DraftKings Week 2 Recap

JEREMYHEIN's Winning Lineup 237.84

Patrick Mahomes Meme

Quick Facts

JEREMYHEIN won with 237.84.

  • He stacked all of his QB’s and used risky players in each stack.
  • The best possible was 279.44 and it was 153.98 to cash.
  • Pro’s favored T.J. Yeldon and Tevin Coleman much more than the herd.
  • The Herd overplayed Jared Cook

JEREMYHEIN’s Lineup Strategy

JEREMYHEIN entered 26 lineups and his best lineup scored 237.84.

JEREMYHEIN's Winning Lineup 237.84

It turns out this score was a bit of an outlier. Here’s a look at all of his lineups

JEREMYHEIN's Point Distribution

You can see what herd did in green and his lineups in blue. Notice that his distribution looks approximately normal with only one lineup scoring above 200 points.

Typically you want to engineer your lineups to have a point distribution with a fat high scoring tail like how DFS Pro CONDIA did in week 8 last year.

CONDIA's Point Totals

We’ll give JEREMYHEIN some slack. Perhaps with more entries his point totals would match CONDIA. And besides JEREMYHEIN is the week 2 champ after all.

Analyzing JEREMYHEIN Lineup Construction

JEREMYHEIN entered 49 players

QB QB Percentage RB RB Percentage WR WR Percentage TE TE Percentage DST DST Percentage
QB Deshaun Watson 23 RB Todd Gurley II 39 WR Antonio Brown 39 TE David Njoku 39 DST Texans 31
QB Matthew Stafford 12 RB Melvin Gordon III 35 WR DeAndre Hopkins 31 TE George Kittle 23 DST Vikings 19
QB Patrick Mahomes 8 RB Corey Clement 31 WR Marvin Jones Jr. 31 TE Jordan Reed 16 DST Buccaneers 12
QB Kirk Cousins 8 RB Kareem Hunt 23 WR Kenny Golladay 27 TE Travis Kelce 8 DST Jets 12
QB Blake Bortles 8 RB Christian McCaffrey 19 WR Stefon Diggs 19 TE Jonnu Smith 8 DST Broncos 12
QB Cam Newton 8 RB Phillip Lindsay 19 WR Will Fuller V 19
QB Tyrod Taylor 8 RB Chris Thompson 19 WR Sammy Watkins 16
QB Case Keenum 8 RB Duke Johnson Jr. 19 WR Ted Ginn Jr. 16
QB Aaron Rodgers 8 RB Dalvin Cook 16 WR JuJu Smith-Schuster 12
RB Royce Freeman 12 WR Keelan Cole 12
RB Kenyan Drake 8 WR Dede Westbrook 12
RB Adrian Peterson 8 WR Jarvis Landry 12
RB Lamar Miller 8 WR Julio Jones 8
RB Alvin Kamara 8 WR Demaryius Thomas 8
WR Robby Anderson 8
WR Randall Cobb 8

He looks hot on the Texans. He also stacked all of his QB’s:

  • Deshaun Watson-DeAndre Hopkins 23%
  • Matthew Stafford-Marvin Jones-Kenny Golladay 13%
  • Patrick Mahomes-Kareem Hunt-Travis Kelce 8%
  • Blake Bortles-Keelan Cole-Dede Westbrook 8%
  • Cam Newton-Christian McCaffrey 8%
    • (QB-RB stacks work well when QB is mobile)
  • Tyrod Taylor-Jarvis Landry-David Njoku 8%
  • Case Keenum-Demaryius Thomas 8%
  • Aaron Rodgers-Randall Cobb 8%

Interestingly enough, none of these stacks involved an elite QB-WR pair, perhaps Watson-Hopkins but they haven’t returned to 2017 yet.

Did He Use an Optimizer To Create His Lineups?

Some weeks it’s easy to tell, some weeks not. I’d reckon 70% chance he used an optimizer and 30% chance he just manually entered lineups.

Limiting his single player exposure below 40% is telling, and common for multi-entry, but using just 26 lineups makes it difficult to get the magic lineup.

Congratulations to JEREMYHEIN, $510 down and $1 million back is a great day.

Best Lineup Possible: 279.44 Points

We had a little snafu last week where our best possible lineup wasn’t accurate. This week we feel much more confident but please let me know if you find an issue.

