Proof

Every opportunity we have ever surfaced, with the rule that decides it written down before the outcome existed. Nothing here is chosen after the fact — the scan writes each row, and this page has no filter that could hide one.

Corrections: 9 results have been corrected and now read as No result. Each compared two prices quoted against different share counts — a stock split between the entry and the resolution — so the difference was never a return. Together they carried -272.4R, which the record above no longer includes. The superseded rows are still listed below, marked, and never edited — corrected 12 August 2026.

Calls on record

1790

Resolved

146 of 601 right

Awaiting resolution

1189

Recording since

26 August 2026

146 of 601 resolved calls cleared their bar — 24.3%

Every resolved call is in that denominator, including the ones that went nowhere. Counting only clear wins and clear losses would drop them and inflate the rate.

Proof: how every opportunity played out

Graded outcomes: 601

not yet enough data to prove the grade order

confirmed48% hit rate (from 125 calls)
likely51% hit rate (from 398 calls)

Educational analysis, not advice. We hide any rate until we have enough calls to report it honestly.

The ledger

Showing 951 to 960 of 1790, newest first. Ordering is fixed and there is no filter, so nothing can be sorted or hidden out of view.

What counts as right is the bar we committed to when the call was made: the move it has to clear, from the price at the time, by the date shown. It was written down before the outcome existed and cannot be changed afterwards. Grading is a symmetric band, not a single line — clearing the bar is right, moving the same distance the other way is wrong, and landing between the two is neither. A near miss is recorded as its own outcome rather than rounded into a win or a loss, because rounding it either way would quietly flatter or damage every tier.

Instrument	Conviction	Published	Decided by	What counts as right	Outcome	Method	Record hash
NTRANatera, Inc. Common Stock	likely	3 Aug 2026	17 Aug 2026	Falls 5.0% from $261.99	Wrong Correction to an earlier result	proof-methodology-v2	4dd07ce97836…
PAHCPhibro Animal Health Corporation	likely	3 Aug 2026	17 Aug 2026	Rises 6.7% from $36.33	Awaiting result	proof-methodology-v2	2fb9be8c04be…
PAHCPhibro Animal Health Corporation	likely	3 Aug 2026	17 Aug 2026	Rises 6.7% from $36.33	Neither Correction to an earlier result	proof-methodology-v2	df1043d3c604…
PAGPPlains GP Holdings, L.P. Class A Units representing Limited Partner Interests	confirmed	3 Aug 2026	17 Aug 2026	Rises 2.0% from $26.45	Awaiting result	proof-methodology-v2	c032654759ec…
PAGPPlains GP Holdings, L.P. Class A Units representing Limited Partner Interests	confirmed	3 Aug 2026	17 Aug 2026	Rises 2.0% from $26.45	Wrong Correction to an earlier result	proof-methodology-v2	91f74e56e0dd…
SBUXStarbucks Corporation	likely	2 Aug 2026	16 Aug 2026	Falls 2.8% from $105.85	Neither Correction to an earlier result	proof-methodology-v2	804bfa6c2c19…
SBUXStarbucks Corporation	likely	2 Aug 2026	16 Aug 2026	Falls 2.8% from $105.85	Awaiting result	proof-methodology-v2	648343025fba…
CSCOCisco Systems Inc	confirmed	2 Aug 2026	16 Aug 2026	Falls 3.9% from $112.48	Neither Correction to an earlier result	proof-methodology-v2	b1a827ce1859…
CSCOCisco Systems Inc	confirmed	2 Aug 2026	16 Aug 2026	Falls 3.9% from $112.48	Awaiting result	proof-methodology-v2	1599965175d9…
DUKDuke Energy Corporation	likely	2 Aug 2026	16 Aug 2026	Rises 1.7% from $125.43	Neither Correction to an earlier result	proof-methodology-v2	0cc1cea8c924…

Backtested record

replayed history, not live calls

Before any of this went live, we ran the engine over 2019 to 2026 of market history to see whether the signal held up: 1,434 company events, measured against 6,175 ordinary weeks for the same names. The test was designed before it was run, so the results could not be tuned afterwards to look better.

Move size (in normal-swing multiples)	Event weeks	Baseline weeks	How much more often	After removing the market’s move
≥1.0σ	52.4%	23.9%	2.19x (likely range 1.79 to 2.69x)	2.65x (likely range 2.09 to 3.34x)
≥1.5σ	36.5%	9.8%	3.74x (likely range 2.72 to 5.15x)	4.67x (likely range 3.27 to 6.66x)
≥2.0σ	26.2%	4.1%	6.4x (likely range 4.03 to 10.08x)	7.9x (likely range 5.03 to 12.74x)

The catalyst edge holds up whether the market is rising or falling: 4.00x more often in weeks the market rose, 3.29x in weeks it fell.

Momentum is honestly closed: net of the market’s own move, its 10-day directional hit rate is 49.1% (CI 47.0%–51.4%), no better than a coin flip, which is why momentum on its own adds no confidence and we price these trades assuming no edge.

Replay evidence, not the live record: these grades come from replayed history, not from calls the engine published in production. The live track record above grades itself as each call resolves.
Today's list of names only: companies that disappeared before our data began are missing, which can flatter results. Names that delist after 2026 keep their record, so this only improves over time.
Earnings dates as known today: we did not record when past dates were revised, so a small quirk remains that if anything understates the result. New revisions are now captured.
The replayed set of names differs from what the live scan shows day to day, so these rates describe the signal itself, not the exact mix you would have seen on any given day.

How to check this yourself

Each row is hashed over the previous row’s hash. Change any past row and every hash after it stops matching, so the history cannot be quietly rewritten.
Grades are appended, never written over the original. The call and its outcome are separate rows in the chain.
The method version is stamped on every row. Calls decided under different versions are different experiments and we do not pool them into one number.

What this does not claim

The record is young. Until there are far more resolved calls, no hit rate here should be read as a forecast of future performance.
Outcomes are measured on closing prices. They do not model fees, slippage, or the price you would actually have got.
Surfacing an opportunity is not advice and not a recommendation to trade. It is a statement about evidence, published so it can be checked.

Educational analysis, not investment advice. Pulse surfaces and explains structures; it does not recommend securities or know your full circumstances.