DraftKings Week 9 Recap

Quick Summary

  • jab76r wins with 245.42 points
    • 21 lineups with 60 players
    • He likely used a minimum exposure constraint
  • The Dolphins D was better than the Bears
  • Maurice Harris was better than the much talked about Adam Humphries
    • You have to decide whether you can predict breakout games or would rather base your lineups off of past performance (Bales looks like a past performance player).
  • Don’t read into ideal number of entries too quickly
    • Strategy more important
  • 165.68 to cash
  • Best possible lineup: 284.04 points

jab76r Wins With 21 Lineups and 245.42 Points

jab76r's Winning Lineup 245.42 Points

jab76r entered 21 lineups in the Week 9 Milly Maker.

Points
count 21.000000
mean 166.920000
std 30.348216
min 126.620000
25% 150.300000
50% 161.220000
75% 186.340000
max 245.420000

It wasn’t a very flukey win because his next highest lineup was 218.94 points.

jab76r Player Pool

He used 60 players across 21 lineups.

QB QB Percentage RB RB Percentage WR WR Percentage TE TE Percentage DST DST Percentage
QB Ryan Fitzpatrick 43 RB Alvin Kamara 57 WR Cooper Kupp 33 TE O.J. Howard 43 DST Chiefs 19
QB Patrick Mahomes 19 RB Kareem Hunt 52 WR Marvin Jones Jr. 33 TE Travis Kelce 19 DST Bills 19
QB Drew Brees 14 RB Christian McCaffrey 28 WR Michael Thomas 28 TE Trey Burton 14 DST Bears 19
QB Cam Newton 9 RB Melvin Gordon III 28 WR Courtland Sutton 28 TE Kyle Rudolph 14 DST Vikings 9
QB Jared Goff 5 RB Tarik Cohen 19 WR DeSean Jackson 28 TE Cameron Brate 9 DST Broncos 9
QB Mitchell Trubisky 5 RB Todd Gurley II 19 WR D.J. Moore 19 TE Benjamin Watson 5 DST Texans 9
QB Deshaun Watson 5 RB Isaiah Crowell 19 WR Brandin Cooks 19 TE Jeff Heuerman 5 DST Jets 9
RB Phillip Lindsay 9 WR Calvin Ridley 19 TE David Njoku 5 DST Steelers 5
RB Latavius Murray 9 WR Sammy Watkins 14 TE Greg Olsen 5
RB Adrian Peterson 9 WR Chris Godwin 14 TE Michael Roberts 5
RB Kenyan Drake 5 WR Laquon Treadwell 9 TE Vance McDonald 5
WR Taylor Gabriel 9
WR Mike Evans 9
WR Adam Humphries 5
WR Anthony Miller 5
WR Tyler Lockett 5
WR Keenan Allen 5
WR Tre’Quan Smith 5
WR Aldrick Robinson 5
WR David Moore 5
WR Antonio Brown 5
WR Kenny Golladay 5
WR DeAndre Hopkins 5

The amount of players he had at 5% (1/21 lineups) is a bit higher than I’m used to seeing. I think he may have a min exposure rule in his optimizer or manual process.

Other thoughts:

  • He took a chance on Jeff Heuerman having a breakout game — which he did
    • No other predicted breakouts
  • He didn’t do anything too fancy besides stacks but did always stack
  • Stacked his QB with WR and TE in most lineups except the following
    • Cam Newton with McCaffrey
    • Jared Goff with Kamara and Michael Thomas
      • No Gurley in this one — mobile QB’s typically have positive correlation with RB point totals

Jab76r Performance Other Weeks

num entered payout paid
week_num
1 1 -20.0 20.0
3 21 -60.0 420.0
4 57 -310.0 570.0
5 21 -200.0 420.0
6 20 -90.0 200.0
8 21 -70.0 420.0
9 21 1000070.0 420.0

Most weeks he enters 20 or 21 lineups. Week 4 was an exception, most likely due to the reduced entry fee of $10. But in Week 6 he kept it to 20 lineups instead of increasing. This suggests he views 20 lineups as an ideal number for his strategy.

