DFS Projections Explained: How to Use Them to Build Better Lineups

A DFS projection estimates a player’s fantasy production under a set of assumptions. Check the platform, slate, update time and meaning of the number before using it. Expected points, median outcomes and ceiling estimates describe different things, and none is a promise of the final score.

For DraftKings NFL, projections help compare opportunity and price across a large player pool. They become more useful when you understand the uncertainty around roles and how one selection changes the rest of a legal lineup.

Mean and median are not interchangeable

A mean projection is a probability-weighted average across possible outcomes. A median marks the middle of a distribution: at least half its probability is at or below that value and at least half is at or above it. Large occasional scores can pull the mean above the median.

Projection providers do not all use the same definition. Read the source’s description instead of assuming a generic “projection” column is a median. A player’s most likely individual score is another concept and can differ from both.

Worked example: the same mean can hide different outcomes

Consider two deliberately simplified players. Their scores are hypothetical and are not intended to describe a realistic full NFL distribution.

Player A and Player B both average 15 points

Player	Possible outcomes	Mean calculation
A	14 points 50% of the time; 16 points 50% of the time	(0.50 × 14) + (0.50 × 16) = 15
B	5 points 60% of the time; 30 points 40% of the time	(0.60 × 5) + (0.40 × 30) = 15

Player B’s median is five points, despite his 15-point mean. If an isolated decision required more than 20 points, B could achieve it in this toy example and A could not. If it required more than 10, A would always clear it while B would clear it only 40% of the time.

A real lineup has several players, uncertain cutoffs, salary constraints and opponents. The example shows why a single average cannot determine the better selection. It does not establish that volatile players are always better in tournaments or stable players are always better in cash games.

Ask what “floor” and “ceiling” actually mean

A source might define ceiling as a particular upper percentile, while another might use an informal upside estimate. Those numbers are not directly comparable unless the definitions match. Floor likewise should not be interpreted as the lowest score a player can possibly produce.

If a source defines a ceiling as the 90th percentile, roughly 10% of its modeled outcomes are above that threshold. That is different from a maximum or a guarantee that the player reaches the number regularly. Injuries, role changes and model error can also create outcomes the estimate handles poorly.

When no definitions are available, treat floor and ceiling as qualitative shorthand. Check the role that could produce those outcomes rather than trusting the labels.

Start with the assumptions about opportunity

For an NFL running back, carries, targets and valuable touches near the goal line influence the opportunity. For a receiver or tight end, routes, targets and the type of targets matter. Quarterbacks can reach fantasy points through passing and rushing, so similar passing expectations need not imply similar fantasy profiles.

Distinguish a player being active from a player having a secure workload. A returning starter might play limited snaps. A reserve promoted after an injury may share work. A projection built on one role can become unreliable when the team uses another.

Matchup and game environment add context, but a high game total is not a fantasy-point allocation. It does not tell you which player scores the touchdowns or how much work a particular receiver receives.

Scoring rules can change the expected value

Use DraftKings-specific projections for the correct contest format. Receptions, passing-touchdown values, yardage bonuses and deductions all affect the translation from football statistics to fantasy points. A generic fantasy ranking is not necessarily adjusted to the same rules or salary cap.

Bonuses also illustrate why a projected stat line is not an exact future box score. If a hypothetical receiver has a 50% chance of reaching the three-point 100-yard bonus, the expected bonus contribution is 0.50 × 3 = 1.5 points. He will actually earn either zero or three bonus points, not 1.5.

Likewise, an expected receiving-yard total below 100 does not imply a zero chance of the bonus. Different outcome distributions can have the same average yardage and different chances of crossing that threshold.

Compare complete salary trades

Points per $1,000 of salary can help you spot inexpensive production. It does not solve a lineup. Positional eligibility, available alternatives and the marginal value of spending more all matter.

Worked example: move salary between two positions

Suppose you have $12,000 left for one running back and one receiver. All projections below are hypothetical means for the same slate.

Both pairs cost $12,000

Pair	Running back	Receiver	Total projected points
A	$7,500; 22 points	$4,500; 13 points	35
B	$6,000; 19 points	$6,000; 18 points	37

Moving from A to B gives up three projected points at running back and gains five at receiver: a net gain of two. The highest-projected individual running back does not produce the highest-projected pair.

That establishes only the mean-point comparison. Workload uncertainty, joint upside, ownership and the contest’s payout can still affect which complete lineup you prefer.

Adding projections does not describe the lineup’s range

Expected points can be added even when players are correlated. If a quarterback has a mean of 22 and a receiver has a mean of 18, the pair’s mean is 40. The relationship between them changes the distribution around that mean, not the arithmetic of adding their expectations.

Medians and upper percentiles do not generally add in the same way. Adding nine individual “ceilings” does not give you a reliable lineup ceiling, because those outcomes may not occur together. A quarterback and receiver can share scoring events, while other combinations may depend on competing game scripts.

This matters in tournaments, where the chance of an extreme joint result can be valuable. It also matters in flat-payout contests, where concentrating too much uncertainty may affect the chance of clearing a cutoff.

When projection sources disagree

First check whether the numbers use the same scoring, slate and update time. Then look for differences in role assumptions, injury treatment or expected opportunity. An old projection and an updated one are not two equally current opinions.

Agreement does not prove low uncertainty. Several providers can share data, news or similar assumptions and all miss the same workload change. Disagreement can point to a question worth investigating, but simply averaging sources does not automatically correct it.

Be equally careful with small gaps inside one source. A difference between 18.7 and 18.4 points may be smaller than the uncertainty in the workload assumptions. Treat it as an estimate to weigh alongside salary and fit, rather than a definitive ranking.

Use a short pre-lock check

1. Confirm the projections match DraftKings and the exact slate.
2. Check the update time and whether important news is included.
3. Identify uncertain roles that materially affect your lineup.
4. Compare alternatives as complete salary and position changes.
5. Review the lineup’s joint outcomes and likely popularity for its contest.
6. Recheck player status and the contest’s available swap options.

DraftKings’ late-swap guidance explains when eligible unlocked selections can be changed. A projection update does not reopen a locked player.

After the slate, compare the assumed role with what happened and separate news changes from ordinary outcome variation. A single projection miss is expected. Repeated misses in similar situations may reveal an assumption worth revisiting.

What to read next

• NFL lineup construction applies projections to a complete nine-slot roster.
• Ownership and leverage adds the competing field to the comparison.
• Cash games vs. GPPs explains why the payout changes how you use the numbers.