Corey Coleman and Leonte Carroo top NFL success scores

Kevin Cole applies his wide receiver prospect model to the 2016 draft class to see who is most likely to have early NFL success.

| 1 year ago
(AP Photo/Charlie Riedel)

(AP Photo/Charlie Riedel)

Corey Coleman and Leonte Carroo top NFL success scores

My last post detailed the construction of a pre-NFL combine logistic regression model for wide receiver prospects. That post also went through the historical results used to train and test the model.

In this post we truly get into the prediction business, applying the model to the 2016 wide receiver draft class. Before we begin, let’s review what our model found to be the most significant variables for predicting early-NFL success (measured as at least one top-24 PPR year in a receiver’s first three seasons).

Dozens of collegiate stats were narrowed down to four that maximized predictive power:

1. Draft age

2. Career market share of receiving yards (i.e. a prospect’s total receiving yards for all years as a fraction of total team receiving yards)

3. Receptions per game (final season)

4. Receiving touchdowns per game (final season)

All four variables were found to be significant in the model, with younger draft age, higher career market share, and more receiving touchdowns all being indicators of future success. Interestingly, fewer receptions per game is a positive in the model, likely because the model prefers receivers who accumulated yards and touchdowns through efficiency, not high volume.

Let’s now apply our model to the 2016 wide receiver draft class. Here are the top-15 scores by our model.

Name College Year Age Career MS Rec/Gm Rec TD/Gm Predict
Leonte Carroo Rutgers 2016 22.4 0.36 4.9 1.3 0.71
Corey Coleman Baylor 2016 22.0 0.26 6.2 1.7 0.62
William Fuller Notre Dame 2016 22.2 0.29 4.8 1.1 0.49
Hunter Sharp Utah State 2016 22.2 0.37 6.5 0.8 0.45
Tyler Boyd Pittsburgh 2016 21.7 0.43 7.6 0.5 0.43
Rashard Higgins Colorado State 2016 21.7 0.35 6.2 0.7 0.41
Roger Lewis Bowling Green 2016 22.6 0.30 6.1 1.1 0.40
Pharoh Cooper South Carolina 2016 21.3 0.26 5.5 0.7 0.28
Laquon Treadwell Ole Miss 2016 21.0 0.23 6.3 0.9 0.27
Thomas Duarte UCLA 2016 21.3 0.19 4.1 0.8 0.26
Derunnya Wilson Mississippi State 2016 21.8 0.21 4.5 0.8 0.23
Josh Doctson TCU 2016 23.6 0.28 7.8 1.4 0.23
Devin Lucien Arizona State 2016 23.0 0.30 5.5 0.7 0.20
Kenny Lawler California 2016 22.0 0.14 4.3 1.1 0.20

The “Predict” column gives the model score for each prospect (between 1 and 0). The higher the score, the more likely a prospect will have a top-24 PPR season in his first three years. You can find more details on the model here.

Leonte Carroo tops the 2016 class with a robust model prediction of 0.71. For context, this score would place Carroo in the top-10 of the entire 2000-2013 data set used to train and test the model, right between first-round picks Dez Bryant and Lee Evans. Carroo has a dominant career market share, in the top-5 of the 2016 class. In addition, the model likes his high touchdown totals, efficiently produced off relatively few receptions.

Potential draft positions are very fluid at this point, but Carroo is currently the 10th wide receiver off the board at NFL draft prediction market Play the Draft, going in the late third or early fourth round. Our model score indicates that, based on production only, Carroo would be a steal in that range, and could be a great value pick in upcoming rookie dynasty drafts.

Corey Coleman currently has buzz as a potential first-round pick in the NFL draft, but isn’t seen as being an elite prospect. Our model loves Coleman, especially his 1.7 receiving touchdowns per game (leads all 2016 prospects). Touchdowns are one of the more random stats in a smaller sample, so whether the 5-foot-11-inch, 185-pound prospect can continue to the find the end zone regularly in the NFL is still yet to be seen.

William Fuller is a top prospect and has been compared to recent Hall of Fame inductee Marvin Harrison. Fuller doesn’t have a particularly strong career market share, but the model likes him touchdown production and efficiency (on fewer receptions).

Hunter Sharp is an afterthought at this point in the draft cycle, listed as a likely UDFA and the 42nd wide receiver in in NFL Draft Scout rankings. But, Sharp did create some buzz at practices leading up to the East-West Shrine game. It’s possible that Sharp moves up during the pre-draft process, especially if his NFL combine results reflect his outstanding 37 percent career market share.

Tyler Boyd is currently seen as a top-10 wide receiver prospect, and the model backs that up. His prediction score of 0.43 isn’t outstanding, but definitely solid. Boyd has the highest career market share of the 2016 class, but is hurt by his lack of touchdowns and that he functioned as more of a possession receiver in his final college season.