He also hadn’t won in the Milly Maker until Week 9. And this is the type of performance you would expect from a good player. Lots of little losses and then a big win.

IT Certification Category (English)468x60

In my opinion the touts and famous players in the DFS community right now lose a bit more than you’d expect if they were truly great.

Best Lineup Possible: 284.04 Point

player_name position score salary team_name
Drew Brees QB 34.44 6100.0 NO
Alvin Kamara RB 33.60 7300.0 NO
Kareem Hunt RB 33.10 7700.0 KC
Maurice Harris WR 25.40 3300.0 WAS
Michael Thomas WR 42.10 7600.0 NO
Adam Humphries WR 28.90 3600.0 TB
Travis Kelce TE 28.90 6600.0 KC
Tevin Coleman FLEX 32.60 4800.0 ATL
Miami Dolphins DST 25.00 2800.0 MIA
284.04 49800.0

The best lineup possible in Week 9 used the Dolphins, not the Bears. And Maurice Harris was a better play than Adam Humphries.

Travis Kelce makes the list too because of the high number of cheap players that did well (Harris, Coleman, Humphries).

Pro Player Spotlight

No player spotlight this week. I’m on vacation and I focused more on the miscellaneous section.

Jon Gruden Meme

Pro vs Herd

Here is the section where I compare the Rotogrinder’s Top 50 with everyone else (their rankings are heavily skewed to high volume players but it’s the best we have… for now).

Compare their rankings with Milly Maker performance

screen_name profit paid rank
youdacao 21545.0 24000.0 1
awesemo -11845.0 24000.0 2
kingfi -170.0 370.0 3
chess_is_ok -11750.0 24000.0 4
bric75 -9615.0 24000.0 5
papagates -16015.0 24000.0 6
moklovin -8780.0 19570.0 7
thatstunna -12055.0 21000.0 8
chipotleaddict -10305.0 24000.0 9
gosixersgo76 -2445.0 9430.0 10
rikkidee 3405.0 24000.0 11
wakeywakey 325.0 24000.0 12
jayk123x -3145.0 21430.0 13
araven52 7970.0 14190.0 14
teejayortj 7855.0 24000.0 15
nolesman -13885.0 21720.0 16
csuram88 -5605.0 10880.0 17
luckypapa -9170.0 12660.0 18
nilknarf 1650.0 24000.0 19
petrgibbons -4850.0 24000.0 20
shocae -265.0 510.0 21
squirrelpatroldk -7980.0 18620.0 22
petteytheft89 -9090.0 24000.0 23
hoop2410 0.0 160.0 24
bomberfd -135.0 3640.0 25
totoroll33 -1210.0 11290.0 26
cubsfan333 -6210.0 18820.0 27
brandonadams -7965.0 21410.0 28
bdholla89 -4855.0 16200.0 29
birdwings 140.0 480.0 30
scout326 -15705.0 24000.0 31
jbcjbcjbc -10205.0 24000.0 32
healybj19 -130.0 220.0 33
3rd_and_schlong -70.0 330.0 34
ragepaladin -295.0 10300.0 35
skipbidder -6655.0 18480.0 36
joemarino27 4820.0 590.0 37
coachs111 -8370.0 12960.0 39
punisheresz -140.0 140.0 40
fjbourne 4975.0 2090.0 41
jly7125 -8035.0 18460.0 42
ianj300 70.0 1080.0 44
ehafner -13290.0 24000.0 45
dinkpiece -6045.0 16000.0 46
mrnastytime91 805.0 21000.0 47
bales -845.0 4370.0 48
erocksya2 -25.0 560.0 49

araven52 and teejayortj look interesting. I will add them to the list of players to review.

