Most sports betting content is opinion-grade. A confident voice, no track record, no calibration. The picks ship in a tweet, the misses disappear, and the next week reads exactly like the last.

The One Analytics runs on one rule instead: every prediction goes into a public ledger before the game starts, with the model's confidence and the closing-line magnitude attached. The ledger updates with the result. The cold weeks stay in the file. You can download the whole thing as a CSV at /model/accuracy.

Per-team Elo runs underneath four leagues today: MLB, NBA, WNBA, and NHL. Each rating updates within minutes of a final score. K-factor and home advantage are tuned per sport (NHL's thresholds are tighter than NBA's because hockey scores are sparser and OT/SO carries more variance).

The shipped surfaces:

• /best-bets — every league's sharpest model-versus-market gaps tonight, ranked by edge magnitude. Edges under 3pp sit inside model noise; sharps tend to live at five or higher on the same side.
• /box-scores — every live, final, and upcoming game across all leagues, sorted by what's happening right now. Click any card for the full elite box-score treatment: stat compare strip, top-performer cards, both-team player tables at the top of the page.
• /model/accuracy — the live track record. 7-day, 30-day, and 90-day hit rates by confidence bucket, the calibration plot, the full sample CSV. The MLB back-test sits at 55% on 2,932 games; the cross-sport hit rate updates every page load.
• /methodology — the actual Elo loop, step by step, with an interactive calculator that runs the same math the production cron applies after every final.
• /draft-2026 — 318 prospects ranked by an analytics composite that weights combine percentiles, age-adjusted college dominance, position scarcity, and an outlier bonus for prospects who break the historical comp distribution.
• /horse-racing — the full Grade 1 ladder profiled with race history, recent winners, and the betting trends that decide each race. Triple Crown deep pages, the Breeders' Cup, every major prep, plus a complete bet-types reference.

Receipts over takes. Every model edge has a timestamp from before the game started. The prediction goes into a public table. The result updates the same table. There's no way to retroactively cherry-pick the dashboard.

Cross-sport in one place. Most betting sites specialize. We don't. The cross-sport leaderboard at /best-bets merges the sharpest edges from every modeled league into a single ranked view. Same with the box-score index at /box-scores — every live game across MLB, NBA, NHL, WNBA, NFL, and NCAA in one ranked feed.

Free during beta. No signup. No paywall, no email gate on picks, no aggressive subscription push. Affiliate revenue from the sportsbook CTAs funds the work. Picks behind a paywall is a worse business model than picks that are right.

Honest data provenance. Every stat traces back to a public source: ESPN scoreboards, MLB Stats API, NBA Stats, the Odds API. The refresh schedule is documented on the methodology page. There are no hand-cleaned aggregates pretending to be live.

MLB is the canonical receipt: 55.1% hit rate across 2,932 games in back-test. That number is the headline because baseball has the sample size — 162 games per team gives an Elo derivative something to converge on.

NBA followed once we had a full year of closing-line data to back-test against. WNBA and NHL came online with the same Elo skeleton and per-sport tuning. NFL gets the prop-modeling treatment instead (Elo alone doesn't beat the consumer-site benchmarks we hold ourselves to, given the smaller sample and the injury variance).

On any given night, the model's top edges typically sit at 5 to 8 percentage points against the closing line. That's where sharps tend to live. The full edge column is at /best-bets.

NFL prop model shipping over the summer ahead of Week 1. Same posture as the moneyline edges: predictions logged before kickoff, public hit rate by confidence bucket.

Live race ingest for horse racing. The reference content is live today (Triple Crown deep pages, BC Classic, every major prep) but live odds + model picks need a paid data provider. That's on the roadmap.

Email digest for the cross-sport edges and the previous day's track-record update. Opt-in, no signup wall to see the picks, just a once-a-week recap of what the model called and how it landed.

One builder. Model assistance. Shipped from a spare-bedroom studio against a laptop. The whole codebase is the result of six months of focused work: per-team Elo across four leagues, the closing-line ingest, the track-record table, the cosmic-register design system, and the writing posture that runs through every page (lead with facts, drop the em-dashes, kill the AI vocabulary).

If you bet seriously, manage a fantasy league, or run a sports content site that wants honest receipts, I'd like to hear what you think. Email [email protected] or open the track record and start clicking.