandon WeedenQB50003.190.63762.8%9512
Phillip SupernawTE30003.091.02942.6%7016
Rob HouslerTE30003.031.01245.1%7719
Greg LittleWR30003.021.00546.5%7835
Michael CampanaroWR30002.940.97942.3%6933
Joe WebbQB50002.890.57856.8%9011
Matt SpaethTE30002.890.96344.6%7517
Ryan BroylesWR30002.680.89347.6%8037
Jed CollinsRB30002.630.87655.7%8918
Brandon WilliamsTE30002.600.86549.7%8421
Brandon PettigrewTE30002.590.86344.7%7618
Jack DoyleTE30002.550.85155.5%8823
Corey FullerWR30002.520.84046.9%7936
Marion GriceRB30002.480.82664.2%9623
Brenton BersinWR30002.370.79138.2%5726
Fitzgerald ToussaintRB30002.340.77960.6%9321
Bernard PierceRB30002.330.77657.0%9119
Josh CribbsWR30002.260.75248.6%8238
Rex BurkheadRB30002.160.72259.8%9220
Jason CampbellQB50002.120.42370.1%10213
Devin StreetWR30002.040.68164.4%9739
James HannaTE30002.030.67848.9%8320
Gavin EscobarTE30001.940.64753.8%8622
George WinnRB30001.890.63169.7%10126
Lance DunbarRB30001.810.60379.6%10428
Dwayne HarrisWR30001.730.57768.2%10040
Robert HughesRB30001.710.57165.3%9824
Cedric PeermanRB30001.650.55162.4%9422
Kyle JuszczykRB30001.520.50767.6%9925
Josh HarrisRB35001.260.35970.9%10327
Mike TolbertRB30001.250.41684.2%10630
Will JohnsonRB30001.150.38381.6%10529
Tyler CluttsRB30001.040.34886.0%10731
Le'Veon BellRB90000.000.000
Dane SanzenbacherWR30000.000.000