How BetFantasy.ai works

What does BetFantasy.ai actually compute?

BetFantasy.ai turns your league settings into one ranked draft board. The pipeline runs in four steps: it projects season points for every draftable player under your scoring rules, measures each player against a replacement-level baseline at their position, folds those gaps into a single cross-position ranking, and estimates how likely each player is to still be available at your upcoming picks. The result is a table you can sort and filter, not a black box: each column below has its own section explaining exactly what it means.

What is value-based drafting?

Value-based drafting is the idea that a player's draft value is not their raw projected points but their points above what a replaceable player at the same position would score. A quarterback projected for 320 points sounds better than a running back projected for 240 — but if replacement-level quarterbacks score 280 and replacement-level running backs score 150, the running back is worth +90 against his baseline while the quarterback is worth +40. Positions differ in depth, so raw points mislead; gaps against a positional baseline compare fairly across positions. BetFantasy.ai's baselines derive from historical value cliffs at each position, scaled to your league's size.

What is Impact Delta?

Impact Delta is that gap, computed for your league: the player's projected points under your scoring settings measured against replacement-level baselines for their position — a starter-level and a bench-level baseline, averaged, and never reported below zero. It is the board's core value number. The bigger the delta, the further the pick moves your roster ahead of what a waiver-wire fill-in would give you at the same slot; a delta near zero means replacement level is close enough that the pick buys you little.

What is Prime Ranking?

Prime Ranking is the board's single cross-position ordering. It applies the same points-above-baseline idea with its own positional baselines — one fixed, and one that tightens as players at the position come off the board — and blends in draft-context factors: positional scarcity as the draft progresses, and roster construction under your league's starting lineup. The top of the list answers "who should I take right now?" rather than "who scores the most points?", and like the board's other formula columns it is deterministic: the same inputs produce the same ranking every time. The one column that works differently — the model-generated AI Rating — is covered in its own section below.

What is Avail%?

Avail% is the probability that a player is still undrafted when one of your upcoming picks arrives. It is computed from where the player typically goes in real drafts — average draft position across multiple platforms — measured against how many picks away your next turns are in your league's draft order, including snake-order reversals. A player at 95% will almost certainly be there next round; a player at 20% is a now-or-never decision. That is the question Avail% exists to answer: whether you can afford to wait.

What is the AI Rating?

The AI Rating is the board's one model-generated column, and it is built in two layers. The first layer is a documented, deterministic signal core: plain math over the board's own fields — projections, ADP consensus and disagreement, sportsbook season lines, age, rookie status, and injury designation. The second layer is a large language model that reads that signal sheet and applies a bounded adjustment, with a short note saying what it weighed. Each number that appears in a note comes from the board's own data; the model is not allowed to introduce outside figures.

A rating is labeled with the model that generated it and shows its generation date, so you can judge its freshness yourself. A player without a rating shows a dash rather than an invented number.

Where do the numbers come from?

Projections come from the BetFantasy Edge Model's imported stat sheet. Rows carrying a real imported projection show an EDGE badge; players without one get an estimate interpolated from their draft position instead, marked EST. EDGE rows are the stronger signal — EST rows should be read as ordering, not as forecasts.

The ADP consensus is built from five named sources: ESPN, Sleeper, Fantasy Football Calculator, MyFantasyLeague, and Yahoo — the last collected from FFToday's published draft-position table (rank order, refreshed with the rest of the pool) rather than from Yahoo directly.

There is no single consensus number that all five feed, because an average may only ever combine values from one scoring format, and no source publishes all three. ESPN publishes one Full PPR draft population. MyFantasyLeague publishes Standard and Full PPR. Yahoo, through FFToday, is published in Half-PPR only. Sleeper and Fantasy Football Calculator publish all three. So the board's three averages are three different blends: Standard averages Sleeper, Fantasy Football Calculator and MyFantasyLeague; Full PPR averages those three plus ESPN; and Half PPR averages Sleeper, Fantasy Football Calculator and Yahoo. Each per-source column shows only what that source actually published, and says so when it is showing you a format other than the one selected.

Season win totals from sportsbook consensus can also appear as reference context — they are not official statistics or projections, and BetFantasy.ai takes no wagers. Each source refreshes on its own schedule — none of them live; the per-source disclosure lives in the data accuracy section of the Terms, which is kept accurate against the running pipeline.

Does the model learn or update itself?

No. The BetFantasy Edge Model is a fixed formula set: it does not train on results, does not recalibrate during the season, and nothing you do in the app changes how any player is scored. When a number on the board moves, it is because underlying data moved — an ADP feed refreshed, an injury status changed, or a new stat sheet was imported. That is a deliberate design choice: a fixed model can be documented, tested, and checked, and its mistakes are inspectable rather than mysterious.

The AI Rating column is the one exception to the formula rule, and it does not learn either. Its adjustment and note come from a large language model — the model's name is shown with the rating — reading the board's own fields. Each rating is dated, and it is regenerated when a new run is published rather than retrained: the model keeps no memory of its past outputs and does not learn from results.

Is it really free?

Yes. The full board works without an account, everyone sees the same numbers, and there is no paid tier. BetFantasy.ai is a draft tool, not a sportsbook — it accepts no wagers and holds no money. Open the board and see for yourself, or read what the product deliberately does not do.