HOW WE TEST

Method, not promises.

Backtest and live results disagree? Begin with the same trade definition and cost basis. Below: what the planned Checker measures, and what it cannot establish.

This page describes our testing approach and the planned Checker’s documented analysis. It is not a claim that every video runs a full strategy test. As of 9 November 2026, CET (Europe/Vienna). General education. No signals, no personal investment recommendations.

1. Rules first, results second

A test starts with written rules: entry, exit, initial risk, instrument, timeframe, the costs already inside the P&L and the costs we add. Where possible we fix a rejection criterion and an untouched holdout period before looking at results, and we count how many variants were tried. Every extra variant we look at makes a good-looking result less convincing.

The Checker includes a rules template and a SHA-256 fingerprint of the rule text. A matching fingerprint shows identical text. It does not prove when the text was written or that results were not inspected first.

2. Costs, counted once

We distinguish gross inputs from net inputs; we do not reconstruct an unknown gross result from a net-only file. Round-trip commission, spread and slippage can change the sign of a small edge. Costs that are already reflected in executed prices or in a broker's net profit must not be subtracted again. That is why the Checker asks one plain question before anything else: does this file already include your costs?

• Gross mode: net = gross P&L − mapped costs + signed swap − any extra stress you add.
• Net mode: reported net P&L − extra stress only. Mapped fees are not applied again.
• Stress: modelled costs at 0×, 1× and 2×. A stress scenario is not a measurement of your real slippage and not a forecast.

The free calculator uses separate break-even arithmetic: if a win pays +kR, a loss costs −1R and round-trip costs are cR, the break-even win rate is (1 + c) / (1 + k). With 1:1 payouts and 0.1R costs that is 55%.

3. An average is not evidence on its own

The planned Checker reports uncertainty alongside the sample mean:

• A simple interval: mean ± 1.96 × standard deviation / √n. An approximation, not a guarantee.
• A bootstrap with 2,000 resamples (fixed seed, so the result can be reproduced). With complete dates, trades are resampled in exit-day groups. Without dates, individual trades are resampled as if independent. Dependence across days, extreme tails and small group counts can make both intervals unreliable.

Neither interval can correct for trying many strategies, look-ahead bias, silently dropped losing trades or unrealistic fills. More trades do not fix those problems. There is no magic trade count that proves an edge.

4. Three cautious labels

• Not enough evidence: the interval overlaps zero or validation conditions are incomplete. This is not “almost profitable”.
• Fails after costs in this sample: both intervals are wholly negative, costs are confirmed and at least 30 trades exist. It does not prove failure in every future market.
• Survives costs in this sample: both lower bounds are positive and conservative conditions are met: costs declared complete, one prior variant, an untouched holdout, at least 30 trades and 30 exit days, no rejected or duplicate rows. These are software gates based partly on user declarations; the tool cannot verify those declarations or prove a future edge.

5. Drawdown means closed-trade drawdown

Drawdown is the largest peak-to-trough fall of cumulative realised P&L, ordered by exit time where dates exist. It is not intraday equity, mark-to-market risk or a prop firm’s trailing drawdown rule. A trade export alone cannot certify that an evaluation was or will be passed.

6. Data and examples

Worked examples in our videos and on this site are labelled as worked examples. Software demonstrations use invented (synthetic) trades. When an episode uses historical market data, it names the source, period, sample size and cost assumptions, and uses data only where we have the right to. We never present a simulation as live performance.

7. What we do not do

• No signals, copy trading, managed accounts or broker affiliate links.
• No income, win-rate or “pass your challenge” promises.
• No invented customers, testimonials or reviews.
• No upload of your trade file: the Checker runs offline in your browser. Trade inputs never leave the browser. The Checker uses no telemetry, cookies or browser storage.

8. Corrections

If we get a number wrong, we correct it and say what changed and when. A public changelog will be published with the first Checker release. Report an error: [email protected].

Trading can result in loss. Education and positive arithmetic do not establish a reliable edge.

Apply the method to your own record

Keep one completed trade, its cost basis and your selection history together. If you cannot reconcile that row, the summary stays unresolved. The planned offline Checker helps inspect supported files; it cannot verify your research history.

Check the exact supported layouts before considering the planned $27 one-time tool. Try the free arithmetic calculator without email or file upload.