Instrument-maker · pre-launch

Measurement instruments for the claims trading makes — feeds, backtests, desk risk, chart vocabulary.

I went looking for honest, instrumented trading tools. I couldn't buy them, so I built them.

Broker feeds audited, backtests tested for honesty, desk risk held to its own written rules, chart vocabulary measured rather than believed: each instrument publishes its measurement method before it runs, hashes the artifact it measured from, and states in writing what it does not establish. Zero performance claims — not in the copy, not in the examples, not ever.

Hadal (/ˈheɪ.dəl/) — from the hadal zone, the ocean's deepest realm.

Explore the instrumentsFree calculatorsThe glossary

Or read the method first

method before measurement · hash before publication
Proof strip

Three measured facts belong here.

The strip below is where a measured fact stands beside the hash of the artifact it was computed from. Not one of them exists yet — so it carries the three commitments instead: what will be published, and the page that will carry it. Each seal below becomes that artifact's hash on the day it lands, and not one day earlier.

  1. 01/03Pre-registration

    The feed battery, written down before it is run

    Every dimension, threshold and exclusion rule of the fidelity battery, fixed and hashed before the first broker feed is touched — so the auditing method cannot be tuned afterwards to flatter a result.

    sha256NOT YET PUBLISHED

    Feed & Execution receipt page
  2. 02/03Computed artifact

    An overfit audit published with what it audited

    The probability of backtest overfitting, the deflated Sharpe and the effective number of independent trials — each published beside the backtest they were computed from and the code version that produced them.

    sha256NOT YET PUBLISHED

    Backtest Honesty receipt page
  3. 03/03Raw corpus

    Broker portrait #1, re-derivable by a stranger

    One broker feed measured on the pre-registered battery and published as a profile, with the tick corpus downloadable and content-hashed so the published figures can be recomputed independently.

    sha256NOT YET PUBLISHED

    The Observatory
Founding thesis
"I searched for it, couldn't buy it, and built it — module by module. First-frustrated-customer validation."

The category — honest, instrumented trading tooling — doesn't exist because the market's incumbents discovered that promises sell better than measurements and stopped trying.

The addressable customer is the burned-serious minority — traders who want an instrument, not a lottery ticket. They are a small fraction of the market and the only fraction that retains.

What I measure vs. what gets sold
Publication cadence floorMeasured per broker
Two-sided coordinationDetected
Spread regime detectionClassified
Freeze/rollover behaviourInstrumented
What the industry sells instead — these three rows describe other people's review sites, not Hadal:
Industry standard disclosures★★★★★
Typical "reviews"Affiliate links
Methodology publishedNever
Instrument Suites

Seven suites.
One discipline.

Each instrument earns its place by solving a problem nobody else will touch. I don't resell external analytics or white-label third-party models — nothing here is someone else's engine with my name on it, and whatever has been built was built by me. What is running, what is on hold and what is still a design differs across the catalogue, and each instrument's own page states which it is. The mathematics is mostly the field's own: the probability of backtest overfitting and the deflated Sharpe ratio are López de Prado's, the annualisation correction is Lo's, HAR-RV is Corsi's. Where I arrived somewhere the literature had already reached, I say so. Where I got there on my own, I publish the reasoning so you can check that too. What makes any of it worth paying for is the part nobody does — guarding the assumptions, refusing to render a number when they fail, and publishing the method so both can be checked. The field is not short of mathematics. It is short of anyone enforcing the conditions the mathematics requires.

All 21 instrumentsBrowse by suiteHow each one measures

Calibration platecounts derive from the catalogue · none is a performance figure
21Instrumentsin 7 suites
0Performance claimsby policy, forever
97Terms definedpublished glossary
1Fabrication caughtpost-mortem published

Instruments, suites and defined terms are read from the published catalogue when this page is built. The zero is pinned by policy. The one is a published post-mortem. Nothing on this plate describes a result.

Those are instruments you run yourself. If you would rather one measurement was run for you — on your own statement, by me, ending in a written report — that is the assay desk. The whole menu and every anticipated price are published there. Intake is not open yet, and the page says so before it lists a single assay.

Simulated demonstration

A broker feed has a grammar. I measure it instead of trusting it.

The tick panel here is a simulation of what a measured broker feed looks like: every tick timestamped, every gap visible, every pattern named. The first real portraits publish with their raw, content-hashed artifacts.

Cadence floor
NOT YET PUBLISHED measured value
Event clock
ms-resolution instrument spec
Non-crossing
enforced instrument spec
EURUSD · simulated tick streamSIMULATED

SIMULATED Synthetic ticks drawn in your browser from a fixed pattern. No broker, no feed, no measurement — the panel shows the shape an instrument reads, never a result it produced.

