DraftKings Week 14 Recap

Week 14 was a week when the winning lineup benefited from player replaceability. Brett Hundley (A-A-RON out), Giovani Bernard (Joe Mixon out), Trey Burton (Zach Ertz Out). In Week 7, we saw this strategy back fire but it panned out well in Week 14.

<scroll down to the bottom for more thoughts on player replaceability>

Quick notes:

  • Winner 220.60 TwoGun used an optimizer plus player exposure strategy
  • 147.66 to cash
  • Best lineup possible 266.98
  • Poor week for us humans. There were at least 197,628 better lineups than the winner!

Let’s begin by looking at TwoGun’s lineups.

TwoGun’s Winning Lineups

TwoGun's Week 14 Winning Lineup 220.60

TwoGun entered 57 lineups and drew heavily on Giovani Bernard. Here’s his top picks.

Percentage Player
90 Giovani Bernard
57 DeAndre Hopkins
47 Todd Gurley II
27 Kareem Hunt
27 LeSean McCoy
27 Mike Evans
25 Travis Kelce
24 Will Fuller V
20 Demaryius Thomas
20 Keenan Allen
18 Trey Burton
18 Golden Tate
18 Packers
18 Jack Doyle
17 Adam Thielen

It looks like his Giovani Bernard near-lock dominated his lineups. His stacks are shown below

Percentage Player
52 DeAndre Hopkins, Giovani Bernard
40 Giovani Bernard, Todd Gurley II
27 DeAndre Hopkins, Todd Gurley II
27 Giovani Bernard, Mike Evans
25 Giovani Bernard, LeSean McCoy
25 Giovani Bernard, Travis Kelce
20 Giovani Bernard, Kareem Hunt
20 Giovani Bernard, Will Fuller V
18 DeAndre Hopkins, Kareem Hunt
18 Giovani Bernard, Packers
17 Demaryius Thomas, Giovani Bernard
17 Giovani Bernard, Golden Tate
17 Adam Thielen, Giovani Bernard
17 Giovani Bernard, Larry Fitzgerald
17 Giovani Bernard, Keenan Allen

I’d like to say that he deliberately set Giovani Bernard-Kareem Hunt at 20%. But looking at only one week is not enough! However, I’m confident he manually set Giovani Bernard player exposure.

Thankfully, we are now tracking QB stacks a bit more dilligently. Like last week’s winner, TwoGun forced his QB-WR’s to stack.

Percentage Player qb
6 Giovani Bernard Carson Wentz
6 Nelson Agholor Carson Wentz
6 Carson Wentz
11 Giovani Bernard Marcus Mariota
11 Marcus Mariota
10 Delanie Walker Marcus Mariota
13 Golden Tate Matthew Stafford
13 Matthew Stafford
11 Giovani Bernard Matthew Stafford
6 Giovani Bernard Nathan Peterman
6 Nathan Peterman
11 DeAndre Hopkins Tom Savage
11 Tom Savage
10 Giovani Bernard Tom Savage
10 Trevor Siemian
8 DeAndre Hopkins Trevor Siemian
8 Demaryius Thomas Trevor Siemian

How to read this: Blank player means that the percentage of QB usage across all lineups. When it’s the same as a pair that is a surefire forced stack! (In the future I’ll start to display this table a bit better)

I also want you to look at TwoGun’s point distribution.

TwoGun's Point Distribution

If you’re entering multiple entries into a big tourney, you want high variance. Look at how TwoGun had lineups below 100 and above 200! It won’t always work out that way but if you have a spread (top lineup – worst lineup of around 30 you aren’t taking nearly enough risk.)

What We Know About TwoGun

All in all we know

  • [very likely] Manually set player exposure
    • supporting evidence: 97 players in his player pool yet Bernard in 90%, round number
    • opposing evidence: his projections source could have favored Giovani Bernard just 90% of the time
  • [definitely] Forced his QB’s to stack
    • evidence: QB’s stacked
  • [likely] Optimizer plus player exposure class of DFS player.