Week 9 Pro Leaderboard

payout lineups_entered payout/num_entries
screen_name
gosixersgo76 180.0 20 9.000000
ianj300 70.0 4 17.500000
3rd_and_schlong -20.0 1 -20.000000
kingfi -20.0 1 -20.000000
hoop2410 -20.0 1 -20.000000
ehafner -2820.0 150 -18.800000
scout326 -2890.0 150 -19.266667
papagates -2910.0 150 -19.400000
skipbidder -2940.0 150 -19.600000
araven52 -3000.0 150 -20.000000

Lots of sixers fans this week

Sixer's Fans

Also, I think papagates may be garbage at DFS. We’ll have to look at him more to know for sure.

Week 9 Herd Leaderboard

payout lineups_entered payout/num_entries
screen_name
jab76r 1000070.0 21 47622.380952
weaz03 99980.0 1 99980.000000
otjosh 50240.0 10 5024.000000
jrozzinri 29980.0 1 29980.000000
jkujo 19970.0 5 3994.000000
powertron -2880.0 150 -19.200000
eritas2 -2940.0 150 -19.600000
equanimity1 -2970.0 150 -19.800000
kajkyle -3000.0 150 -20.000000
ending -3000.0 150 -20.000000

Ending continues to do terrible.

Pro vs Herd Point Totals

index Points_other Points_rg_top_50
count 186952.000000 4133.000000
mean 146.583385 142.756255
std 25.327491 24.917003
min 53.580000 67.560010
25% 129.300000 125.040000
50% 146.540000 141.980000
75% 163.720000 159.640000
max 245.420000 227.240000

The RotoGrinder’s Top 50 lost to the herd this week

Pro vs Herd Player Utilization Comparison

lineup_hash Percentage Pro Percentage Herd pro_favoritism
Kareem Hunt 18 33 -15.0
Bears 6 20 -14.0
O.J. Howard 10 20 -10.0
Sammy Watkins 7 14 -7.0
Adrian Peterson 16 23 -7.0
Tarik Cohen <5% 8 -5.0
Isaiah Crowell <5% 8 -5.0
DeSean Jackson 10 14 -4.0
Lamar Miller <5% 7 -4.0
John Brown <5% 7 -4.0
Cooper Kupp 21 25 -4.0
Greg Olsen 12 15 -3.0
Kenyan Drake <5% 6 -3.0
Ryan Fitzpatrick 9 12 -3.0
Adam Thielen 19 22 -3.0
D.J. Moore 21 24 -3.0
Jordan Howard <5% 6 -3.0
DeVante Parker <5% 6 -3.0
Tyreek Hill 5 7 -2.0
Christian McCaffrey 24 26 -2.0
Demaryius Thomas <5% 5 -2.0
Mike Evans 5 6 -1.0
Jared Goff 15 16 -1.0
Courtland Sutton 24 25 -1.0
Brandin Cooks 10 11 -1.0
Patrick Mahomes 9 10 -1.0
Latavius Murray 8 8 0.0
Jets 5 5 0.0
Drew Brees 7 6 1.0
Alvin Kamara 26 25 1.0
Cam Newton 24 23 1.0
Tre’Quan Smith 9 8 1.0
James Conner 7 6 1.0
Travis Kelce 18 17 1.0
Devin Funchess 11 10 1.0
Nick Chubb 21 20 1.0
Dolphins 8 7 1.0
Benjamin Watson 5 <5% 2.0
Bills 7 5 2.0
Chris Godwin 5 <5% 2.0
Ravens 5 <5% 2.0
Steelers 5 <5% 2.0
David Njoku 9 6 3.0
Julio Jones 13 10 3.0
Peyton Barber 6 <5% 3.0
Jordan Reed 6 <5% 3.0
Chiefs 13 10 3.0
Austin Hooper 7 <5% 4.0
Melvin Gordon III 7 <5% 4.0
Mark Ingram 9 5 4.0
Jarvis Landry 11 7 4.0
Kirk Cousins 10 5 5.0
Emmanuel Sanders 13 8 5.0
Michael Thomas 21 16 5.0
DeAndre Hopkins 11 6 5.0
Kyle Rudolph 22 16 6.0
Marvin Jones Jr. 16 9 7.0
Kenny Golladay 17 10 7.0
Broncos 22 14 8.0
Todd Gurley II 34 26 8.0
Laquon Treadwell 14 5 9.0
Robert Woods 23 11 12.0
Phillip Lindsay 26 8 18.0

