Fantasy Football Projections Should Be Ranges, Not Single Numbers

Most fantasy football projections hand you one number per player and call it a forecast. That number is a guess about the middle of a wide distribution, and it hides the information that actually decides your picks: how bad the bad weeks get, how big the big weeks get, and how likely each one is. This post makes the case that a player is a distribution, not a point, and shows how to draft with ranges instead.

The problem with single-number fantasy football projections

Say two wide receivers both project for 14 points this week. A point projection says they are the same pick. They are not.

Receiver A is a possession target with a steady role. His realistic range runs from 9 to 19 points. Receiver B is a deep threat. His range runs from 3 to 28. Same median, completely different players. If you are a heavy favorite, you want A and his floor. If you are a big underdog, you need B and his ceiling, because a safe 14 probably loses anyway.

The single number erases that entire decision. And it gets worse over a season. Point projections cannot tell you:

• Which late-round pick has a real shot at a top-tier finish, and which one is just a safe bench body.
• Which "consistent" veteran is actually a coin flip week to week.
• Whether a breakout candidate's upside justifies his terrifying floor.

Every one of those questions is a question about the shape of a distribution. None of them can be answered with one number.

A player is a distribution: floor, median, ceiling, boom, bust

NovaPredict projects every player as a full range of outcomes, summarized in probability bands:

• Floor: the 10th percentile outcome. Roughly the bad week, or the disappointing season, that happens about 1 time in 10.
• Median: the middle of the range. Half of outcomes land above it, half below.
• Ceiling: the 90th percentile outcome. The great result that happens about 1 time in 10.
• Boom% and bust%: the odds of a spike week or a dud, stated as probabilities instead of vibes.
• Tier-1 odds: the chance a player finishes as a true difference-maker at his position.

On top of the bands sits a DNA grade, a week-to-week consistency grade built from multiple seasons of game logs, from steadiest to swingiest. Two players with identical season totals can sit at opposite ends of that scale, and they win leagues in very different ways. This range-first approach is also the foundation of what AI fantasy football actually is rather than a chatbot guessing at names.

Where do the ranges come from? The starting point is the Vegas betting market, the sharpest public forecast in sports, rebuilt into a full range of outcomes for each player. Proprietary adjustments only ship when they beat a market-only baseline on data they have never seen. The full discipline is laid out on the methodology page, and the case for the market as a baseline is in why player props are the sharpest starting point in fantasy.

How ranges change your actual decisions

In the draft. Your first few picks carry your season, so weight floors and consistency. Your last six picks are lottery tickets, so weight ceilings and tier-1 odds. A late pick with a 15 percent shot at a top-12 finish is worth more than one with a slightly higher median and no path to mattering. This is really what drafting as a process, not a hunch, looks like. This is the core of finding busts before they bust: the danger sign is rarely the median, it is a fat left tail at an early-round price.

In weekly lineups. Favorites should start floors. Underdogs should start ceilings. A point projection cannot even express this strategy. A range makes it obvious.

In trades. A player coming off three spike weeks looks great by recent points. His range tells you whether those weeks were his ceiling talking or his median moving.

What "most accurate fantasy football projections" should even mean

Here is the uncomfortable part for the industry: accuracy claims about point projections are mostly noise, and back-tested accuracy claims are worse. Anyone can tune a model until it looks brilliant on last season.

Grading a range is harder and more honest, and it is checkable. NovaPredict locks every projection before kickoff and grades it in public with scores built for distributions:

• CRPS, which grades the entire predicted range against what actually happened, not just the middle.
• Coverage, which asks how often reality lands inside the floor-to-ceiling band, measured against an 80 percent target. Miss high or low too often and the bands are wrong, and the record will say so.
• Brier score and skill, which grade the boom and bust probabilities against what a naive guess would score.

That live record starts filling in Week 1 of the 2026 season. Until then, no accuracy percentage gets claimed, because none has been earned in public. What can be claimed is the discipline: nothing ships unless it beats the market out of sample, and everything gets graded, misses included.

So when you evaluate any projection source, ask one question: if their number is wrong, how would anyone ever know? If there is no locked, graded record, the answer is nobody would, and that is the answer. We go deeper on that standard in how to actually judge projection accuracy, and on the limits of the numbers themselves in what analytics can and cannot do.

FAQ

What is a floor and ceiling in fantasy football? A floor is a player's 10th percentile outcome, the bad result that happens about 1 time in 10. A ceiling is the 90th percentile outcome, the great result that happens about 1 time in 10. Together with the median, they describe a player's realistic range instead of a single guess.

What does boom or bust mean in fantasy football? A boom is a spike week well above a player's typical output, and a bust is a dud well below it. Boom% and bust% state the odds of each. A boom-bust player can have the same average as a steady one while being far riskier week to week.

Why are range projections better than point projections? Because fantasy decisions depend on risk, not just averages. Ranges tell you which players are safe, which have league-winning upside, and which are hiding a scary floor behind a pretty average. A single number cannot distinguish any of those.

How do you measure the accuracy of a range projection? With distribution-aware scores like CRPS, and with coverage, which checks how often actual results land inside the projected floor-to-ceiling band. Both require projections locked before games and graded afterward in public.

See every player as a range

Stop drafting off single numbers. Open the 2026 rankings and see every player's floor, median, ceiling, boom and bust odds, and consistency grade on one board.