How to Use a DFS Optimizer (Without Letting It Think for You)
An optimizer is a calculator, not an oracle: it faithfully maximizes whatever numbers you feed it. Using one well is 20% clicking Optimize and 80% feeding it projections worth maximizing and interpreting the output correctly. This guide is the full workflow as used on this site's own optimizer — every step works identically on any optimizer you'll ever touch.
Quick answer: The workflow: (1) get salaries and build projections you actually believe in, (2) paste them in, (3) exclude players you won't roster for news reasons, (4) lock your correlation plays (stacks), (5) optimize, (6) read the alternatives as a menu of defensible builds — not as four copies of the truth. The optimizer guarantees the math; you own the inputs.
Step 1 — Bring your own projections (this is 80% of the job)
The optimizer's guarantee — 'highest total projection under the cap' — is only as good as the projections. Free tools that include projections are including someone's opinion; paid tools are including someone's better-funded opinion. Either way the honest workflow is the same: export salaries from DraftKings or FanDuel, attach projections from whatever source you trust (your own model, a screening process, a paid provider), and paste the result. On this site, the CSV box takes Name, Position, Team, Salary, Projection and the whole table stays editable inline. If a projection looks wrong in the table, fix it there — the solver re-runs instantly.
Step 2 — Exclusions are your news layer
Optimizers don't read injury reports (and ones that claim to still miss late scratches). The human step before optimizing is the exclusion pass: any player you would not roster for real-world reasons — injury risk, benched, weather, you just don't trust him — gets excluded so the math can't hand him to you. A good habit: run the optimizer once with no exclusions, see who the math loves, then research those specific five or six names instead of the whole 140-player slate.
Step 3 — Locks are for correlation, not confidence
Locking a player forces him into every lineup. The legitimate use is correlation: you want your QB with his WR1, so you lock both and let the optimizer build the best legal seven around them. The illegitimate use is 'I love this guy' — that's how people pay a premium price and then discover the optimizer would have found a better fit for the same money. As a rule: lock pairs, not singles.
Step 4 — Read the alternatives like a menu
On our sample slate, the optimal lineup projects 114.8 and the four alternatives land within half a point of it. That's the normal shape: near the optimum, many lineups are nearly equal. So don't ask 'which is THE lineup?' — ask 'which of these five nearly-equal builds has the shape I want today?' One alternative might concentrate salary in two studs (higher ceiling, lower floor); another might spread it evenly (the cash shape). The optimizer has done the arithmetic; the portfolio choice is yours.
The four mistakes that waste optimizer output
- Trusting the projections more than you should — the optimizer's precision is arithmetic, not prophecy; a 114.8 'optimal' lineup is still a distribution, not a promise (the Range of Outcomes tool shows how wide).
- Optimizing once at 11:55am with stale news — late scratches turn optimal lineups into 8-man rosters; re-run after final inactive lists.
- Locking too much — every lock shrinks the solver's options; lock a stack pair and you've spent your lock budget wisely, lock five players and you're hand-building with extra steps.
- Ignoring ownership — in GPPs, the optimal-with-default-projections lineup is by definition the one the field's optimizers also found. The alternatives menu is where low-ownership leverage lives.
A worked example on the sample slate
Load the optimizer with the sample slate (146 players) and press Optimize: you get the exact optimum — 114.8 points at $50,000 of the $50,000 cap. Now lock the QB and his best receiver and re-run: the score drops (locks always cost projection — that's the price of correlation), but the lineup's shape changes toward the GPP build you wanted. The delta between unlocked and locked scores is the actual price of your stack, stated in points — most people never price it, and are surprised how expensive 'I like both of them' really is.
What this site doesn't do
- This site's optimizer never auto-updates projections or salaries — the default table is a labeled sample, and your pasted numbers are the real inputs.
- It will not tell you who to roster — lock/exclude decisions are yours; the solver only does the arithmetic honestly.
- It is not betting advice, and nothing here promises profit. DFS contests can lose money as easily as win it.
Frequently Asked Questions
Do I need an optimizer to win at DFS?
How many lineups should I generate?
What's the difference between optimizing for cash and GPP?
Can an optimizer account for injuries and news?
Why did my 'optimal' lineup score badly?
More tools used in this guide
Build the highest-projected legal lineup under DraftKings ($50,000) or FanDuel ($60,000) rules — exact solver, lock/exclude players, CSV import, no signup.
Monte-Carlo simulate any 9-man lineup: floor (P5), median, ceiling (P95) and the odds of clearing your target score — deterministic and instant.
Turn any salary + projection into points per $1,000, the points needed at your target multiplier, and a surplus/deficit verdict.
More guides
The actual algorithm inside every DFS optimizer: knapsack constraints, why greedy fails, how FLEX enumeration works, and what 'optimal' legally promises.
The honest beginner's path: cash vs GPP, the value multiplier, why 50/50s teach more than milly makers, and the bankroll rules nobody enjoys reading.