The best lineup outscored JEREMYHEIN by over 40 points, further supporting this author’s belief not to worry about being too contrarian with your lineups. There are way too many better lineups out there and you should focus on finding them.

player_name position score salary team_name
Patrick Mahomes II QB 41.84 6100.0 KC
Matt Breida RB 27.90 4800.0 SF
Christian McCaffrey RB 30.90 7000.0 CAR
Adam Thielen WR 34.10 7100.0 MIN
Stefon Diggs WR 38.90 6900.0 MIN
JuJu Smith-Schuster WR 34.10 6400.0 PIT
Travis Kelce TE 32.90 5900.0 KC
Jesse James FLEX 27.80 2800.0 PIT
Green Bay Packers DST 11.00 2500.0 GB
279.44 49500.0

Thoughts

The best possible lineup features Adam Thielen, Stefon Diggs, AND their opponent… the Green Bay Packers defense. Showing an exception to the rule not to play against your own defense.

In this case the blocked punt touchdown may have caused positive correlation to the points scored. But on further review, on average it leads to less total points scored (source: Sports Data Direct Matchup and Play by Play data since 2011)

Total Points vs Punt Recovery Touchdown
Point totals are discounted to account for points that were scored by the punt td recovery and wouldn’t contribute any more to overall point totals. Read about the Skellam distribution and dependent Poisson distributions for more on why this may be true.

The Patrick Mahomes -Travis Kelce stack worked wonders. Also, Matt Breida presented enormous value with little downside at the RB position.

Pro’s

chess_is_ok

We’re now tracking chess_is_ok. He took home the top prize in  Week 13 last year  and he won the Draft Kings World Championship. His strategy featured single player utilization capped at 50% and stacking all his QB’s.

In Week 2, chess_is_ok entered 150 lineups and used 96 players. This size player pool implies that he used exclusion criteria to reduce his player pool or just an everyday optimizer. Perhaps he is a Karl Popper falsification disciple but for now that’s a reach.

He entered 150 lineups and lost $480. Here’s his player usage

QB QB Percentage RB RB Percentage WR WR Percentage TE TE Percentage DST DST Percentage
QB Cam Newton 15 RB Todd Gurley II 46 WR Antonio Brown 38 TE George Kittle 23 DST Texans 16
QB Nick Foles 11 RB Tevin Coleman 46 WR Devin Funchess 21 TE David Njoku 20 DST Titans 14
QB Ben Roethlisberger 10 RB Alvin Kamara 41 WR Demaryius Thomas 21 TE Zach Ertz 20 DST Bills 13
QB Matthew Stafford 10 RB James Conner 29 WR Kenny Golladay 19 TE Ricky Seals-Jones 17 DST Vikings 11
QB Patrick Mahomes 9 RB LeSean McCoy 17 WR Larry Fitzgerald 19 TE Austin Hooper 15 DST Saints 11
QB Blake Bortles 8 RB Dalvin Cook 13 WR DeAndre Hopkins 18 TE Jared Cook 13 DST Steelers 9
QB Matt Ryan 8 RB Melvin Gordon III 11 WR Julio Jones 16 TE Jack Doyle 13 DST Jets 8
QB Case Keenum 7 RB T.J. Yeldon 9 WR Randall Cobb 14 TE Travis Kelce 7 DST Buccaneers 7
QB Tyrod Taylor 6 WR Dante Pettis 14 TE Austin Seferian-Jenkins 6
WR Rashard Higgins 13
WR Mohamed Sanu 13
WR Nelson Agholor 11
WR Jarvis Landry 11
WR JuJu Smith-Schuster 9
WR Donte Moncrief 8
WR Pierre Garcon 8
WR Geronimo Allison 7
WR Golden Tate 7
WR Emmanuel Sanders 7
WR Sammy Watkins 6
WR Marvin Jones Jr. 6
WR Quincy Enunwa 6
WR Mike Wallace 6

Wow! Look at all those WR’s. chess_is_ok must view them as hard to predict.

