Evidence-first deskThe method
From claim to settled outcome
The pipeline below is the actual state machine behind the portal, not a diagram drawn for this page. Nothing reaches you that has not moved through all five stages, and nothing leaves the record once it has entered.
The pipeline
Each stage leaves an artefact behind
A stage is only complete when it has produced something durable and inspectable. That artefact is what the next stage reads, and what you can audit later.
01
INGESTClaims are collected, not curated
Publications, filings, social velocity, and prediction-market pricing land in one normalized claim store. Nothing is dropped for being inconvenient.
Leaves behind — A normalized claim, with its source02
SCORECorroboration decides weight
A claim only carries weight when independent sources agree. Conflicting evidence is flagged as conflicting rather than averaged away.
Leaves behind — An evidence score, and its inputs03
PUBLISHThe setup is fixed before it triggers
Entry zone, invalidation level, and targets are written down and timestamped up front. There is no room to reinterpret the plan afterwards.
Leaves behind — A timestamped ticket04
TRACKState changes are recorded, not edited
Forming, active, invalidated, expired, closed. A setup moves between states in the open; it is never quietly removed from the queue.
Leaves behind — A full state history05
SETTLEThe journal counts everything
Expectancy, hit rate, and drawdown are computed from the complete record, losers included, with the sample size shown next to every figure.
Leaves behind — A settled row in the journal
What we refuse to do
The constraints are the product
Most of the design work went into what the system is not allowed to claim. These four rules are enforced in code, not written into a style guide and then forgotten.
We do not call a feed live unless it is
Every data panel carries its own health state. When a provider is stale or unconfirmed, the panel says so instead of quietly showing the last known price as though nothing had changed.
We do not delete setups
An invalidated setup keeps its place in the queue and its row in the journal. Removing it would improve the numbers and destroy the only reason to trust them.
We do not publish a level after the move
Entry, invalidation, and targets are timestamped before the setup can trigger. A ticket amended later shows both versions, side by side, permanently.
We do not average away disagreement
When sources conflict, the claim is marked conflicting. A confident-sounding average of two opposite readings is worse than an honest admission that it is unclear.
Evidence score
How a setup earns its number
Four inputs, weighted and shown. A score you cannot decompose is a score you have no reason to believe, so every published setup exposes the parts it was built from.
01
Source quality
Primary filings and exchange data outrank aggregators, which outrank commentary. The tier is recorded on the claim itself, not applied afterwards.
35% weight02
Corroboration
Independent sources reaching the same conclusion raise the score. Two outlets repeating one wire count as one, because they are one.
30% weight03
Market structure
Liquidity, spread, and depth at the proposed entry decide whether the setup is tradeable at all. Thin books disqualify it regardless of the thesis.
20% weight04
Freshness
A claim decays. The score falls as the evidence ages, and the setup expires on its own rather than lingering as a stale idea.
15% weight
- Exchange filingPrimary
- Wire reportSecondary
- Aggregator echoDuplicate
- Prediction marketPricing
Two outlets repeating one wire count as one source, because they are one
The stack above is what corroboration actually looks like: a primary filing outranks a wire report, and an aggregator repeating that wire adds nothing at all. Prediction-market pricing enters as its own input rather than as confirmation of anyone’s opinion.
The method is easier to judge from inside
Open the portal and read a live queue against the evidence that put it there.