What's new
Fantasy Football - Footballguys Forums

Welcome to Our Forums. Once you've registered and logged in, you're primed to talk football, among other topics, with the sharpest and most experienced fantasy players on the internet.

Results of letting FBG choose my lineups everyweek (1 Viewer)

palmerville

Footballguy
Setup:

Thought it would be entertaining to write a small script that uses the FBG projections to blindly generate lineups that I would play in cash games.  This is basically extending the interactive value chart functionality that fills in partial lineup to generate the full optimal lineup given each projection.  So take every possible valid lineup combination and score it for each prediction (DD, MT, SB, AV).  Play the highest scoring lineup of each.

Data starts in week 4 as that's when I had the idea.  Deposited $25 in both Yahoo and Fanduel.

Ending up playing in over 400 contests, hope the results are interesting.

Rules:

  1. Only play 50/50 and Double-up cash games
  2. Only play main slate games
  3. Wager 75% of bankroll every week
  4. Enter contents ranging from $1 - $50 
  5. Play 4 lineups (DD, MT, SB, AV) in each site.  Multiple contests per lineup.
  6. Equally divide wager across the 4 lineups
  7. Set lineups at 11:15 (Central)
Results:

Fanduel:


Contests


226


Wins


155


Losses


71




Avg AV Score


133.69


Avg DD Score


129.85


Avg MT Score


124.37


Avg SB Score


119.51




Avg Win


137.18


Avg Loss


103.02


Min Win


96.94


Max Loss


129.78




 


AV


DD


MT


SB


Wins


49


42


30


34


Losses


18


13


22


18


%


73%


76%


58%


65%



Starting bankroll: $25.00

Ending bankroll: $245.45

Yahoo:


Contests


179


Wins


111


Losses


68




Avg AV Score


118.44


Avg DD Score


122.48


Avg MT Score


126.12


Avg SB Score


118.18




Avg Win


130.16


Avg Loss


101.86


Min Win


101.94


Max Loss


131.52




 


AV


DD


MT


SB


Wins


24


27


33


27


Losses


26


16


9


16


%


48%


63%


79%


63%

Starting bankroll: $25.00

Ending bankroll: $64.28

Conclusion:

Bankroll was subject to wild swings as it would go for a few weeks winning 100% and then followed by just as many weeks where it would win 0%.  Also not every prediction (DD, MT, SB, AV) preformed as well as the others.  For Fanduel, AV and DD were clear winners with SB lagging.

Full weekly lineup results:


 


Fanduel


 


 


 


Yahoo


 


 


 


 


AV


DD


MT


SB


AV


DD


MT


SB


4


126.70


108.10


122.34


138.02


144.20


149.20


134.44


134.44


5


130.24


131.64


132.94


106.14


124.20


153.00


111.80


112.38


6


122.08


129.78


102.88


113.48


134.58


138.74


121.38


129.28


7


150.28


143.02


137.48


138.48


135.32


131.52


131.82


129.94


8


122.38


122.12


109.90


78.48


97.38


108.18


114.18


69.78


9


154.48


104.14


166.48


141.71


132.88


132.88


167.98


131.28


10


156.84


160.84


136.52


139.82


108.04


124.24


143.74


106.24


11


110.58


139.58


85.48


115.08


117.24


117.78


139.90


168.28


12


119.64


96.94


89.24


106.14


88.14


101.94


102.82


111.84


13


156.96


146.26


151.96


144.98


133.46


132.06


133.06


122.62


14


151.34


154.24


113.32


127.24


110.64


99.94


76.54


108.34


15


100.34


94.46


114.84


84.58


91.18


86.80


106.98


100.04


16


147.46


152.52


155.42


140.38


124.82


135.72


139.92


128.82


17


122.40


134.20


122.40


98.62


116.10


102.70


141.10


101.18


 


 


 


 


 


 


 


 


 


 


133.69


129.85


124.37


119.51


118.44


122.48


126.12


118.18






 
Nice, but seems like a lot of work each week to come up with the ultimate lineup based on projections.  Why can't FBG just do that work for you and spit out a lineup,(or 3 or 4).   But then, I guess those lineups would be too heavily used.

 
Last edited by a moderator:
Nice job.  I'm a nerd software developer so I'd like to know more about the script/process you used to generate the ideal lineups.  Did you use javascript to extend the existing IVC functionality, or did you import the data into something else (Excel, database, etc.) and then use some other software to determine the lineups?

