Honest version

What this does, and what it does not

A companion to the Risk Disclosure · Terms · Privacy

Most trading products are sold on their best day. This page is the other list — written because the fastest way to lose someone’s trust is to let them discover a limit for themselves that we already knew about.

The one-line version. Trade X is research and automation software. It reads charts, computes indicators, argues with itself, backtests, and can place trades on your broker account through your terminal. It is not a signal service, it does not manage money, it has never been shown to be profitable, and the measured results of our own strategy vault are below breakeven.

What it does

  • Computes five non-repainting indicators on real market data, and shows you the same numbers the AI reads. No hidden second set.
  • Runs a five-stage AI council over a chart — data, analyst, a bull/bear debate, a judge, a decision — and streams its reasoning so you can see why, not just what.
  • Backtests a strategy over a window with explicit costs, and gives every published figure a run id you can use to regenerate it.
  • Publishes our whole strategy vault, including the strategies that failed our own credibility bar.
  • Separates our measurement from the vendor’s claim, and shows you when the two disagree.
  • Places trades on your own MetaTrader terminal, under a Risk Supervisor that can only ever reduce exposure.
  • Says “unknown” when a figure is unknown, rather than printing a zero.

What it does not do

  • It does not tell you to buy or sell. There is no signal service here and no instruction to act. The output is an argument with a confidence attached.
  • It does not claim a win rate or a profit. Not anywhere, not in any tier. If you find a number on this site implying one, it is a bug — tell us.
  • It does not hold your money or touch your broker password. Ever. The terminal logs in on your machine; we only publish instructions to it.
  • It does not manage your account. We are not a broker, adviser, fund or fiduciary, in any jurisdiction.
  • It does not know the future. An AI can be confidently, fluently wrong, and it is most convincing exactly when it is.
  • It does not present a backtest as a live result. They are labelled differently because they are different things.
  • It does not guarantee a fill, a price, or that a stop will hold. Gaps and slippage are your broker’s reality, not ours.

Our own vault, measured by us

Every figure below is read live from /api/quantx/vault when this page loads — the same endpoint the library and the screen use. Nothing here is typed in by hand, which is the point: a page about honesty that hard-codes its numbers goes stale and starts lying by omission. If the endpoint is unreachable you will see a dash, never an old number.

—loading

Reading the vault…

Read that expectancy honestly: it is below breakeven. Most of what we have measured does not clear our own bar, and we publish it anyway, because a research partner who shows you only the winners is not showing you research. The bar itself is stated plainly — at least 60 trades, a profit factor above 1.2, and positive expectancy — so you can disagree with it.

What that screen is not. It is a single in-sample window. There is no out-of-sample split and no walk-forward in those figures. Treat it as a filter that removes obviously broken strategies, not as evidence that the survivors work. A strategy showing a profit factor of 86 on six trades is noise, not an edge, and several of them do.

Where the AI can be wrong

  • Fluent and wrong. The council produces well-argued cases for both sides by design. A confident tone is a property of the model, not evidence about the market.
  • Numbers versus words. The figures are computed in code from real bars. The explanation around them is model-generated and can misread its own inputs.
  • Stale or thin data. An analysis refuses to run on generated prices or on too few real bars — but a market can also simply be untradeable in ways no indicator shows.
  • Regime change. Everything measured here was measured on the past. Edges decay, and the ones that decay quietly cost the most.

If you only do one thing

Run it on a demo account first, for longer than feels necessary. Every failure mode in the Risk Disclosure — configuration error, duplicate instances, a sleeping VPS, a broker symbol that does not match — shows up on demo at no cost, and shows up live at full price.

Questions, or a number on this site you think is wrong: [email protected]. We would rather fix it than defend it.