FAQ & Glossary

How the numbers are made, what the words mean, and what we will not do. Last updated 18 September 2026.

Short version: the Herd is a statistical model that produces every projection. The Seer is an AI analyst that ranks and explains using those numbers and never invents one. We publish our misses.

1. Scoring formats

Projections are produced per scoring preset, and the one in use is always the one your league is set to.

2. Projection, floor and ceiling

Projection
The expected points for that player in that week, race, tournament or card, under your league's scoring.
Floor
A bad-but-plausible outcome, not a worst case. For NFL it is the model's ~15th-percentile prediction.
Ceiling
A good outcome, not a best case. For NFL it is the model's ~90th-percentile prediction.
Rest of season
The sum of the remaining games or events, projected the same way.

Floor and ceiling are produced by separate quantile models rather than by adding a margin to the projection, so the gap between them is itself information: a wide gap is a volatile player, a narrow one is a steady player. Other sports derive them differently where the shape of the sport demands it — golf from the spread of finishing positions, hockey scaled by the chance a goalie actually starts — so compare floor and ceiling within a sport, not across sports.

3. Why a projection looks low

Three things routinely pull a number below where you expect it, and all three are deliberate.

4. How we compare to ESPN and other sites

We load ESPN's projections and show them next to ours wherever both exist, so you can see the disagreement instead of taking our word for it. Those third-party numbers are display only: they never feed our model, our rankings or the Seer's advice. If you want to know who has been closer, do not take either side's marketing — read the accuracy page, which reports mean absolute error, RMSE and rank correlation for our projections, both backtested on a full past season and week by week for the current one, among projected starters. We publish the weeks we did badly too.

5. When numbers update

Every projection carries the time it was produced, so if a number looks stale you can check rather than guess.

6. The Herd and the Seer

The Herd is the projection model — gradient-boosted models trained on our own history, currently served as a blend of two members whose weights were chosen by measuring them against held-out seasons. The Seer is the AI analyst. It reads the Herd's numbers, the usage and matchup data behind them, injury status and recent news, and then ranks, chooses and explains. It cannot invent a projection or a player: the candidate set is computed before the model is asked anything, and its answer is validated against that set. If the Seer tells you something we do not have a number for, it is designed to say so rather than guess.

7. What we will not do

No betting. We do not publish odds, props, or wagering advice, and we do not take entry fees. Leagues never hold money. Prize contests are free to enter with an alternate method of entry — the official rules cover those.

8. Still stuck

Ask the Seer directly — questions about scoring, floor and ceiling, or when things update are answered from this page. For anything else, support@fantasyxai.ai.