How DFS lineup optimizers work
What an optimizer actually computes, where its answers come from, and why the projections and the rules you give it matter more than the software.
A daily fantasy lineup is a small puzzle: pick a set of players who fill the roster positions, stay under the salary cap, and score as many points as possible. With a few hundred players on a slate there are more legal lineups than anyone could check by hand. A lineup optimizer is software that searches that space for you.
The mathematics, briefly
Under the hood, most optimizers solve an integer program. Each player is a yes-or-no choice; the roster rules and the salary cap are constraints; the objective is the sum of the chosen players' projected points. A solver finds the combination with the highest total that satisfies every constraint — exactly, not approximately. Your own instructions are simply more constraints: a lock forces a player in, an exclusion forces him out, an exposure cap limits how many lineups he can appear in.
An optimizer is only as good as its inputs
- Projections. The optimizer maximizes whatever numbers it is given. If a projection is wrong, the optimizer will confidently build around the mistake. Good tools say where their projections come from and let you replace them with your own.
- Rules. The optimizer has no opinion about football. Stacking a quarterback with his receivers, keeping a defense away from the offense it faces, and leaving out a cheap player with no role are knowledge someone has to put in — either you, as rules, or the tool, as defaults.
- The objective. Pure projection is right for a single cash lineup. For a tournament, many optimizers also weigh ceiling, ownership and correlation, because the goal there is to finish first, not to finish above average.
From one lineup to a portfolio
The naive way to build twenty lineups is to find the best lineup, forbid it, find the next best, and repeat. The result is twenty near-copies of one idea: the same core with one or two players swapped. A portfolio optimizer instead generates a large pool of strong candidate lineups and then chooses the set together, with rules about exposure and overlap applied across the whole portfolio. That is the difference between twenty entries and twenty chances.
Doing it here
- Pick a slate. Published slates come with Martian Flow projections; you can upload your own instead.
- Choose Just Pick For Me to let the optimizer decide everything you do not, or set only the rules you actually believe in — a lock, a fade, a stack.
- Build, then open Portfolio Review to see what the optimizer did: exposures, stacks, how similar the lineups are, and whether each rule was met.
An optimizer makes construction better, not predictions. It cannot know who will score; it can make sure your lineups are valid, varied and built around the reasoning you gave it.