We’ve seen a lot of these optimizer plus player exposure guys win but that doesn’t mean it’s the best strategy. It could be more indicative of one of the major trends in the DFS world.

Personal Development Category (English)468x60

Pro Player Leaderboard

CONDIA is back!

payout lineups_entered payout/num_entries
screen_name
twogun 1017115.0 57 17844.122807
thatstunna 6685.0 150 44.566667
mrnastytime91 3530.0 150 23.533333
iampgm 845.0 150 5.633333
condia 160.0 94 1.702128
papagates -2300.0 150 -15.333333
ehafner -2330.0 150 -15.533333
chipotleaddict -2365.0 150 -15.766667
fallfromgrace -2440.0 150 -16.266667
al_smizzle -2440.0 150 -16.266667

Well CONDIA only made $160 but still won something. Al_Smizzle had another bad week. And ChipotleAddict fluctuates a ton! Besides that tommyg1979 only entered 3 lineups.

CONDIA Strategy Checkin

CONDIA used Trevor Siemian and Demaryius Thomas as a surefire stack in 11% of his lineups, that means he always played Siemian and Thomas together.

He only played 94 lineups so now we get to see how his lineups change versus the number of lineups. Sure enough they did.

Normally CONDIA plays QB’s in ~8% of his lineups. A quick computation 150/94*(standard_percentage) shows that we’d expect him to play QB’s around ~12.7% if his risk function is linearly based on his lineups.

Here’s what CONDIA did with his QB’s, min significance 5%

Percentage qb
17 DeShone Kizer
16 Alex Smith
13 Blaine Gabbert
11 Trevor Siemian
8 Jameis Winston
7 Marcus Mariota
6 Andy Dalton

(12.7% is a a bit hairy but it roughly fits the data). More convincingly, he took more risk at QB this week. None of these QB’s are truly top tier. Smith stands out as the best while the others can be very hit or miss.

CONDIA’s single player utilizations were far riskier than normal too.

Percentage Player
82 Giovani Bernard
48 Larry Fitzgerald
44 Kareem Hunt
41 LeSean McCoy
39 DeAndre Hopkins
39 Lamar Miller
32 Demaryius Thomas
28 Todd Gurley II
25 Mike Evans
19 Golden Tate
19 Chargers
19 A.J. Green
18 Bills
17 DeShone Kizer
17 Stefon Diggs

In past weeks, we haven’t seen any of his players over 40% utilization. Maybe he’s taking some big risks before retiring this year? Hope he keeps playing

Al_Smizzle What You Doin?

WYD meme

WYD=what you doin

Al_Smizzle’s player pool is consistent around 43 players (this week it was 44). But he’s just not picking the right player right now.

Percentage Player
65 Giovani Bernard
54 Todd Gurley II
45 Sterling Shepard
44 DeSean Jackson
41 Dez Bryant
34 Larry Fitzgerald
34 Travis Kelce
28 Melvin Gordon
27 DeAndre Hopkins
26 Josh Gordon
26 Bengals
26 Lamar Miller
25 Dak Prescott
25 Bills
25 Golden Tate

Sterling Shepard ouch. Same goes for Bengals D. Al_Smizzle’s strategy requires a bit of luck and football savyness. Perhaps his winning observations haven’t hit in recent weeks. We’ll try and find out what his go to players are after the season. For a pro like him, I’d assume he sticks to a strategy rather than chase things.

Bales Enters 18 lineups again

This guy loves spending $360. 18 seems like a lucky number. Here’s his single player utilization

Percentage Player
49 Giovani Bernard
49 LeSean McCoy
38 Sterling Shepard
38 Melvin Gordon
32 Bills
32 Austin Seferian-Jenkins
27 Todd Gurley II
27 Matthew Stafford
27 Corey Coleman
21 DeAndre Hopkins
21 Alshon Jeffery
21 Josh Gordon
21 Travis Kelce
21 Theo Riddick
21 Golden Tate

His top lineup was 196.14 and worst was 96. So he got quite the spread from his 18 entries.