 
Nice job.  I'm a nerd software developer so I'd like to know more about the script/process you used to generate the ideal lineups.  Did you use javascript to extend the existing IVC functionality, or did you import the data into something else (Excel, database, etc.) and then use some other software to determine the lineups?
Process is basically:

  1. Download IVC pages (Fanduel and Yahoo)
  2. Python to parse pages and load all players into sqlite db.  (overkill but wanted to play with sqlite and had an initial thought to load all generated lineup into the db to play with results, but that wasn't needed nor where they interesting)
  3. Update settings to exclude players in certain games (Only played main slate of games)
  4. Python script builds and calculates every possible lineup, first testing to fit under salary cap and then for total points*
  5. Spits out the top lineup for each projection (DD, MT, SB, AV)
*Doesn't really do all lineups as that would take hours, it first pre-processes the players to remove those that wouldn't get selected anyway.  Example:


Name


Pos


Game


Salary


MT


Aaron Rodgers


QB


GB@DET


8800


22.7


Drew Brees


QB


NO@ATL


8200


20.8


Russell Wilson


QB


SEA@SF


7600


20.5


Matt Ryan


QB


NO@ATL


8700


19.8

We can discard Matt Ryan as he would never be in a top lineup.  His salary is higher and points lower than Russell Wilson.  It gets a little more involved for RB/WR as you are selecting for multiple position slots.

 
Great job.  Did you do any other analysis reviewing the weeks or season?

Week 14 was one time of a few this year on FD where the GPP cut lines were lower than the cash contests cut lines.  So the chalk went nuts, as did DD and AV scores above.  MT and SB were crushed that week.  Did you notice the optimal DD lineup more consistently pulled in chalk plays?  Did the others seem to reach more?  

 
I experimented on a smaller scale during the middle weeks of this past season.  Based on my own selection process, I determined my preferred:

  • 3 QBs
  • 6 RBs
  • 9 WRs
  • 3 TEs
  • 3 Def
  • 3 PK
and used Excel to generate all possible lineups on Fanduel, and their corresponding fbg projections.  I then cherry picked 9 that I liked the most and played each lineup in both a 50-50 and 1/4K Sun NFL Mini Tourney at only $1 entry levels.  One of my criteria, for example, was trying to spread out my exposure to any one player.  I didn't want each lineup to have David Johnson, no matter how much of a roll he was on.  I did this for 4 or 5 weeks, and came out ahead.   So not a big $ investment or return, but an interesting exercise, and I did not experience the wild swings of the original author.  Some of my steps took too much time to repeat each week, so I could only do it certain weeks.  I plan to hone the process in the off-season, so I can try it in earnest in 2017 on a more consistent/regular basis.

 
Last edited by a moderator:
Process is basically:

  1. Download IVC pages (Fanduel and Yahoo)
  2. Python to parse pages and load all players into sqlite db.  (overkill but wanted to play with sqlite and had an initial thought to load all generated lineup into the db to play with results, but that wasn't needed nor where they interesting)
  3. Update settings to exclude players in certain games (Only played main slate of games)
  4. Python script builds and calculates every possible lineup, first testing to fit under salary cap and then for total points*
  5. Spits out the top lineup for each projection (DD, MT, SB, AV)
*Doesn't really do all lineups as that would take hours, it first pre-processes the players to remove those that wouldn't get selected anyway.  Example:


Name


Pos


Game


Salary


MT


Aaron Rodgers


QB


GB@DET


8800


22.7


Drew Brees


QB


NO@ATL


8200


20.8


Russell Wilson


QB


SEA@SF


7600


20.5


Matt Ryan


QB


NO@ATL


8700


19.8

We can discard Matt Ryan as he would never be in a top lineup.  His salary is higher and points lower than Russell Wilson.  It gets a little more involved for RB/WR as you are selecting for multiple position slots.
Nice.  I did something similar in years past, but what I did differently was use a linear programming class to do the math.  You give it all of the players and projected scores, your constraints (position lineup, player cost etc) and it will give you the best lineup.  The learning curve was pretty high for me to get it working, but once I did, it was great as it was 100% based on the numbers.  

Here's the link to lp_solve if you're interested

http://lpsolve.sourceforge.net/5.5/

 

Users who are viewing this thread

Top