How many NFL DFS lineups should I enter in a Sunday GPP?

For a big-field GPP like DraftKings' Millionaire Maker (hundreds of thousands of entries), one to five lineups is fine if you're bankroll-constrained — the odds of any single lineup taking down the top prize are essentially independent, so 20 correlated lineups isn't ten times as many shots as 2. Serious multi-entry portfolios (20–150 lineups) reward correlated construction: each lineup should win in a distinct game script, or you're paying entry fees to hedge against yourself.

Should I always stack a QB with his own receiver in NFL DFS?

In GPPs, yes almost without exception — a quarterback's fantasy production is a direct function of his passing volume, which is his receivers' catches. Naked-QB builds win the same universes as QB-WR stacks but pay 30–50% less because they miss the joint upside. Cash games are the one exception: if your goal is a high floor, an anti-correlation build (a stud QB paired with a running back on a run-heavy opponent) can be defensible.

What is a bring-back in NFL DFS and when does it matter?

A bring-back is a player from the opposing team of your stacked game — usually a receiver — added to your lineup to capture the shootout script. If your QB-WR stack wins because their game turned into a 34-30 track meet, a receiver on the OTHER team almost certainly went off in that same universe. Bring-backs typically add 10–15 points of ceiling to a stack in shootout scripts and cost you a small amount of floor. Essential in tournaments; skip in cash.

How does salary cap efficiency actually work in NFL DFS?

Salary cap efficiency is measured as fantasy points per $1,000 of salary — the standard threshold for a cash-viable player is around 3.0 (a $6,000 receiver needs to hit 18 points to "clear value"). But cap efficiency alone will fill your roster with punt plays and cap your ceiling. The correct usage is a floor constraint: EVERY player should clear ~2.5x floor, and you spend your remaining flexibility on high-ceiling stars regardless of their $/pt efficiency.

Are DraftKings NFL and FanDuel NFL the same game strategically?

No — the same slate produces meaningfully different optimal lineups. DraftKings uses full-PPR scoring, which inflates every receiver's expected value versus FanDuel's half-PPR. A 6-catch, 60-yard, no-TD receiver scores 12 on DK and 9 on FD — a 33% ceiling difference for the same game. Optimal DK builds lean receiver-heavy; optimal FD builds lean toward TD-dependent stars and rushing-focused game scripts.

NFL DFS Strategy: How Sharp Players Actually Build Sunday Lineups

By DFS Simulator TeamPublished July 27, 202612 min read

NFL DFS is decided by three variables the average entrant under-weights: game-script correlation (stack your QB with his receivers), portfolio diversity (your 20 lineups should win in 20 different universes, not one), and salary efficiency measured against a floor threshold rather than maximized in isolation. Everything else — projections, ownership, exposure caps — is machinery in service of those three.

NFL is the highest-volume DFS sport by a wide margin — one main slate per week, nine-figure prize pools on the biggest Sundays, and the sharpest average opponent pool. It's also the sport where correlation matters most: touchdowns are shared events, passing yards create receiving yards, and blowouts crater everyone's ceiling on both sides of the losing team. This guide walks through what actually decides tournaments — not the projections themselves, but the construction rules that convert projections into portfolios with real EV.

The stacking-correlation problem, made concrete

Take a starting quarterback projected for 22 fantasy points and his top receiver projected for 16. If you treat those two projections as independent, the joint expected value of rostering both is 38 — the sum. But their outcomes are not independent. A four-touchdown quarterback game is (with high probability) a two-touchdown receiver game. A 320-yard passing game is (with high probability) a 100-yard receiver game. The joint upside distribution is far heavier-tailed than the sum of two independent distributions with the same means.

The math: if QB and WR have a joint correlation coefficient of ~0.55 (typical for a top-target WR), the 95th-percentile combined outcome is roughly 15% higher than the same percentile of an independent-outcomes model. That 15% is the difference between a lineup that cashes and a lineup that wins.

This is why correlated simulation is not an optimization — it's a correction against the default assumption that fantasy scores are independent. Every naked-QB build in a large-field GPP is quietly paying the correlation tax.

