DraftKings DFS Analytics
ML-driven projections and lineup optimizers, built for how DraftKings contests actually pay out.
Wilke Analytics runs unified machine-learning models and constraint-based optimizers for MLB, NFL, and NBA — projections, ceiling estimates, and lineup construction in one place, built by a quant for his own play and opened to subscribers.
What's inside
Unified LightGBM models
A single global model per sport trained across the full player-game history, with per-player bias correction — not thousands of thin per-player models.
Ceiling models
A separate quantile model estimates upside independently of the mean projection, so GPP lineups can target ceiling instead of guessing at variance.
Vegas-aware stacking
Implied team totals and game environments feed directly into stack selection and player weighting — the market's read on scoring, not just box-score history.
Confirmed-lineup protection
Optimizers gate on confirmed starting lineups and late-scratch signals before a slate locks, instead of trusting stale projection data.
Constraint-based optimization
OR-Tools solvers build cash and GPP lineups under real salary, position, and correlation constraints — not a spreadsheet sort by points-per-dollar.
MLB, NFL, and NBA
The same modeling and optimization approach applied across three sports, tuned to how each one's DraftKings contests actually get won.
Built by a quant, for his own play
Wilke Analytics started as a personal system for playing DraftKings MLB, NFL, and NBA contests — models, optimizers, and data pipelines built to be right rather than to look impressive. It's now open to subscribers who want the same tools. No fabricated track record, no promised profits: DFS is a game of skill with real variance, and these are analytics to help you make better decisions inside that variance, not a guarantee of winning.
Product
MLB Lineup Optimizer
Contiguous batting-order stacks, Vegas implied-total tilts, and separate cash and GPP modes.
NFL Lineup Optimizer
QB + pass-catcher stacks with bring-backs, allocated by game environment, optimized toward ceiling.
Research
We publish the same data-driven research that drives our own optimizer decisions — real DraftKings contest data, not theory.
What Wins DraftKings MLB GPP Contests: A 76,000-Lineup Study
We analyzed ~76,000 opponent lineups across 24 DraftKings MLB GPPs to find what actually separates top finishers from the field.
MLB vs. NFL DFS Stacking: Team Correlation vs. Game Correlation
Why MLB stacking targets a team and NFL stacking targets a game — evidence from real DraftKings contest data.