Pro and Herd Meta

The pro’s beat the herd by a solid 3 points on average.

index Points_herd Points_pro
count 172523.000000 3818.000000
mean 130.725991 134.328910
std 21.851224 21.510833
min 7.200000 63.200000
25% 115.560000 119.320000
50% 130.320000 133.200000
75% 145.420000 148.095000
max 218.840000 220.600000

The biggest difference between the herd and pro’s was Josh Gordon.

lineup_hash Percentage Pro Percentage Herd pro_favoritism
Josh Gordon 18 36 -18
Michael Crabtree 6 20 -14
Marquise Goodwin 8 21 -13
Travis Kelce 20 31 -11
Chargers 6 13 -7
Adam Thielen 9 15 -6
Bengals 5 10 -5
DeAndre Hopkins 21 26 -5
Alex Smith 9 14 -5
Melvin Gordon 13 18 -5
Tyreek Hill 13 17 -4
Jamaal Williams 6 10 -4
Larry Fitzgerald 25 28 -3
Mike Davis 6 8 -2
Jimmy Graham 5 7 -2
Marvin Jones Jr. 12 13 -1
Evan Engram 5 6 -1
Stephen Anderson 12 10 2
LeSean McCoy 30 28 2
Vikings 12 9 3
Bills 14 11 3
Jaguars 10 6 4
A.J. Green 22 18 4
Kareem Hunt 15 10 5
Sterling Shepard 13 7 6
Dez Bryant 15 9 6
Matthew Stafford 11 5 6
Keenan Allen 30 23 7
Golden Tate 17 9 8
Lamar Miller 32 24 8
Giovani Bernard 60 51 9
Mike Evans 19 7 12
Todd Gurley II 28 15 13
Trey Burton 13 <5%
DeSean Jackson 13 <5%
Demaryius Thomas 12 <5%
Alshon Jeffery 12 <5%
Carlos Hyde 12 <5%
Packers 11 <5%
Jameis Winston 11 <5%
Jamison Crowder 10 <5%
Browns 7 <5%
Leonard Fournette 6 <5%
Doug Baldwin 6 <5%
Carson Wentz 5 <5%
Theo Riddick 5 <5%
Titans 5 <5%
Frank Gore 5 <5%
Cowboys 5 <5%
Samaje Perine 5 <5%
Blaine Gabbert 5 <5%
Kirk Cousins 5 <5%
Alfred Morris <5% 13
Jimmy Garoppolo <5% 12
Philip Rivers <5% 11
Hunter Henry <5% 9
Cameron Brate <5% 8
Stefon Diggs <5% 6
DeShone Kizer <5% 5
Nelson Agholor <5% 5
Trent Taylor <5% 5
Austin Seferian-Jenkins <5% 5
Derek Carr <5% 5

While most will agree Josh Gordon had a good week, compared to his salary there were better choices. And Josh Gordon wasn’t in our better than the winning lineup player pool.

The pro’s played Giovani Bernard 9% more than the herd! But more importantly, their devotation towards TGII has been paying off.

The herd’s biggest stack was Alex Smith-Travis Kelce at 8% and they used 347 players across their 172,652 lineups.

The pro’s biggest stack was Jameis Winston-Mike Evans at 8% and they used just 188 players across their 3,818 lineups.

Interestingly, the pro’s stacked Blaine Gabbert-Larry Fitzgerald in 5% of their stacks and Blaine Gabbert was used in 5% of lineups making that a potential surefire stack. (potentially only because I round my percentages to the nearest number.)

Best Lineup Possible: 266.98

At least 197,628 better lineups than the winner’s 220.60!

