Milly Maker Recap Methodology
Often times new readers will ask me what and why I do certain things in the recaps. This post should help explain everything I do on the blog and serve as a glossary to new readers.
I don’t expect my analysis to be above reproach so please feel free to send me suggestions, flag an issue, or critique what I’m doing. With this post my methods will be out in the open in an easy to critique and understandable format.
Milly Maker Recaps
Too much of DFS analysis focuses just on the top lineup and not enough on more predictive information. My goal with the recaps is to investigate strategies from the Milly Maker.
In my view there are 3 main ways to get that information with various trade offs.
- Who won?
- The recaps start with me looking at the winners and people on the leader board. I’ll go back and review their past Milly Maker performance and see if I can identify patterns with their bankroll, stacking, treatment of injuries, or other data points.
- The optimal lineup
- I compute the optimal lineup using the pydfs-lineup-optimizer and a pruned player pool that guarantees finding the optimal lineup with much less complexity (algo here — boy do I close my issue and submit the pull request).
- Often times you’ll see me mention the player pool of better than winner’s lineups. This player pool uses the pruned player pool that guarantees the optimal lineup is included and then I look at the player utilization for all lineups greater than the actual Milly Maker winner’s.
- Ideally I would compute all possible lineups but that is very computationally expensive so I do use this player pool subset instead
- When reading the player utilization tables a percentage of 100% means this player is indisputably the best player to play. As the percentage goes down to 0 there are more options.
- As a rough approximation higher utilization -> better play. Often times there are some surprising results given how points are distributed. Expensive guys can be worth it as well as cheap QB’s who get little to no points.
- Performance of “Sharks” or other highly regarded groups
- I compare the behavior of the “sharks”, the RotoGrinders top 50, to everyone else, the “herd”
- I analyze point distributions and player utilization differences to look for major discrepancies and hypothesize reasons why
- NEW in 2019 I will be using my own DFS player rankings that do a better job accounting for players who don’t play high volume or have more attractive point distributions (high variance)
- I also spotlight certain players from time to time in order to better understand their strategy
Why I Do It
All of these sections help us understand the crazy world of the NFL Milly Maker from different lenses.
The “who won” section gets players on our radar and shows the upper limit of human performance that week.
The optimal lineup shows us what’s possible but may never be attainable due to the inherent unpredictability of football and lineup limits.
And the player spotlights help us uncover different strategies that may be useful but were not profitable in that given week. Remember, 1 Milly Maker win is equivalent to 333 weeks of max entering, or ~19 years of football. Given the extreme payouts we have to be very careful before completely dismissing a “shark” as a bad DFS player.
Can you do a FanDuel recap?
No. They don’t provide data in an accessible format to analyze.
That’s not a stack you F***!
When I look at stacks I start from the most frequent pairings. Sometimes these aren’t “stacks” like when 2 players who don’t in the same game are paired in a high percentage of lineups. I still treat this as a “stack” but hypothesize whether it was deliberate or not.
Some DFS players start from a tight cohort of 2 players and use that as the base for risky players
Feel free to suggest your own ideas or ask any questions with a comment below