Process Receipts

Receipts are the marketing.

A receipt is cheap to publish and expensive to fake. That asymmetry is the whole strategy — and it is legible to exactly the customers I want.

04 published01 in force, unpublished02 not yet published

Zero performance claims
Constitution

Agent-governance constitution

Every agent session that touches this product is bound by a written constitution: no fabrication, no self-attestation, receipts or it did not happen.

PUBLISHEDRead the disciplines
Forensic audit

Agent fabrication caught

In July 2026 an agent working on my core engine fabricated results and declared the system "COMPLETE" when nothing worked. I caught it, documented it, and published the post-mortem.

PUBLISHEDRead the post-mortem
Definitions

Every term defined before it is used

The vocabulary a measurement depends on is published as a glossary, so a reader can check what a word means here before deciding whether a claim built on it is worth anything.

PUBLISHEDOpen the glossary
Build rail

Fabrication fails the build

Placeholder hashes, invented provenance and unverifiable dataset claims are build failures. The site cannot ship a number that does not trace to a real artifact.

Enforced in CI · not itself a published artifact
Pre-registration

Feed fidelity battery

The battery’s definitions and thresholds are written down and hashed before any broker is touched, so the method cannot be tuned to flatter a result.

NOT YET PUBLISHEDThe feed receipt page
Kill ledger

Kills publish with registrations

Hypotheses that fail publish with the same visibility as the ones that survive. Four engine-registry kills already stand in the ledger, with effect sizes, intervals and artifact digests; the Observatory’s own first entries arrive with the live pipeline.

PUBLISHEDRead the ledger
Walk-forward

Out-of-sample verification

In-sample calibration is not evidence. Every measurement faces out-of-sample verification before publication — no exceptions, no post-hoc adjustments.

NOT YET PUBLISHEDThe methodology page
Broker Feed Observatory

Profiles, not rankings.

A portrait, not a podium. Each broker is measured on the same dimensions and published as a profile you read — never squeezed into a single score, because one editorial number is the easiest thing in this industry to buy.

The value is that a stranger can check my work: every published figure ships with the data and the method needed to re-derive it independently.

RecomputableContent-hashedWarts published first

Those three words describe the standard a portrait must meet before it publishes. They are policy, not results — no portrait has been published against them yet.

Broker Portrait · schema
PRE-LAUNCH
Cadence floorNOT YET PUBLISHED
Event clock resolutionNOT YET PUBLISHED
Bid-ask coordinationNOT YET PUBLISHED
Non-crossing checkNOT YET PUBLISHED
Tick depth (sample)NOT YET PUBLISHED
Spread regimesNOT YET PUBLISHED

Observatory latest

Latest broker portrait
NOT YET PUBLISHEDNo portrait exists. The pre-registered method publishes before broker #1 is measured — that ordering is the point, and it is checkable: read the method.
Roster
PUBLISHED · ALPHABETICAL · UNRANKEDThe brokers queued for the battery are listed in name order and carry no position, score or badge — a queue is scheduling, never a judgement: open the Observatory.
Field notes
Field notes publish the question and the method. The numbers publish with the artifact that settles them, and not before: all field notes.

"What does your broker's feed actually do when you measure it instead of trusting it?"
— The founding question

Discipline culture

"I caught my own system fabricating results. Then I built the constitution that makes it impossible."

The core asset is not any single module. It is the discipline culture — mandatory internal structures, receipts, pre-registration, retraction with original errors preserved verbatim, the register that caught a fabricating agent and two honest ones compressing.

Receipts promise nothing about returns, which is precisely why they are safe to publish and hard to counterfeit. Filtering is the feature — these instruments tell you things you may not want to hear, and that selects the buyer. What survives the funnel retains.

Read the full storyView methodology
Early Access

Be first at launch

Join the waitlist for early access, first look at pricing, and build updates. No spam — just signal.

Submitting stores your email address, and the name and answer if you give them — nothing else — on Hadal's own infrastructure. No third-party mailing provider.

Unsubscribe anytime, from the first email.

Standing policy

  • No promise of profit, ever. Hadal makes no performance claims and carries no implied edge. Past measurements describe instrument behaviour — never future returns.
  • You own risk management. Hadal cannot control it and does not insure it. Good tools do not fix bad discipline — and this site says so.
  • Analytical tools, for discretionary use. Nothing here is investment advice or a recommendation to trade. Every decision, and every outcome, is yours.

The same statement stands in the footer of every page.Terms Privacy