An assay is only as good as the artifact it runs on, and the artifact is the one part of the job that only you can do. This page is published before intake opens — deliberately, in the same order as everything else here: you should be able to see exactly what an assay will ask of you before there is anything to pay for.
Two rules apply to every assay, before the per-assay detail:
Send the export, not a screenshot. Every assay needs the machine-readable file your platform produces — CSV, HTML statement, native report. A screenshot cannot be hashed meaningfully, cannot be parsed without transcription, and transcription is where numbers change. If your platform can display it, your platform can export it.
Do not clean the file. Gaps, requotes, cancelled orders, the losing trades — the defects are usually the finding. A tidied artifact does not make a better report; it makes a report about a file that never existed. The artifact is hashed on receipt, so the report is pinned to exactly what you sent — which protects you as much as it binds you.
Strip what the assay does not need. Cleaning the data is forbidden; removing your identity is encouraged. The assays measure fills, rules, structure and cost — never who you are — so where your platform’s export allows it, remove or redact your name, account number and broker account ID before sending. A statement that says ACCOUNT ****4821 audits exactly as well as one that names you, and the less personal data the desk holds, the less anyone has to protect. This is privacy engineering, not a compliance dodge: the measurement is identical either way.
Overfit Assay — the trade list, and the number that matters more
Export the complete backtest trade list: entry and exit timestamps, direction, size, and per-trade result. From MetaTrader, the Strategy Tester report (HTML or the detailed CSV). From cTrader, the backtest results export. From anything else, any per-trade CSV with those columns.
Then the part no platform can export: an honest count of how many variants you tried before this one. Different parameter sets, different symbols, different date ranges, versions you abandoned — each one is a trial, and the probability of backtest overfitting is arithmetic over that count. Understate it and the report will flatter you in exactly the way you paid it not to. There is no way for the desk to check this number; it is the one input taken on your word, and the report says so.
Execution-Cost Assay — the account statement
Export the full account history for the period you want measured, not the summary: every order with its timestamps, requested and filled prices, and state (filled, cancelled, requoted where recorded). MetaTrader: the detailed account history report. cTrader: the history export. Broker web portals usually offer a fuller statement than the terminal — if both exist, send both.
Prop-Breach Forensics — the statement and the rules
Two files: the evaluation account’s statement (the full order-level account history, complete), and the firm’s rule set as it was published to you — the page, PDF or email that stated the drawdown definitions, the daily loss rule, and any consistency rules. The interaction between those rules is usually the actual cause of death, and the analysis is only as good as the rule text it reconstructs against.
Dataset Forensics Assay — the file itself, and where it came from
Send the historical data file exactly as you hold it — do not re-save it through another tool first, because re-saving overwrites the structural fingerprints the audit reads. State where it came from (vendor, broker export, download link if public) and what, if anything, has already been done to it. Provenance is part of the audit, and “I don’t know where it came from” is an admissible and useful answer.
Stress Replay Assay — parameters, not the strategy
The Stress Replay Assay needs your sizing rules and exposure, not your entry logic: position size per trade (fixed, percentage, or the rule that decides it), maximum concurrent positions, the instruments you hold, account size, and leverage. The strategy’s signals stay yours; the replay is about what your sizing survives, and it never needs to know why you enter.
Feed Fidelity Assay — a quote log from your own terminal
The capture method bounds what the assay can claim, so this one is specific:
- cTrader: quote data can be exported per symbol from the platform’s tick data facilities; capture at least five full trading days on the symbols you care about, and note your account type — pricing differs by tier, and the report will state which tier was measured.
- MetaTrader 4/5: run a tick-logging EA (any minimal logger that writes symbol, bid, ask, and a local timestamp per tick) for the same window. The assay needs arrival order and time, so the log must be written as ticks arrive, not reconstructed from history afterwards — downloaded history is the venue’s account of itself, and this assay exists to measure what actually reached you.
Note your timezone and any VPS involved: a capture’s clock is part of its provenance.
Verdict Recompute — everything that exists, as it exists
Send the claimed result (the number, the report, the screenshot — here a screenshot is admissible, because it is the claim rather than the evidence), whatever inputs the claimant provided, and the stated method if one was given. Incompleteness is not a blocker — “cannot be recomputed from what was provided” is a verdict, and it is the one this assay most often returns.
Single-Trade Investigation — one trade, in full
Send the trade’s complete ticket as your platform records it — entry and exit, timestamps to the finest resolution the export carries, order type, and every fill confirmation — plus the platform’s journal or log excerpt for the session if you can export one, and the disputed window described in your own words: what you expected, what happened, when.
Timestamps matter more here than in any other assay: note your platform’s server timezone and whether a VPS was involved, because the reconstruction aligns your clock against the corpus clock before anything else is computed. Do not clean, annotate or trim the exports — the gap between what you expected and what the record says is the subject, not a defect to tidy away.
Outage-Window Report — the logs that prove you tried
Your platform’s Journal (MT4/MT5: the Journal tab, exportable from the terminal’s log folder) records every connection failure, rejected action and retry with timestamps — that file IS the outage evidence, so export it before the platform rotates its logs. Send it with the account statement covering the positions held through the window, plus screenshots or status-page captures if you took them at the time. State the window as you experienced it, in your own words, with your local timezone.
Payout-Denial Recompute — the record and the rules
Send your full account statement or platform export for the account in question, the firm’s rule text exactly as published to you — the dashboard page, PDF or FAQ that defines the consistency rules, drawdown terms and split — and the firm’s stated denial reason verbatim, ideally the email itself. The recomputation is only as good as the rule text’s completeness: if the firm cites a rule you cannot find in writing, say so, because that absence is itself part of the finding.
Execution-Chain Attribution — three logs, one event
Send the platform Journal and Experts logs covering the event (MT4/MT5 write both; the Experts log carries your EA’s order requests and the server’s answers), the statement rows for the trades in question, and the EA’s own log file if it writes one. Note the VPS provider and region if one was involved. The assay reads timestamps, requests and fills — never the EA’s code, which stays yours.
Copy-Trade Divergence — your fills against their record
Send your copying account’s statement and journal for the comparison window, and the address of the provider’s public record page (the MQL5 signal page, or the platform equivalent your subscription copies). Note your subscription’s copy settings where you know them — fixed lot versus proportional, and any delay setting — because sizing rules are part of the divergence anatomy, not noise in it.
What happens on receipt, for every assay: the artifact is content-hashed before it is opened, the hash goes in your report, and the data-handling terms — retention, deletion, the submission channel itself — are published in the same act that opens intake. If what you hold does not match any format above, that is not a dearer assay; it may be a configured engagement, and the honest answer might be that the artifact cannot support the question — in which case you will be told that before any work is billed, not after.
Documentation — page 9 of 9, in reading order