Moving down the board to some of current favorites of draft watchers, Laquon Treadwell and Josh Doctson end up with prediction scores in the 0.20-0.30 range, which is respectable, but nothing to get excited about. The model likes Treadwell’s age, but his career market share is sorely lacking. Doctson has shown the ability to score touchdowns, but he’s a relatively old prospect and his lower market share pushes down his score.

Name College Year Age Career MS Rec/Gm Rec TD/Gm Predict
Michael Thomas Ohio State 2016 23.3 0.20 4.3 0.7 0.10

Michael Thomas is currently view by some as a potential first-round pick (though he didn’t go in the first round of our most recent mock draft), but he’s far down the prediction rankings, with a score of only 0.10. Thomas’ score is hurt badly by his older age and low career market share. Thomas’ career market share is probably understating his ability, driven down by the fact that he only had 22 yards in 12 games as a freshman, and competed for yards with early second-round pick Devin Smith last year.

Below is a sortable table of the 2016 wide receiver class with the relevant variables and prediction scores.

Kevin Cole is a Lead Writer for PFF Fantasy. You can follow him on Twitter at @Cole_Kev

NameCollegeYearAgeCareer MSRec/GmRec TD/GmPredict
Leonte CarrooRutgers201622.40.364.881.250.71
Corey ColemanBaylor2016220.266.171.670.62
William FullerNotre Dame201622.20.294.771.080.49
Hunter SharpUtah State201622.20.376.450.820.45
Tyler BoydPittsburgh201621.70.437.580.50.43
Rashard HigginsColorado State201621.70.356.170.670.41
Roger LewisBowling Green201622.
Pharoh CooperSouth Carolina201621.
Laquon TreadwellOle Miss2016210.236.310.850.27
Thomas DuarteUCLA201621.
Derunnya WilsonMississippi State201621.80.214.540.770.23
Josh DoctsonTCU201623.
Devin LucienArizona State2016230.35.50.670.2
Kenny LawlerCalifornia2016220.144.331.080.2
Michael ThomasSouthern Mississippi201621.90.24.310.690.19
Jalin MarshallOhio State201620.90.1630.420.18
Robby AndersonTemple201623.10.3150.50.18
Demarcus RobinsonFlorida201621.80.273.920.170.17
Jordan WilliamsBall State201622.10.2560.670.17
Marquez NorthTennessee201621.
Carlos HarrisNorth Texas201622.
Donovan HardenGeorgia State201622.
Malachi JonesAppalachian State201622.
Amara DarbohMichigan201622.40.254.460.380.14
Alonzo RussellToledo201623.80.2630.420.13
Bralon AddisonOregon201622.70.194.850.770.12
Bryce TreggsCalifornia201622.20.173.460.540.12
Dhaquille WilliamsAuburn201624.
Teddy RubenTroy2016230.225.250.750.12
Jehu ChessonMichigan201622.50.163.850.690.11
Joe HansleyColorado State201622.
Tajae SharpeMassachusetts201621.50.329.250.420.11
Cayleb JonesArizona201623.30.264.310.380.1
Daniel BravermanWestern Michigan201622.80.248.3110.1
Jay LeeBaylor2016230.162.920.620.1
Kolby ListenbeeTCU201622.40.1530.50.1
Michael ThomasOhio State201623.30.24.310.690.1
Sterling ShepardOklahoma201623.40.246.620.850.1
Chris MooreCincinnati2016230.163.640.640.09
Gabe MarksWashington State201622.90.1981.150.09
Geronimo AllisonIllinois201622.50.265.420.250.09
Marcus JohnsonTexas201621.90.1520.170.09
Pig HowardTennessee201623.30.20.500.09
Quinshad DavisNorth Carolina201622.10.193.930.290.09
Simms McelfreshAppalachian State201623.40.23.450.550.09
Byron MarshallOregon201622.
Dennis ParksRice201621.90.163.670.330.08
Mitch MathewsBrigham Young201624.
Devon CajusteStanford201623.
Edmarques BattiesMTSU201624.60.236.3110.07
Aaron BurbridgeMichigan State201622.
Alex EricksonWisconsin201623.70.35.920.230.06
Darius PoweCalifornia201622.30.093.620.620.06
Jenson StoshakFlorida Atlantic201622.30.214.670.170.06
Kj BrentWake Forest201622.
Melvin RayAuburn201623.20.1620.20.06
Antwane GrantWestern Kentucky201623.20.143.930.50.05
Cody CoreOle Miss201622.
Devin FullerUCLA201622.
Dom WilliamsWashington State201623.60.155.770.850.05
Jaydon MickensWashington201622.20.184.460.150.05
Macgarrett Kings JrMichigan State201622.
Malcolm MitchellGeorgia201623.90.234.460.380.05
Mekale MckayCincinnati201622.
Mose FrazierMemphis201622.80.25.380.310.05
Nelson SpruceColorado201623.60.286.850.310.05
Rashon CeaserLouisiana-Monroe2016230.
Ricardo LouisAuburn201622.30.153.540.230.05
Tevaun SmithIowa201623.40.172.910.270.05
Tres HoustonArkansas State201625.80.1930.830.05
Braxton MillerOhio State201623.60.141.920.230.04
Davonte AllenMarshall201623.30.174.830.420.04
Demarcus AyersHouston2016220.176.930.430.04
Dezmon EppsIdaho201623.90.3810.170.330.04
Dj FosterArizona State201622.60.174.540.230.04
Jakeem GrantTexas Tech201623.70.186.920.770.04
Jarvis BentleyTroy201623.20.1130.430.04
Jordan PaytonUCLA201622.80.1960.380.04
Kenneth ScottUtah201623.80.1730.310.04
Kj MayeMinnesota201622.40.175.620.380.04
Max MccaffreyDuke201622.10.1140.380.04
Paul TurnerLouisiana Tech201623.
Ryan LongoriaGeorgia Southern201623.40.171.500.04
Tevin JonesMemphis201623.50.1220.360.04
Tj ThorpeVirginia201623.10.152.560.110.04
Bobo BeathardAppalachian State201623.
Brandon SheperdOklahoma State201623.
Brandon SwindallUtah State201623.80.081.750.380.03
Cameron DickersonNorthwestern201622.70.081.2500.03
Casey MartinSouthern Mississippi201623.80.195.710.50.03
Daje JohnsonTexas201622.20.113.360.090.03
Durron NealOklahoma201622.70.113.380.230.03
Gary ChambersArizona State201623.30.1220.150.03
Imani DavisAkron2016220.
Joe MorrowMississippi State201623.
Marvin ShinnSouth Alabama201624.
Maurice HarrisCalifornia201623.
Miles ShulerNorthwestern201622.80.11.2500.03
Quenton BundrageIowa State201624.80.183.420.330.03
Richard MullaneyAlabama201623.40.112.530.330.03
Ryan BurbrinkBowling Green201623.40.154.310.310.03
Trevor DavisCalifornia2016230.
Autrey GoldenTexas-El Paso2016230.061.580.170.02
Danny AnthropPurdue201623.10.154.750.170.02
Jemond HazelySDSU201622.90.112.3300.02
Jordan FredrickWisconsin201623.80.081.1700.02
Jordan ThompsonWest Virginia2016230.12.9100.02
Kyle KleinKansas State201624.70.162.380.120.02
Stephen AndersonCalifornia201623.40.123.420.170.02
Von PearsonTennessee2016250.
Amir CarlisleNotre Dame201623.80.072.580.080.01
Bubba PooleUtah201624.20.051.8500.01
Chris ShillingsBall State201623.40.072.890.110.01
Deandre ReavesMarshall201624.20.064.310.310.01
Kj MyersWest Virginia201623.60.02100.01
Max MorrisonCincinnati2016240.1240.230.01
Shamier JefferySouth Carolina201624.80.051.7500.01
Tevin JonesMemphis201623.500.0900.01
Deante GrayTCU201622.30.08NaNNaNNaN
Michaelee HarrisAkron201623.9NaNNaNNaNNaN
Tim PatrickUtah201622.60.12NaNNaNNaN
  • Justin Wright

