Roadmap.
No dates. Each phase unlocks when the ledger proves the one before it — the same standard we hold every prediction to. The gates below are checked against the live record, including the ones we're currently failing.
Gates read from the ledger · N 26
Now — Prove the record
In progressOne model, one market, every prediction logged before first pitch. Nothing scales until the sample says the edge is real. Right now it doesn't — and that's on the page, not under it.
| Gate | Ledger reads | Requires | Status |
|---|---|---|---|
| Graded sample size | 26 | N ≥ 150 | In progress |
| Closing-line value | -8.8¢ | avg CLV > 0 | Not met |
| Model version held stable through the sample | mlb-k-model-v1 | no mid-sample edits | Met |
| Contract counts on every new prediction | 0.0% | 100% of new rows | Not met |
Plainly: the record is currently CLV-negative. If it stays negative at N 150, the model doesn't graduate — it gets versioned, fixed, and the count starts over. A roadmap that couldn't fail wouldn't be worth publishing.
Real money, capped
Locked by § 01A small live bankroll on our own account, sized at half-Kelly with a hard 10% cap per bet. The point isn't income — it's proving live fills match what the model priced.
| Gate | Ledger reads | Requires | Status |
|---|---|---|---|
| § 01 fully met | no | all 4 gates above met | Not met |
| Live fill price vs. price the model saw | — | slippage ≤ fee-adjusted edge | Locked |
| Kill switch and drawdown throttle exercised live | — | tested, logged | Locked |
Every bet stays human-confirmed. Any edge reading over 15% halts the system — at that level we assume a bug before we assume a jackpot.
Sell only what's actionable
Locked by § 02The graded record stays free, forever. Paid access covers one thing: the pre-outcome feed, while it's still bettable. Flat fee only — never a cut of anyone's outcome. Subscriptions open when every gate below clears, not before.
| Gate | Ledger reads | Requires | Status |
|---|---|---|---|
| Paywall enforced at the route level | — | pre-outcome feed gated | Locked |
| Legal review complete | — | attorney sign-off, entity, 21+/jurisdiction gating | Locked |
| Signing-key custody | — | ingest key held solely by Brier | Locked |
| Every displayed figure traces to a graded row | — | no exceptions | Locked |
The founding 100 slots stay reserved until this section clears. We won't take a subscription for a feed the record hasn't earned.
New markets, one at a time
Paper trials runningThree candidate models — NRFI/YRFI, first-5-innings, batter-hits-NO — are publishing to the same hash-chained ledger as paper picks. Each graduates alone, on its own record, or not at all.
NRFI / YRFI
| Gate | Ledger reads | Requires | Status |
|---|---|---|---|
| Paper sample size | 0 | N ≥ 150 | In progress |
| Paper CLV | — | avg CLV > 0 | In progress |
| Independent versioning | — | own lineage, own count | In progress |
First 5 innings
| Gate | Ledger reads | Requires | Status |
|---|---|---|---|
| Paper sample size | 0 | N ≥ 150 | In progress |
| Paper CLV | — | avg CLV > 0 | In progress |
| Independent versioning | — | own lineage, own count | In progress |
Batter hits NO
| Gate | Ledger reads | Requires | Status |
|---|---|---|---|
| Paper sample size | 0 | N ≥ 150 | In progress |
| Paper CLV | — | avg CLV > 0 | In progress |
| Independent versioning | — | own lineage, own count | In progress |
Winter markets (NFL, NBA) follow the same door: paper first, own sample, own grade. A model that fails on paper never touches money or a subscriber feed. Paper picks are excluded from every homepage stat until they graduate.
Under research: managed deposits
ExploratoryWe're researching whether an investment-style product — where you deposit funds and the model trades them for you, no betting on your part — can be offered legally.
This would be managed money, which sits under a different and much heavier body of regulation than anything else on this page: registered-intermediary requirements, custody rules, and jurisdiction-by-jurisdiction licensing. So the gate here isn't a ledger metric — it's written clearance from regulatory counsel that a compliant structure exists. Until then, this doesn't exist, isn't offered, and takes no deposits. If counsel concludes it can't be done cleanly, this section is removed and we say so.
The record becomes the product
Long gameMultiple seasons, multiple markets, one unbroken chain. At that point the moat isn't the model — anyone can build a model. It's years of timestamped, honestly-graded history that no one can manufacture retroactively. API and data access grow from there.
What never changes
- —The predictions table is append-only at the database level. No phase adds an edit button.
- —Every public number traces to a graded, pre-registered row — or it doesn't appear.
- —Subscription pricing stays flat-fee. Nothing here is ever paid as a share of anyone's winnings.
- —Failed gates are published, not buried. This page shows the misses at the same size as the hits.
- —House trading and customer-facing products run on the same model but separate ledgers and separate money. Always.
Gate readings update with the ledger · verify any row →