How It Works — AIOmni

Step 01

Connect your league

Link Sleeper, ESPN, Yahoo, MyFantasyLeague, or Fleaflicker via official OAuth (or a session cookie where the platform doesn't expose OAuth). AIOmni reads scoring rules, roster construction, waiver type, draft order, trade deadlines, and playoff structure on the first sync — without you typing a single setting.

Step 02

AI internalizes your settings

Your league context is appended to every model prompt: 4pt vs 6pt passing TDs, full / half / no PPR, TE Premium multipliers, SuperFlex slots, FAAB vs rolling waivers, bonus thresholds, IDP rules. Claude Opus parses these alongside your live roster before formulating any answer.

Step 03

Get personalized advice

Start/sit, waiver targets, trade grades, draft picks — every output is specific to your league's scoring and your actual roster. No generic top-100 list. No copy-pasted advice. Two managers in the same league get different answers.

Under the hood

What gets read on every request.

A snapshot of what AIOmni ingests before generating an answer — live data plus your league context.

Your League

Scoring rules · roster construction · waiver type · draft format · trade deadline · playoff bracket · weekly lineup deadlines.

Your Roster

Active starters · bench · IR · taxi (dynasty) · recent transactions · waiver budget · projected lineup vs current.

Live Data

Injuries from official sources · snap counts · Vegas implied team totals · weather at stadium · advanced stats · college prospects.

Recency

News from Rotowire, PFR, CBS, ESPN, and beat writers — deduplicated and tagged. Anything past 48 hours is flagged stale.

Season Memory (Pro)

Your decisions, tendencies, and outcomes across the full 52-week season — every conversation gets sharper as the year progresses.

The Model

Claude Opus for reasoning (Coach, Trade, Draft). Claude Haiku for quick lookups. Sonnet reserved for future. Sources cited inline.

AIOmni Formula

A transparent, documented ranking algorithm.

Most ranking platforms ship black-box numbers. AIOmni Formula is a 4-stage algorithm anyone can read in the source code. Every output traces to a specific step.

01
Projected Points — convert ADP/consensus rank into a season point estimate via empirical decay curves with format-specific scoring coefficients.
02
Value Over Replacement — subtract positional replacement level (driven by league construction) so positional scarcity is captured.
03
Time-Value Factor — for dynasty, apply age-based depreciation curves to compute Expected Lifetime Value.
04
Tier Clustering — natural cliff detection on VOR gaps surfaces "trade up / trade down" decision points without hand-tuning tier breaks.

Reproducibility

Given the same input data, every AIOmni user sees the same base rankings. No drift, no hidden randomness, no opinionated tweaks layered on top. The math is the math.

Format accuracy — the algorithm knows your league settings and adjusts replacement level accordingly, not via hand-tuned position boosts.

# services/rankings/aiomniEngine.ts

Reads your league. Then advises.