BB Return mt5: the optimised settings kept working out of sample, but not nearly as well

BB Return mt5 returned +1,311.1% on the window its inputs were fitted to, and +301.3% on a later window the optimiser never saw. It is still positive out of sample, which is more than most manage, but the gap is where the honest number lives.

It did not close a single losing trade

Over 54 trades the report shows a gross loss of zero and a win rate of 100%. At the same time the equity curve was down 37.0% at its worst. Those three figures can only be true together in one situation: losing positions were never closed. They were held until they came back.

A trade that is held rather than closed does not appear in the win rate, in the gross loss, or in the balance drawdown. It appears only in the equity drawdown — which is why that is the number quoted above, and the one this site uses everywhere. A system that always waits shows a perfect record right up until the one move it cannot wait out.

Parameters change how long it can wait. They do not change what happens at the end of the wait: no setting puts a floor under a position that is never closed. Read everything below with that in mind.

How the two windows were run

The optimiser only ever saw the first window. The winning inputs were then run once over the second, with no further tuning. Any number quoted as a result on this page comes from the second window.

Fitted window against unseen window

The left column is what the search produced by construction — inputs were chosen because they made those numbers. Only the right column says anything about a period that had not happened yet when the settings were picked.

How concentrated the profit is

80% of the profit came from 16 of 23 trading days — 69.6% of them. The return is spread across enough sessions that no single day carries it.

What the input list shows

These are read out of the report itself, not out of the sales page. A vendor can rewrite a description; the inputs recorded in a strategy tester report are whatever the run actually used.

• UseRecovery = true
• RecoveryFactor = 6.0
• MaxRecoveryTrades = 8

Where this leaves it

The out-of-sample number is positive, so the settings are not worthless. But the distance between the two columns is the part worth remembering: most of the fitted result was the fit. If you saw only the left column on a sales page, you would be looking at a number that had already spent itself.