The best possible lineup scored 266.98 while using a salary of just $43,500.

player_name position score salary team_name
Dak Prescott QB 29.58 5600.0 DAL
Rod Smith RB 36.00 3900.0 DAL
Jordan Howard RB 31.50 5400.0 CHI
DeAndre Hopkins WR 39.90 8500.0 HOU
Tyrell Williams WR 26.20 3700.0 LAC
Davante Adams WR 30.40 6600.0 GB
Trey Burton TE 24.10 2900.0 PHI
Jonathan Stewart FLEX 31.30 3500.0 CAR
Denver Broncos DST 18.00 3400.0 DEN
Totals 266.98 43500.0

Optimal player pool is shown below

Name Percentage
DeAndre Hopkins 0.898
Rod Smith 0.644
Dak Prescott 0.511
Jordan Howard 0.427
Jonathan Stewart 0.420
Davante Adams 0.401
Jamaal Williams 0.364
Tyrell Williams 0.352
Cooper Kupp 0.305
Kendall Wright 0.305
Adam Thielen 0.283
Mitchell Trubisky 0.277
Demaryius Thomas 0.272
Denver Broncos 0.270
Todd Gurley 0.266
Trey Burton 0.255
Golden Tate 0.237
Tyler Lockett 0.226
Torrey Smith 0.209
T.J. Yates 0.199
Detroit Lions 0.189
Giovani Bernard 0.164
Eric Ebron 0.160
Chicago Bears 0.160
Jared Cook 0.151
Kansas City Chiefs 0.147
Washington Redskins 0.134
Wayne Gallman 0.125
Garrett Celek 0.117
Adam Shaheen 0.114
O.J. Howard 0.114
Oakland Raiders 0.099
Rhett Ellison 0.089
Kyle Juszczyk 0.040
Shane Vereen 0.029
Andy Janovich 0.021
Elijhaa Penny 0.013
Bryce Petty 0.013

There’s a few strange names on there, Elijhaa Penny and Rod Smith especially. After the season, I’d really like to look into how many repeat players are in these top player pools.

Also notice that Josh Gordon is not on this list!

Thoughts on Player Replaceability

First off, can we all agree that Ezekiel Elliott is overhyped as a top RB? Alfred Morris is averaging 5 YPC with 97 touches and Eliott has 5.1 with 322 touches. And Rod Smith (who?!?) dominated last week. The Cowboys rushing attack has looked only a half-step worse with him out.

Zach Ertz is a bit trickier than Elliott. Ertz doesn’t seem like a game breaker and has played similarly most of his career. Although this season, Wentz has helped him get more touchdowns. To complicate things further Trey Burton seems to do well with Ertz out. But should you play Burton like you would Ertz?

And it certainly looks like people play replacements more than they would the starter. I seriously doubt Joe Mixon would have gotten more than 25% usage and Giovani Bernard is more like a 10% used player at best. Yet he got played in 50% of lineups last week!

Programming Category (English)468x60

You’d think the proper percentage should be less than adding the two normal percentages together because the defense can key in on one player. But perhaps there is some nonlinear relationship that comes into play because of touches. (In our optimimal lineups, Giovani Bernard was used 16% not 50%. Jordan Howard was a better play — I wonder if that has to do with time of possession differences when the opposing team can’t run as much, the Bears played the Bengals.

Or is that just a crappy narrative I’m trying to fit to one week of data.

The Almighty Belichick

Bill Belichick is a master of understanding who is replaceable and who isn’t. While I think thinking you know everything about football can backfire, I did try it. Here’s some books I bought (if anyone has some other good ones please share. I’m always interested in Matt Bowen type analysis and books like the ones below.)

Here’s a great collection of Belichick’s interviews on players/football. FYI his dad wrote a book on football scouting too

Football Scouting Methods

Bill Belichick’s dad wrote a great book on scouting. The biggest takeaway for me was how to be consistent with any scouting efforts. While I don’t think it is directly applicable to winning DFS, Jonathan Bales book is way better for that, it’s an interesting read. I’m slowly compiling a list of questions I have about each DFS pro’s play style so I can adequately scout and analyze them. Get Belichick’s book here.

Announcements

Thanks for reading! I continue to tweak the content of these posts. If you really miss a section that I used do more of let me know.

I’d love to hear your ideas on how to make these the best they can be.

Programming Category (English)468x60

Matteo Hoch

Matteo Hoch is the founder of Sports Data Direct and maintains a personal blog called Ergo Sum where he writes about data visualization and coding. You can generally find him in the DraftKings Sunday $3 early only contests under username mhoch2

Leave a Reply

Your email address will not be published. Required fields are marked *