Portfolio construction — the 20-lineup problem

Multi-entry NFL tournaments reward what portfolio theorists call "spanned wins": each of your 20 lineups should take down first place in a DIFFERENT game-script universe. If 18 of your lineups all need the Chiefs to score 35+, you own one lineup's worth of variance in 18 spots — you're paying $2 × 18 entry fees for a $2 × 1 payoff.

Pick 4–6 stacks that win in different scripts (high-total shootouts, low-total defensive slugfests, run-heavy garbage time, road blowouts).
For each stack, build 3–5 lineup variants — different bring-backs, different flex plays, different tight ends.
Cap exposure per player at ~35% (never let one injury shred more than a third of your portfolio).
Cap exposure per game-stack at ~20% (never let one script collapse more than a fifth of your portfolio).

DFS Simulator's exposure caps enforce these limits at build time so a hand-tuning pass at 1 PM Sunday doesn't drift the portfolio into concentrated risk. The correlated Monte Carlo output ranks the resulting portfolio by spanned expected value — the metric that matters, not raw sum-of-EV.

Bring-back logic: the shootout hedge that isn't a hedge

Traditional finance calls a bring-back a hedge, but in DFS it's actually a compounding bet. If your QB-WR stack needs a 30-point game to win a GPP, that game is almost certainly reaching 60+ combined points — which means a receiver on the OTHER team is almost certainly having a big game too. A bring-back captures that correlated upside; the alternative (an uncorrelated player at the same salary) hedges against your stack failing, but at the cost of the exact universe where your stack wins.

In GPPs, always bring back. In cash games, never bring back — you want players whose floors are independent of your other rosterable positions, not correlated.

Cap efficiency as a floor constraint, not a maximum

The common mistake is to sort by points-per-dollar and fill the roster top-down. That gets you a lineup full of $4,500 punt plays projected for 12 points — mathematically "efficient" at 2.7 pts/$1k, and mathematically incapable of winning a GPP. The winning frame:

Set a floor: every player must clear ~2.5 pts/$1k in floor projection (about 22 points for a $9,000 running back, 15 for a $6,000 receiver, etc.).
Above the floor, sort by CEILING — the 90th percentile of simulated outcomes, not the mean.
Spend your salary on the highest-ceiling players who clear the floor. Punt plays get one FLEX slot, not five.

This is why the DFS Simulator run-summary shows both floor and ceiling per lineup — the ceiling column is the tournament answer; the floor column is the cash-game answer. Same simulation, different sort.

Cash vs. GPP: the split that most players get wrong

Cash games (50/50s, double-ups) pay the top ~44% of entries. High floor wins; ceiling is irrelevant because the payout is flat. GPP (tournament) pays the top ~20%, but the money is concentrated in the top 0.1% — you need ceiling, and you need to differentiate from the field on ownership.

A common trap: playing the same lineup in both. Your cash build should be your highest-floor, lowest-variance construction. Your GPP build should be your highest-ceiling, highest-variance construction. They should share maybe 3 players out of 9. If your cash and GPP lineups are 80% the same, one of the two is misconstructed.

The workflow, top to bottom

Load the NFL slate on DraftKings or FanDuel.
Adjust projections you disagree with (news breaks the consensus projections all week — the sim edit column is designed for last-second fixes).
Set exposure caps: 35% per player, 20% per game stack.
Configure 4–6 stacking rules: which QB-WR pairs, which bring-back positions, which run-game onslaught builds.
Run 10,000 iterations (Pro) or 50,000 (Elite). The correlated model prices your stacks joint-distribution, not mean-sum.
Sort the output by CEILING for tournaments, by FLOOR for cash. Enter accordingly.
Export site-formatted CSV. Upload. Watch. Adjust ownership reads next week from the leaderboard.

What separates winning players over a season

Talent, patience, and bankroll discipline. The math of GPPs guarantees that any individual week's outcome is a lottery — even a perfectly constructed portfolio has <20% ROI variance per week. Winning players survive the down weeks; losing players tilt, chase, and over-lever after a cold stretch.

NFL DFS Strategy: How Sharp Players Actually Build Sunday Lineups

Frequently asked questions

Keep reading

Put the theory into practice