How to Use a DFS Optimizer (Without Letting It Think for You)

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.

More tools used in this guide

More guides