Thoughts:

  • Shockingly Phillip Lindsay saw a lot of Pro love against a well ranked Texans run defense.
  • D.J. Moore was a very popular pick
    • Both the pro’s and herd thought that he would benefit the most from a big Cam Newton day
    • Probably because it appeared Curtis Samuel was being usurped by the rookie D.J. Moore but that narrative proved false.
  • Todd Gurley managed to have more usage at 9800 than Kamara at 7300.
    • The Saints are first in yards allowed against the run
    • 31st against the pass.

Misc

How Many Lineups Should You Enter?

Many people think that in multi-entry you should max enter if you can afford it. There was some interesting commentary you should read in /r/dfsports.

Also, Samsondfstruth posts payouts versus number of lineups entered and the 50-99 lineups group does well but this can give you confusing results.

Personal Development Category (English)468x60

I would caution you from making any inferences on this data beyond that it is possible to win with any number of lineups at this time. And this is why.

Take a look at another grouping of entries.

paid payout percent_of_entrant percent_of_paid percent_of_winnings percent_of_payout
number of entries
1 10389650.0 -3522840.0 64.287034 25.418303 19.406893 -64.152823
2-9 16215350.0 -2909855.0 32.375162 39.670892 37.603825 -52.990034
10-24 5282650.0 -637345.0 2.492254 12.924016 13.128504 -11.606397
25-49 2247930.0 2261075.0 0.452269 5.499566 12.743294 41.175399
50-149 2701100.0 370335.0 0.223563 6.608247 8.680452 6.744001
150 4038000.0 -1052695.0 0.169719 9.878976 8.437032 -19.170146

Here I used different number of entry groupings and you could draw a different conclusion from Samson’s post. The 25-49 group has a better percent of winnings compared to their percent of paid.

But as you increase the number of lineups in a group the expected value of payouts goes way down because of the rake.

paid payout percent_of_entrant percent_of_paid percent_of_winnings percent_of_payout
number of entries
1 10389650.0 -3522840.0 64.287034 25.418303 19.406893 -64.152823
2-35 22880970.0 -1045820.0 35.180824 55.978346 61.710231 -19.044948
36-149 3566060.0 130030.0 0.362424 8.724374 10.445844 2.367917
150 4038000.0 -1052695.0 0.169719 9.878976 8.437032 -19.170146

It would be better to look at what percentage of players make money in each group rather than combining everyone because

  • more players is negative EV
  • Wins are very bursty
    • This skews the payout by a large amount in a small group that has the Milly Making lineup in it
  • Number of entry groupings are a poor proxy to strategy
    • CONDIA would manually enter 150 lineups!
    • Some players use an optimizer for 1 lineup

Also, just changing the number of lineups you enter with disregard to strategy is not a great idea. You need to know what strategies are used successfully by different players at different bankrolls. And apply a strategy that works with your bankroll. Don’t adjust your bankroll to match some bad statistical inference.

Business Category (English) 468x60

I’ll build on this idea more through my rankings system.

Rankings Coming Soon

Ratings are hard because of the insane payout structure. One first place win makes 333 weeks of max entering in the Milly worth it.
Still you’d expect the better players to come close sometimes. Here’s what I’m thinking right now. Feel free to comment with your ideas

  • Should look at weekly snapshots only
    • Should be comparative within the weeks e.g. compare player’s point totals to the average for that week rather than a season long average
    • Comparisons should bounded in such a way that follows Pareto distributions
  • Composed of points, variance, volume
  • Consist only of players that have played in majority of contests
    • this implies they think they may have an edge rather than a one and done player.
  • Should be scaled to not overcompensate for volume yet still factor in volume in case of similar players. perhaps use payout/entry, n_entries and a payout/week
  • one-dimensional player ratings to start

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 *