How CourtEdge Simulates NASCAR DFS

NASCAR is the most simulation-friendly sport in DFS — and the most misunderstood. Here's exactly how our engine turns 38 cars into thousands of simulated races, and how those races become DraftKings lineups that actually win.

Most NASCAR players stare at a projection column and roster the six biggest numbers. That's how you finish 60th. NASCAR scoring is built around positions gained and laps controlled — two things a single projection number hides. Our engine doesn't project a driver to a point total; it simulates the whole race, thousands of times, and reads the point totals out the back. Here's the build.

1. Start with the scoring, because it's weird

DraftKings NASCAR scoring rewards things that have nothing to do with raw speed:

Two of those — fastest laps and laps led — are the dominator points. The other big lever, place differential, pulls in the opposite direction: you can't gain 25 spots if you start on the front row. That tension is the entire game, and a flat projection can't express it. A simulation can.

2. The core: a finishing-position distribution for every car

For each driver we build a probability distribution over where they finish, not a single number. The inputs:

• Practice & qualifying speed — the strongest near-term signal of car quality this week.
• Track-type history — a driver's record at this kind of track (superspeedway, short track, intermediate, road course), not just this exact track.
• Starting position — the anchor that place-differential upside is measured from.
• Equipment & recent form — team speed trends over the last several races.

From those we get a skewed distribution: a fast car starting P30 has a fat tail toward the front (and the wreck risk toward the back). We then run the field tens of thousands of times. Every trial is a full finishing order — which automatically produces place differential (start minus finish) for all 38 cars at once.

3. Layering in dominators

Laps led and fastest laps don't spread evenly across the field — they concentrate in a handful of cars that run up front. So on top of the finishing sim, we model a dominator probability for the contenders: given track type and starting spot, how likely is this car to control a big share of the 0.45-per-fastest-lap and 0.25-per-lap-led pool?

At tracks where the leader checks out, dominator points can be half a driver's score. Our sim allocates them by simulated laps-led share, so a pole-sitter at a dominator track carries a genuinely different scoring profile than the same driver at a chaos track.

4. Track type rewrites the whole model

This is where most tools fall down — they run one model for every race. We don't:

The same driver is a different DFS play week to week, and the engine treats him that way.

5. From simulated races to six-car lineups

Once every car has thousands of simulated scores, a DraftKings lineup ($50,000 cap, 6 drivers) is just an index sum across each trial. That lets us score thousands of candidate lineups against the same simulated races and rank them by what the contest actually pays:

• Cash: optimize floor and median — the lineup that clears the cash line most often. Usually a couple of safe dominators plus reliable place-differential.
• GPP: optimize ceiling × leverage — the lineups that win when the chaos breaks right, weighted against projected ownership so you're not just chalk.

The classic NASCAR construction — pairing one or two dominators (front-runners banking laps-led) with three or four place-differential plays (fast cars starting deep) — falls out of the math naturally, because that's the build that maximizes simulated points under the cap.

NASCAR isn't won by picking the fastest cars. It's won by owning the right mix of cars that control the race and cars that climb through it — and being slightly different from the field when they do.

The short version

We simulate the entire race tens of thousands of times, driver by driver, with the finishing distribution, dominator share, and track type all baked in. Then we read DK points out of those races and build the lineups that win cash and GPPs under the cap. No flat projections, no one-size-fits-all model.