Averaging into a position on a price grid

Forty-two pyramiding and averaging configurations were tested against a constant-size baseline at matched risk. None beat it. The ones that appeared to were carrying more risk, and the comparison was not matched.

The deeper problem is what the null looks like. For any strategy that defers its losses — which is exactly what a grid does — a shuffled-bar placebo comes back strongly positive. Randomise the bar order, destroying every real relationship, and the grid still makes money on paper, because deferring a loss converts it into an open position that a finite sample often never resolves.

So the correct question is never 'is the grid profitable'. It is 'is the grid more profitable than the same grid run on shuffled data'. Tested that way, the answer has been no every time.

Continuation itself decays fast: measured on the same tape, the expected follow-through fell from roughly +0.22 ATR to +0.07 ATR by the time price had travelled one ATR. There is less to add to than the premise assumes.