How we test

Three stages, in this order

Each stage is cheaper than the next, so we run them in order of cost. Nothing reaches a ranking without passing all three.

Stage 1 — Structural screening (free, applied to everything)

We read the vendor's own documentation, specification table and category assignment, and look for mechanisms with no loss ceiling: martingale multipliers, grid baskets, averaging into adverse moves, "recovery" modes, position locking, and explicit absence of a stop loss.

Negations are excluded — a page saying "no martingale is used" is not a martingale detection. This sounds obvious; getting it wrong inflates the failure rate by several percentage points and disqualifies exactly the candidates most worth testing.

What this stage proves: that a system will eventually reach zero, given a large enough one-directional move. What it does not prove: that a system without those mechanisms is profitable.

Stage 2 — Backtest reproduction

Products that survive stage 1 are re-tested against the vendor's own claims, under conditions the vendor usually does not use:

• Real tick data, not MetaTrader's interpolated 90% modelling. Interpolated ticks inflate scalping results several times over.
• Realistic variable spread from the broker we actually test on, not a fixed one-pip assumption.
• Vendor default parameters, with no optimisation. A result optimised on the same data it is reported against is not a result.
• A period that includes a stress event. Any window can be chosen to look good; the point is to include one where it did not.

We also compute two figures the standard report does not show:

• Profit concentration — how few trading days produced 80% of the profit. A smooth equity curve is often a handful of lucky days plus a long flat stretch.
• Cost autopsy — commission and swap as a share of gross profit. Systems that hold gold positions overnight frequently hand the entire gross profit to the broker.

What this stage proves: that a claim does not survive honest conditions. What it does not prove: that a surviving system will make money live. A backtest cannot demonstrate an edge — it can only remove candidates.

Stage 3 — Live and demo accounts

What survives stage 2 goes onto a real terminal on a real account, one account per product, identical starting balance, identical broker, started on the same day. Anything less than that is not comparable: a product on a $100,000 account will show a smaller drawdown percentage than the same product on $10,000, and a product started a month later traded a different market.

Results are recorded continuously, including the ones that fail. Every figure we publish is labelled with whether it came from a demo account, a live account, or the vendor.

What we cannot tell you

• Whether a system will be profitable. No test can. Thirty days of survival is not evidence of an edge, and we will not present it as one.
• How it behaves at your broker. Spread, slippage and execution differ. A scalping system especially may not reproduce.
• Whether the vendor will keep it working. Products are abandoned, and updates change behaviour.

Corrections

If a figure on this site is wrong, tell us. We will correct the page and note what changed and when, rather than editing it silently.