Load the slate, choose the build mode and create the lineup portfolio.
ACTIVE PROTOTYPESPORTS ANALYTICS / SIMULATIONMLB + NFL ACTIVE · NBA + NHL INTEGRATION UNDERWAY
FanDuel ATI Engine
Working multi-sport DFS desktop software. MLB + NFL are active today, with NBA + NHL integration underway. The engine covers lineup building, slate intelligence, portfolio construction, late swap, simulation, exposure controls and learning from completed contests.
INSIDE THE ATI ENGINE
Real screens from the working desktop application.
Choose a workflow stage to see the actual software used for slate preparation, live management and postgame learning.
01 / 05
Build Lineups
Creates contest-aware lineups for currently supported MLB and NFL slates, with NBA and NHL being added to the same multi-sport workflow.
WHAT THIS SCREEN SHOWS
Slate files, reserved contests, optional projection override, build mode, slate status and adaptive Monte Carlo settings are brought together in the primary operating screen.
WORKING SOFTWARE
Five screens. One connected operating workflow.
Update intelligence, projections, ensembles and calibration inputs.
Manage late news, lock state, lineup watchers and affected-entry repairs.
Import completed contest history and carry bounded lessons forward.
Keep testing and non-standard settings away from the normal live workflow.
THE PROBLEM
What I was trying to solve
A lineup generator can produce valid entries and still perform badly. The harder problem is portfolio construction: player exposure, correlation, contest type, late news, diversification, slate size, historical results and avoiding repeated mistakes.
THE APPROACH
How I structured it
The engine evolved through live testing and repeated post-slate review. Features were added after specific failures: lineup concentration, inactive-player risk, stale projections, last-minute CSV problems, insufficient diversity and weak learning from prior slates.
WHAT I DESIGNED
From concept to working system.
Portfolio generation
Build multiple unique lineups while controlling concentration and maintaining viable alternatives.
Simulation
Use Monte Carlo concepts to compare candidate lineups under uncertainty rather than relying only on one projection.
Exposure controls
Track player usage, MVP exposure and portfolio concentration across the full set of entries.
Late-swap workflow
Support updating later-game players as news changes while respecting already-locked roster positions.
Contest awareness
Treat small-field contests differently from large-field tournament entries.
Historical learning
Track engine version, contest type, score, finish, ROI and other results so each slate can inform the next.
TOOLS / CONCEPTS
Technology used or explored
PythonSimulationOptimizationCSV workflowsPortfolio analyticsLate swapHistorical tracking
WHAT I LEARNED
This project reinforced that optimization is not the same as prediction. A model needs reliable inputs, constraints, feedback loops and portfolio-level reasoning. It also taught me how quickly a 'working' tool can expose a new systems problem once it meets real use.