Here’s a look at his frequent pairs

Percentage Player
20 RB Tevin Coleman, RB Todd Gurley II
20 RB Tevin Coleman, WR Antonio Brown
17 RB Alvin Kamara, RB Tevin Coleman
15 RB Todd Gurley II, WR Antonio Brown
15 QB Cam Newton, WR Devin Funchess
13 RB Todd Gurley II, WR Demaryius Thomas
13 RB Todd Gurley II, TE George Kittle
12 RB Todd Gurley II, WR Devin Funchess
12 RB James Conner, RB Todd Gurley II
12 RB Todd Gurley II, TE David Njoku
12 RB Tevin Coleman, WR Demaryius Thomas
12 RB Tevin Coleman, WR Kenny Golladay
12 TE George Kittle, WR Antonio Brown
12 RB Alvin Kamara, TE Zach Ertz
12 WR Julio Jones, WR Mohamed Sanu

The closest parallel I can find to chess_is_ok is CONDIA but even CONDIA wouldn’t try a QB-WR stack much higher than 8%. This suggests chess_is_ok may be more confident in his projections/his view of player performance while still using an optimizer.

It doesn’t look like chess_is_ok went heavy on game stacks (choosing multiple players from the same game).

Time will tell if these lineup strategies hold true for the season but we’re excited to keep watching chess_is_ok enter contests.

Al_Smizzle Lost $1,560

Al_Smizzle lost $1,560 on 150 entries. True to form, his player pool was just 53 players across 150 lineups (chess_is_ok used almost twice that!).

Here’s Al_Smizzle’s player usage.

QB QB Percentage RB RB Percentage WR WR Percentage TE TE Percentage DST DST Percentage
QB Patrick Mahomes 15 RB James Conner 55 WR Dante Pettis 40 TE David Njoku 26 DST Texans 26
QB Matt Ryan 15 RB Melvin Gordon III 30 WR Michael Thomas 29 TE George Kittle 26 DST Redskins 25
QB Tyrod Taylor 15 RB Tevin Coleman 26 WR Mike Williams 25 TE Jack Doyle 25 DST Rams 25
QB Jimmy Garoppolo 15 RB Todd Gurley II 26 WR Quincy Enunwa 24 TE Jared Cook 10 DST Chargers 24
QB Drew Brees 11 RB Alvin Kamara 26 WR Mohamed Sanu 21 TE Zach Ertz 8
QB Case Keenum 9 RB Christian McCaffrey 20 WR Demaryius Thomas 20 TE Travis Kelce 6
QB Ben Roethlisberger 8 RB James White 19 WR Devin Funchess 20
QB Deshaun Watson 8 RB Kareem Hunt 15 WR Emmanuel Sanders 19
QB Matthew Stafford 6 RB Peyton Barber 10 WR Cooper Kupp 19
RB Dalvin Cook 9 WR Jarvis Landry 19
WR Kenny Golladay 19
WR Antonio Brown 17
WR Tyreek Hill 15
WR Julio Jones 15
WR Tyrell Williams 10
WR Courtland Sutton 10
WR Chris Godwin 9
WR JuJu Smith-Schuster 9
WR Corey Davis 7
WR Kenny Stills 6

Kinda surprising to see how poorly he did even with Patrick Mahomes as his top QB. Looking at his stacks, it looks like Al_Smizzle is starting to use more deliberate stacks.

  • Patrick Mahomes-Tyreek Hill 12%
  • Jimmy Garoppolo-George Kittle 12%
  • Drew Brees-Michael Thomas 9%
  • Case Keenum-Emmanuel Sanders 8%

Remarkably, all these stacks aren’t equal to the amount of QB utilization but just a few % lower. This implies Al_Smizzle is forcefully ensuring QB’s aren’t always stacked. There were also a few triple stacks but it’s too hard to tell if those are deliberate or chancy due to his small player pool

We’ll keep track to see if this trend continues.

ending Wins $2,520

ending hasn’t opted into Rotogrinder’s tracking but consistently seems to do well. Last year he won week 12 with a CONDIA esque approach.

