Lineups must spend at least this much (cap is $50,000 on DK, $60,000 on FD -- the slider adjusts automatically when you switch sites).

Excludes any player projected below this many points from the pool entirely, before the solver runs (0 = no floor). Locked players are always exempt.

Requires the lineup's summed ownership to be at least this much (0 = no floor; same units as "Total Own" on a built lineup, e.g. 130). Set very high to see a "super chalk" build for comparison against the true optimal.

Excludes a player from the pool if his participation (share of his team's last 5 played games he actually appeared in) falls below his position's floor here -- catches a backup or a still-injured player a point projection alone can miss. ON by default. Locked players are always exempt -- lock in a player you know is a real exception (e.g. a starter confirmed back from injury). Format: POSITION:FLOOR comma-separated, 0.0-1.0. Clear the field to disable entirely.

Session 10.5's mean-variance objective: sum(mean) - lambda*sum(sigma²). 0 (default) maximizes projected points only. Positive = floor-seeking (cash); negative = upside-seeking (GPP). Requires a projection build with --sigma-recalibration -- the solve fails loudly if the pool has no per-player sigma. Built-in presets' lambda values are Session 10.5b's swept defaults (DK, 2018-2021, 65 weeks, holdout; see ROADMAP.md Ad Hoc Session A5) -- cash/SE-3Max at 0.063 (best on their own metric, though the 3-Max case was not a clear winner over 0.0), MME at -0.005 (positive lambda measurably hurts large-field GPP).

Ad Hoc Session A6: any non-DST player whose simulated 10th-percentile outcome (statline_p10) falls below the Floor Threshold is a "dart" -- capped at Cap % of this batch's lineups, unless he's on this batch's own stack team (that correlation is the trade you're paying for, not accidental dead-money risk). DST is always exempt -- every DST's floor simulates as a flat 0 regardless of matchup, so this filter can't tell a good matchup from a bad one there; use the DST's own opponent implied total for that instead. Real example that motivated this: a $3,000 min-salary WR with statline_p10 = 0.07 (basically a coin flip on scoring anything) was showing up in 20% of a real 20-lineup MME batch purely for being the cheapest legal filler. Not a backtested/swept value like Lambda above -- a heuristic guard against that specific failure mode (see ROADMAP.md Ad Hoc Session A6). Classic slates only; needs the same --sigma-recalibration build as Lambda.

By default the optimizer refuses to roster a skill player (QB/RB/WR/TE) from the same team your own DST/DEF is facing this week — the two are negatively correlated (your DST scores better when that offense struggles). Check this box to lift that restriction. Recommended for small slates or Showdowns, where the player pool is thin and avoiding the pairing may cost more than it saves.

E.g. cap KC at 2 players. Click an added cap to remove it. If a locked player already exceeds a cap, the build will fail with a clear reason.

E.g. cap KC vs DEN at 4 combined players. Independent from (and additive with) a Game Stack's minimum -- a contradictory combination will fail to build.

E.g. floor KC at 4 to force at least 4 of the roster's players from KC. The other team must still have at least 1 player -- an impossible floor fails loudly before the solver runs, not as a silent no-op.