We publish the model, not the marketing.
Most paid sports-pick services are wrapped in a persona, a story, and a win-rate they can't prove. We chose the opposite trade.
How VEA works
Every pick is published at issue.
Server-timestamped, content-hashed, and posted to a public daily anchor. We can't backdate. The track record updates live, and losing months stay on the page.
We track closing-line value, not just wins.
Beating the closing line is what separates a real edge from variance. CLV is captured at game-start, rolled into the graded log, and surfaced on the track record alongside win rate and ROI. Where the metric reads "Pending," the closing-line capture for that sport is still warming up — we don't fabricate the number to look better.
Calibrated probabilities, sized by Kelly.
Every pick carries a model probability, an implied edge, and a Kelly fraction. Track-record charts the long-run P&L so you can stake-size with full context — no guesswork.
No tipster theater.
No screaming Twitter videos, no inflated win rates, no fake credentials. The model is the product. If it stops working, you'll see it on the track record before we send a marketing email.
One spine, multiple sports
VEA isn't a single model — it's a collection of sport-specific probabilistic models that share one architecture. Each sport has its own Edge engine that produces a posterior probability for every outcome it covers, blends that against the sharp market price, ranks by expected value, and only publishes picks that clear an EV threshold.
The picks flow into VEA through a content-hashed handoff bucket. Each pick is timestamped at issue (server-side, immutable), graded at settlement, and tracked for closing-line value (CLV) — the single metric that separates skill from luck in betting.
Inside a single pick
- 1
Ingest the market
Pull live odds for every covered event from major US sportsbooks. Cross-reference sharp pricing to derive a no-vig fair-line baseline before applying any model adjustment.
- 2
Score the outcome
Run the sport-specific model. Table Tennis = ELO + match-context blend. MLB home-run signal blends historical similarity, matchup context, and park-and-conditions adjustments. Output: a posterior probability or ranking score for every available bet.
- 3
Compare to the market
Gap = model probability minus implied market probability. Translate to expected value per unit staked. Compute Kelly fraction. Anything below our EV threshold gets dropped — only the sharpest picks ship.
- 4
Hash and publish
Each pick gets a server-side timestamp and a deterministic hash over its core fields. The hash is the pick’s permanent fingerprint — once published, we physically cannot alter the pick without the hash changing. Anyone can recompute it from what we display.
- 5
Deliver to your tier
Pro ($49/mo) sees the full daily slate with sport-tabbed filtering. Pro+ unlocks the research suite: matchup heatmaps, hitter zone profiles, pitcher arsenal + zone-by-pitch heatmaps, and the multi-book Odds Graph. Elite adds the real-time sharp-money signal feed.
- 6
Grade at settlement
When the game ends, the resolution service writes the outcome (win, loss, push, void) to the immutable graded log. Units won/lost roll into the track-record stats automatically. No manual editing.
- 7
Track closing-line value
A separate job captures the closing line for each pick at game start. CLV = whether we beat the close (the metric that survives variance). The CLV-beat percentage is published on the Track Record page alongside win rate and ROI.
Every pick produced by VEA flows through this pipeline. The content hash at step 4 is what makes the track record auditable — anyone can verify a pick wasn't altered after the game.
See the model live.
The Track Record page shows every pick we've published, graded, and CLV-tracked — public, no login. Judge the model before you pay.