In week 2, he entered 150 lineups and won $2,520. His player pool consisted of 87 players and his most common players are shown below

QB QB Percentage RB RB Percentage WR WR Percentage TE TE Percentage DST DST Percentage
QB Kirk Cousins 66 RB Tevin Coleman 66 WR Adam Thielen 74 TE George Kittle 36 DST Titans 28
QB Aaron Rodgers 27 RB James Conner 54 WR Randall Cobb 69 TE Rob Gronkowski 29 DST Texans 13
RB Melvin Gordon III 23 WR Geronimo Allison 51 TE Kyle Rudolph 24 DST Buccaneers 12
RB Alvin Kamara 23 WR Stefon Diggs 37 TE Zach Ertz 17 DST Falcons 10
RB Todd Gurley II 19 WR Demaryius Thomas 18 TE David Njoku 7 DST Steelers 9
RB Dion Lewis 12 WR DeAndre Hopkins 13 DST Redskins 8
RB Dalvin Cook 9 WR Jarvis Landry 12 DST Bills 7
RB T.J. Yeldon 8 WR Pierre Garcon 8
WR Corey Davis 6
WR Larry Fitzgerald 6
WR Antonio Brown 6

This single-player utilization is quite interesting. He looks like he takes a few concentrated big risks and then a lot of smaller ones (87 players in his player pool and just 33 players above with 5% utilization or more)

Stacks

  • QB Kirk Cousins, WR Adam Thielen 51%
  • QB Aaron Rodgers, WR Randall Cobb 26%
  • QB Kirk Cousins, WR Adam Thielen, WR Stefon Diggs 23%
  • QB Aaron Rodgers, WR Geronimo Allison, WR Randall Cobb 23%

We haven’t tracked ending much yet but the 23% triple stack looks interesting. We’ll have to see if that’s an outlier or normal for his strategy.

Overall, ending looks like a super risky version of CONDIA. Big player pool yet most of his lineups use similar players ala Al_Smizzle or Tommy G.

Perhaps he’s very confident in predicting a cohort of players and then views the rest as some sort of “edit distance” or Levenshtein distance applied to lineup construction. That would surely make backtesting easier. For now this is just speculation.

Pro vs Herd

The Herd

Here is the section where we compare the Rotogrinder’s Top 50 vs everyone else (the herd).

The Pro’s

The pro’s accounted for 1.6% of all entries and took home $171,895

The best and worst entrants

payout lineups_entered payout/num_entries
screen_name
4u2nvagain 150270.0 5 30054.000000
rockchalk88 49800.0 16 3112.500000
asmitty914 10060.0 4 2515.000000
moklovin -2500.0 150 -16.666667
mrnastytime91 -2630.0 150 -17.533333
awesemo -2760.0 150 -18.400000

It’s always surprising how big names end up on the bottom. Last week it was youdacao and papagates. This time it’s awesemo but hey you can still pay to read all of his picks on his blog?!

Hopefully Awesemo’s performance is just a fluke and he returns to form. Unlike the late greats: CSURAM88, Bales, and SaahilSud (there’s still hope for them in NFL DFS but it’s been a while).

I just hate to see people following the fake expert pattern of lucky win and selling advice instead of using knowledge to make $$$. I want to believe in experts!

The Herd

As a group the herd paid $5,194,720 and lost $849,030.

The best and worst players not in the Rotogrinder’s Top 50

payout lineups_entered payout/num_entries
screen_name
jeremyhein 1000145.0 26 38467.115385
aperls3206 100040.0 3 33346.666667
zeke1234 29980.0 1 29980.000000
fierygold -3000.0 150 -20.000000
powertron -3000.0 150 -20.000000
jdm68a -3000.0 150 -20.000000

Pro vs Herd

Remarkably similar performance this week.

index Points_herd Points_pro
count 259736 4246
mean 137.869966 137.940862
std 22.217274 22.482617
min 5.000000 74.300000
25% 122.240000 121.840000
50% 136.780000 136.880000
75% 152.580000 152.725000
max 237.840000 237.740000

Player utilization differences shown below (full table in Appendix)

lineup_hash Percentage Pro Percentage Herd pro_favoritism
Jared Cook 8 21 -13.0
Tyreek Hill 6 18 -12.0
Alvin Kamara 12 23 -11.0
Jarvis Landry 12 22 -10.0
Adrian Peterson <5% 12 -9.0
Emmanuel Sanders 6 15 -9.0
Jets 7 15 -8.0
JuJu Smith-Schuster 5 13 -8.0
Chris Thompson <5% 8 -5.0
Keenan Allen <5% 8 -5.0
Pierre Garcon 10 <5% 7.0
Julio Jones 18 11 7.0
George Kittle 38 31 7.0
Rob Gronkowski 10 <5% 7.0
Mohamed Sanu 10 <5% 7.0
Redskins 11 <5% 8.0
Christian McCaffrey 20 12 8.0
Antonio Brown 37 28 9.0
T.J. Yeldon 16 6 10.0
Tevin Coleman 41 26 15.0

Interestingly enough James Conner saw similar love from pro’s and the herd at ~ 38% (see Appendix).