    Is there anyway to add the “predict” score for other WR’s on these teams to the model to get more accurate significance from the “market share” variable? Its seems that MS is the one stat here that is highly dependent on the surrounding talent. The reason I ask is that Coleman had a higher caliber second and third WRs this year (Cannon and Lee) while Carroo had no competition.

    • Kevin Cole

      I’ve thought about something similar: making an adjustment if there are more than one quality draft prospect on the same team. I started thinking about it after OBJ and Landry both were super successful in the pros, but they kept each others MS down in college.

      Looking at the rest of the team’s predict scores is a good idea. I’ll take a look.

      I think MS is dependent on surrounding talent, but there seems to be a cap around 30% unless you are an elite talent, no matter how poor your fellow WRs.

  • Doug

    Are these ages supposed to be in years or some type of age score? I have Boyd as 23 (1/15/1993), Coleman as 21 (7/6/1994), and Fuller as 21 (4/16/21994) to just name a few.

    • Kevin Cole

      I probably should have been a little more exactly when the age is calculated as off. I’m using mid-year 2016. The reason is that all the ages in the historical database are just draft year – year of birth. So, on average they should be as of midway through the draft year.

      Coleman and Fuller’s ages as of 7/1/2016 look correct by your dates but Boyd’s is a year off. I got all the 2016 ages from a database someone else complied. That person had Boyd’s DOB as 11/5/94, so that’s what I’m using.

    • Kevin Cole

      FYI, this article says Boyd was 20 when he got a DUI in June 2015. So, he couldn’t have been born in 93. I think my DOB of 11/5/94 is probably right.

  • YayAnalysis!

    Any chance we could see this applied to the 2014 and 2015 draft classes?