The industry has an opinion. We have a divergence score.
NovaPredict is an AI fantasy football decision engine. It starts from the Vegas betting market, turns each player into a range of outcomes, ranks by value, scores where it splits from the market, and grades its own accuracy in public.
What NovaPredict does
The sportsbooks set a line on every player, with real money behind it. We read the whole market and translate it into a full fantasy points distribution for each player, then rank the board by value over replacement, flag the players we price differently than the market, and grade week-to-week consistency. Then we lock every projection and publish how it did. It is one connected system: price from the market, rank by value, publish the record.
What we believe
Distributions, not point guesses
Every player is a full range of outcomes, a floor, a ceiling, and the odds in between, not a single number that pretends to know the future. The range is the honest picture, and it is what our projections are built on.
Value over raw points
Points do not win drafts, value does. We rank by value over replacement, how far a player sits above what you could stream for free at the same position, so the board reflects real edge instead of raw totals.
Where we disagree with the market
The betting market is sharp, so when our numbers diverge from it we say so out loud with a Divergence score. That gap, not agreement, is where the edge lives, and we surface it rather than hide it.
Accountability over hype
We lock every projection before games and grade it in public afterward. No quiet edits, no cherry-picked wins. If the model is wrong, the record shows it.
We only keep signals that beat the market out of sample
The betting market is the baseline to beat, not a number to copy. A signal earns a place in the model only if it improves on the market on data it has never seen. Anything that looks good in hindsight but does not hold up out of sample gets cut. That discipline is the whole point.
An early-access build
NovaPredict is live in early access and built for your draft right now. We make no accuracy claims until the season starts and the public record begins in Week 1. You can pressure-test the model yourself: open the Draft War Room, read where we diverge from the market, and hold us to the record we publish.