The pro’s were more into backups thrust into a starting role and ignoring recency bias (see Jared Cook usage differences and Yeldon and Colememan)

Announcements

That’s all thanks for reading. Let me know your feedback and subscribe to our mailing list.

DFS Pro Player Profiles

I’m still working on the DFS pro player profiles article, I got married last weekend in Vegas and am about to go on my honeymoon so it’s been real busy.

Thoughts: I think the categorization of pro player styles is a bit more continuous than I thought. CONDIA and chess_is_ok are both optimizer players but how they use constraints within the optimizer is telling about their beliefs in their own strengths. So far I think stacks (positional, game), player pool size, remarkable position treatment, and exposure limits are the most telling attributes. I’ll try to create a pro tendency matrix using those attributes.

Let me know if you have thoughts on any other attributes to focus on.

DFS Pro’s Missing

I didn’t write about a few pro’s this week. chipotleaddit, tommyg1979, and Bales. That’s because they all did pretty poorly and their playstyles seem consistent at this point.

Appendix

lineup_hash Percentage Pro Percentage Herd pro_favoritism
Jared Cook 8 21 -13.0
Tyreek Hill 6 18 -12.0
Alvin Kamara 12 23 -11.0
Jarvis Landry 12 22 -10.0
Adrian Peterson <5% 12 -9.0
Emmanuel Sanders 6 15 -9.0
Jets 7 15 -8.0
JuJu Smith-Schuster 5 13 -8.0
Chris Thompson <5% 8 -5.0
Keenan Allen <5% 8 -5.0
Jordan Reed <5% 7 -4.0
Kenny Stills <5% 7 -4.0
Tyrod Taylor <5% 7 -4.0
Drew Brees <5% 7 -4.0
Vikings <5% 6 -3.0
Ted Ginn Jr. <5% 6 -3.0
Deshaun Watson <5% 6 -3.0
Jay Ajayi <5% 6 -3.0
Case Keenum <5% 6 -3.0
Quincy Enunwa 7 10 -3.0
Chargers <5% 6 -3.0
Patrick Mahomes 8 11 -3.0
Nelson Agholor 5 8 -3.0
Rams 6 9 -3.0
Isaiah Crowell <5% 5 -2.0
Philip Rivers <5% 5 -2.0
Mike Williams <5% 5 -2.0
Dion Lewis 7 8 -1.0
Melvin Gordon III 18 18 0.0
Kareem Hunt 8 8 0.0
Dalvin Cook 6 6 0.0
James White 9 8 1.0
Randall Cobb 11 10 1.0
Ben Roethlisberger 9 7 2.0
Nick Foles 5 <5% 2.0
Jamison Crowder 5 <5% 2.0
Todd Gurley II 19 17 2.0
Jack Doyle 7 5 2.0
Golden Tate 6 <5% 3.0
James Conner 41 38 3.0
Dante Pettis 17 14 3.0
Jimmy Garoppolo 11 8 3.0
DeAndre Hopkins 10 7 3.0
Larry Fitzgerald 7 <5% 4.0
Rex Burkhead 7 <5% 4.0
Devin Funchess 12 8 4.0
Michael Thomas 14 10 4.0
Matthew Stafford 8 <5% 5.0
Matt Ryan 8 <5% 5.0
Titans 8 <5% 5.0
Zach Ertz 9 <5% 6.0
Demaryius Thomas 16 10 6.0
Kenny Golladay 19 13 6.0
Texans 29 23 6.0
Travis Kelce 10 <5% 7.0
David Njoku 15 8 7.0
Pierre Garcon 10 <5% 7.0
Julio Jones 18 11 7.0
George Kittle 38 31 7.0
Rob Gronkowski 10 <5% 7.0
Mohamed Sanu 10 <5% 7.0
Redskins 11 <5% 8.0
Christian McCaffrey 20 12 8.0
Antonio Brown 37 28 9.0
T.J. Yeldon 16 6 10.0
Tevin Coleman 41 26 15.0

Leave a Reply