The market is the floor, not the ceiling.
Every fantasy site sells you an expert. We start from the only source with real money on the line, turn it into a range of outcomes for every player, and only add our own signals when they prove they beat the market. Then we grade every projection in public, misses included. That is the whole pitch. The rest of this page is the evidence.
The problem with how fantasy advice gets made
Traditional fantasy tools run on three broken habits.
Opinions with no skin in the game
Expert rankings are guesses, updated when the analyst gets around to it, shaped by groupthink and last week's highlights. Nobody loses anything when they are wrong.
Single-number projections
“He will score 14.2 points” is a fiction. No one knows that. Football outcomes are wide, and a fake-precise number hides the one thing you actually need: how wide.
Backtested hype
Any site can tune a model until it looks brilliant on last season's data, then market the result as accuracy. You have no way to check it, and they know that.
Four pillars, in order
Each one is a decision about how a fantasy number should be built. Together they are why a Nova number looks nothing like the rank on a printable cheat sheet.
Start from the sharpest signal in sports
Sportsbooks pay a real price for being wrong, so their player markets absorb injury news, depth-chart changes, and weather within minutes. That makes the sharp betting market the best public estimate of what happens on a field. We start there, not from opinion, and rebuild each player from it. When the market moves, the board moves.
Every player is a range, not a number
We model every player as a distribution: a floor, a median, and a ceiling, plus the odds of a boom, a bust, and a tier-1 finish. Draft strategy is risk management: buy floors early, buy ceilings late. A single number can't tell you which is which. A DNA grade adds the last piece: how steady or how swingy a player has been, week to week, across multiple seasons.
Rank by value, not points
A QB who scores 22 usually loses to an RB who scores 15, because the wire is full of 18-point QBs and empty of usable backs. Nova Rank orders players by value over replacement, with scarcity priced into every rank, tier, and auction dollar. Tiers aren't decorative lines every ten players; they mark the real value cliffs, so you know when a position is about to collapse.
The Divergence Score, where the edge lives
If we always agreed with the market, you would not need us. The Divergence Score is our signature metric: an uncertainty-adjusted measure of how far our projection sits from the Vegas market on any player. Positive means we're higher, negative means lower. Those gaps are where sleepers, trade targets, and fades come from, measured instead of guessed.
Beat the market, or get deleted
Here is the discipline that governs everything we build. Every proprietary signal we develop must beat a market-only baseline on data it has never seen. Not on a backtest we tuned. On blind, out-of-sample data. If a signal clears that bar, it ships. If it does not, it is cut, no matter how clever the idea was. Most ideas do not clear the bar. That is the point: the market is hard to beat, and pretending otherwise is how fantasy sites lie to you.
Graded in public, misses included
Talk is cheap and backtests are cheaper. Every projection is locked before kickoff and graded against what actually happened, with proper scoring rules, out in the open. This is a distribution product, so we grade the distribution.
Scores the entire predicted range, not just a point guess. It grades the whole shape of the distribution, which is the product.
How often reality lands inside our floor-to-ceiling band, measured against an 80 percent target. The most intuitive calibration stat there is.
Every boom and bust probability is scored against dumb guessing. Skill above zero means the probabilities actually beat the base rate.
Start/sit win probabilities are binned against the realized record, so “72 percent” has to mean 72 percent over the season.
One thing we will never show you: a historical accuracy claim. Reliable historical prop data is effectively unobtainable, so any site flashing a multi-year hit rate built on it deserves your skepticism. Our live record starts Week 1 of the 2026 season and fills in from there, in public, whether it flatters us or not.
Sign in and log Agree, Override, or Skip against Nova on any call. We keep score all season.
No other fantasy platform invites you to outdraft its own engine and keeps the receipts. We do, because the same public grading that holds us accountable makes your wins provable too. Challenge the Model →
What this looks like on draft day
It all converges in the Draft War Room, live-synced to Sleeper. As picks come off the board it recalculates value in real time, shows tier cliffs before they collapse, weighs your roster needs and your opponents' needs, estimates who survives to your next pick, and flips between safe-floor and high-ceiling recommendations on a toggle. Auction leagues get live dollar values from the same engine.
After the draft, the same engine runs your season: weekly ranges, waiver targets separated from one-week noise, trade values built on live value over replacement instead of a static chart, and a lineup optimizer that picks for win probability, not just projected points.
The season is a point pool
Most tools treat the year as seventeen isolated weeks. We treat it as one system. Every player starts with a preseason point budget - their projected full-season total, and each week draws that pool down by what they actually score.
We measure that drawdown against the pace a player was supposed to hold (budget over their active weeks, not a shrinking week-to-week requirement that would punish anyone improving). A player who has banked 30% of their season budget by Week 3 is either a sell-high or a fluke, and we flag both. Someone running cold is a bounce-back candidate if the role still holds, never “due for a boom” because the market thinks they’re owed one. That pace signal flows straight into trades, waivers, and start/sit as a regression weight. See the pace board →
The board is live now.
The market is the floor of what is knowable, and we are the only fantasy platform that treats it that way. Every player is a range, every rank is value-based, and every signal had to beat the market blind before you ever saw it. $29 Founders rate while the 2026 record fills in, locked for life.
Frequently asked
What makes NovaPredict different from expert rankings?
Expert rankings are human opinions updated on human schedules. NovaPredict builds projections from live sharp betting markets, models every player as a range of outcomes, ranks by value over replacement, and grades every locked projection in public.
Does NovaPredict have a proven accuracy record?
Not yet, by design. The live record starts Week 1 of the 2026 season, with every projection locked before kickoff and graded publicly, including misses. NovaPredict does not publish backtested accuracy claims because historical prop data cannot support honest ones.
What is the Divergence Score?
An uncertainty-adjusted measure of how far NovaPredict's projection sits from the Vegas market on a player. Positive means Nova is higher than the market, negative means lower. Large gaps mark potential sleepers, fades, and trade targets.
Is NovaPredict free?
NovaPredict is currently in a free testing build for the 2026 season, including the rankings board, Draft War Room with Sleeper sync, Divergence Watch, and the public accountability record.
New here? Start with why the old way loses, and how to judge any projection →