Roadmap

This page is the honest version. It lists what works, what is being built, and the gaps between the current product and what a fully automated betting assistant would actually be.

What works today

Every pregame bet on the board gets a graded verdict with an edge and a confidence score, across ten leagues and 62 market types. The grading method is published in full, and the coverage matrix states exactly what can and cannot be priced.

Markets the engine cannot ground appear as UNRATED rather than being hidden or guessed at, and grades built on thin samples are marked THIN DATA rather than presented with false confidence.

Being built next

The calibration ledger. The single most important missing page. Of the bets graded at 55%, did roughly 55% win? The computation exists and the settled sample is accumulating. Until it can be shown, no track-record claims are made anywhere in the product — a confidence number you cannot audit is decoration.

Line shopping. Selections currently carry one best price from one book. Showing every book's price on every selection is the highest-value missing feature in the product, because line shopping is the most reliably profitable habit in betting and right now WeBall cannot help you do it.

Sportsbook deep links. The plumbing to build a DraftKings, FanDuel or BetMGM bet slip from a graded selection exists in the API and has no button attached to it.

Parlay correlation. The engine carries measured correlation coefficients for common leg pairings — the kind of data almost nobody publishes — with no interface to use them.

Alerts. A closed set of six pregame alert types — a line crossing a number, a grade reaching a tier, a kickoff approaching — that you arm yourself or through the assistant; the rows exist and the matcher is being wired into the same 15-minute pass that grades the board.

Personalisation. A profile — the sports, teams and activities you told us about — that the assistant, the alerts and the feed all read from one row; onboarding and settings are being built on top of it.

A daily brief. One digest a day on your sports, in a fixed order — the market's view first, then where the model disagrees, then what you armed — composed from rows that already exist rather than written on the fly.

Named gaps, not yet started

These are the things a betting assistant should do that this one does not.

Bet tracking and P&L. Without it there is no personal closing line value, no realised ROI, and no foundation for the three items below. It is the keystone.

A bankroll manager. Kelly sizing exists inside the engine but there is no concept of your bankroll, so it cannot recommend a stake.

Arbitrage and middle screening. Both fall out almost free once multi-book prices exist.

Promotional EV tools. Evaluating boosts and free bets inside the product rather than only in the calculators.

Player pages. Per-player prop history exists in the warehouse with no route to view it.

State availability. There is no geo awareness, so the product cannot tell you which sportsbooks are legal where you are.

A public developer API. The FastAPI surface and an MCP server both exist and neither is documented, keyed, or rate-limited for outside use.

What is deliberately not planned

Live and in-play grading. Everything is pregame by design. Following in-play prices well is a different engineering problem, and doing it badly would be worse than not doing it.

Picks and tout content. WeBall grades prices and shows its working. It will not publish "locks", guaranteed plays, or a subscription tier built on telling people what to bet.

A social copy-betting feed. Following other bettors' picks is a retention mechanic, not an edge.