Methodology

How EdgeIQ builds today’s picks

We publish the same model every subscriber sees — no black boxes, no “secret sauce.” This page walks through the inputs, the components, the confidence calibration, and what the free vs. paid tier actually gets.

1. Inputs per game

Every slate starts from these public inputs. Nothing is paid, nothing is private — the model is reproducible from the data we already publish.

    Elo ratings
    A per-team power rating updated after every game. Larger Elo gap → stronger model favorite.
    Recent form (last 5)
    Win/loss differential across the last five games, weighted by margin.
    Injury reports
    Each reported absence (key skill positions) lowers confidence by ~7% per side.
    Opening line
    The market’s first posted spread / total — used as the baseline for the pick.

2. Component models

Three independent sub-models vote on every pick. When two or three agree, agreement is reported alongside raw confidence — disagreement widens the variance band.

Elo + injury
Power rating adjusted for known injuries. The simplest of the three.
EPA regression
Expected-points-added regression vs. league baseline — finds teams trending beyond their record.
Market deviation
Quantifies how far the model leans from the market — high deviation picks often carry higher variance.

3. Confidence calibration

Confidence is capped at 78% — honest models rarely get more confident than that, and we don’t want anyone treating a free pick as a lock. The board also reports line movement at DraftKings, FanDuel, and BetMGM so you can see the market breathing alongside the model.

  • Confidence is clamped between 50% and 78% — no pick is ever presented as “lock.”
  • Component agreement is reported alongside raw confidence; high agreement widens the realistic edge.
  • Line movement is captured at three books across a six-hour window pre-kickoff.

4. Tier gating

Freevisitors (anonymous and signed-in non-subscribers) get a peek of the slate: one pick per sport per day, with the matchup, market, selection, and line — but no confidence score, no reasoning, no weighted key inputs. The free gate is intentionally narrow so it's hard to mistake the public rank for the model's actual conviction.

$29 Starter subscribers unlock the full daily model output across every sport: model agreement across the three component models, weighted key inputs with sources, and book-by-book line movement across DraftKings, FanDuel, and BetMGM. The daily email digest lands pre-kickoff.

$99 Pro subscribers get everything in Starter plus slate-wide statistical breakdowns (pick count by sport, average confidence, agreement distribution), per-pick player props when live data lands, priority ordering on the daily email digest, bankroll projections, and the private EdgeIQ Discord channel.

Subscriptions are managed via Stripe — checkout at /pricing, self-serve change/cancel under /account/billing. The first 200 subscribers get the $29 mention lock-in; if you cancel, your access stays paid through the end of the period.

Free tier misuse is mitigated server-side in /api/picks: gate + cap live on the server, not the client. Subscribed access is reconciled by a corner-based payment events feed (the platform's sanctioned Stripe webhook substitute), so a Stripe-side cancellation surfaces in the picks feed within about a minute.

5. Limitations

This is a methodology page, not a marketing claim. The MVP does not have live prop-grader data or in-play updates; the slate refreshes before kickoff via a recurring cron job (see polsia.toml). A confidence number is not a guarantee of outcome — anyone treating it that way is reading the site wrong.