Charleston, Illinois

AlphaAudit

AlphaAudit answers two questions about any trading track record: was the strategy any good, and were the trades even real? Give it a list of trades and it rebuilds the results from the fills themselves, measures what the fees really cost, compares the record against simply buying and holding, checks the fills against the market, and gives a verdict. No predictions, no advice. Measurement.

Back to the home page · Our research page

The problem

Trading systems are sold on claims: a win rate, a profit figure, a screenshot of a balance. The claims are easy to state and, for a buyer, almost impossible to check. A record can hide its fees, cherry pick its period, count its trades in the order that flatters it, or rest on fills that never happened at all. What a claim needs is not a better story. It needs an audit: the same fills, rebuilt in order, measured the same way every time, with the assumptions stated where anyone can read them.

Method

  • The input is a trade log. A CSV file, one row per fill: time, symbol, side, quantity, price, fee. Most exchange exports work as they are; the headers are matched flexibly, and anything the tool cannot read fails loudly instead of being guessed at.
  • The record is rebuilt trade by trade, in order. Positions are closed first in, first out, in both directions: a sell closes the oldest long first, a buy closes shorts before it opens a long. From the rebuilt record the tool measures win rate, profit factor, expectancy, maximum drawdown, and the result net of fees.
  • Fee drag is measured, not assumed away. Fees are the whole point of the audit, and the tool never guesses them: if an export has no fee column, one has to be added before the audit runs. The report states the gross result, the fees, and what the fees consumed.
  • The record is compared against buy and hold. A strategy that works hard to end up behind simply owning the asset has its answer, whatever its win rate says.
  • Claims are tested against the measurements. A seller's stated win rate, profit factor, or net figure can be handed to the audit with the log, and the report checks the claim against the measured numbers and says plainly whether it holds.

Were the trades even real?

A separate verification pass checks every fill against real historical market data. It flags a fill printed at a price the market never touched that day, which is the tell of a fake platform; a stock fill dated on a day the market was closed; and a symbol no market data exists for. Each record comes back VERIFIED, SUSPICIOUS, or UNVERIFIABLE. The desktop app pairs this with a broker and platform red-flag checklist, the manual half of scam detection.

Verdicts

  • PASS. At least 30 closed trades, a net positive result, sane fees, and any stated claims verified against the measurements.
  • FAIL. A net result at or below zero, fees that consumed the gross profit, or a stated claim contradicted beyond tolerance.
  • NOT ENOUGH DATA. Fewer than 30 closed trades. The statistics are shown, and the verdict is withheld. This is a feature, not a gap: most supposedly proven systems die exactly here, on a record too short to prove anything.

Public Audit 001

AlphaAudit's first public audit was run on a claimed record of 189 signals: the output of our own house signal engine, handed to the tool as a track record to verify. Rebuilt trade by trade, the record showed a $179,000 loss under an 820 percent fee drag: the costs of trading the system were more than eight times what the trading itself made. Verdict: FAIL.

We publish that audit first, on a record of our own, because an audit tool that has never failed its own house has told you nothing about itself.

Limits and open questions

  • Modeled fees are stated, not hidden. The Audit 001 record paid no real fees; the audit modeled them at 0.6 percent per fill and said so in its own report. A record with real fee data gets a harder audit than a modeled one.
  • Coverage has edges. The audit engine works from trade logs, spot records first; options and futures are not modeled yet, and it audits track records, not ideas: it does not replay a strategy to see what it might do next.
  • The authenticity check is only as good as the historical market data behind it. A record can come back UNVERIFIABLE, and that word means what it says: not proven false, not proven real.
  • AlphaAudit is not a broker and not an advisor. It never says what to buy or sell, and a PASS is a statement about a past record, not a promise about a future one.

Our research page · Back to the home page