Order blocks are easy to find in hindsight and hard to test, because most descriptions leave room for interpretation. A mechanical test needs every choice written down.
Decisions a test must fix:
Which candle: the last opposite-colour candle before the impulse, or the last few candles?
Impulse size: a minimum move after the candle, e.g. 2 × ATR within 5 bars.
Structure break: must the impulse break a prior swing? Which swing size counts?
Zone: candle body or full range including wicks?
Validity: how many bars until the zone expires, and is it void after the first touch?
Entry and stop: edge or midpoint of the zone; stop beyond the far side plus a buffer.
Once these are fixed, results can be compared across pairs and sessions. In practice, small changes to the definition often change results more than anything else, which is itself useful information about how robust the concept is for you.
If you trade order blocks, which definition do you use, and have you tested it mechanically?
My mechanical order block test only kept the last opposite colour candle before a move of at least 1.5 times M15 ATR within six bars. Anything looser produced so many zones that the backtest was just curve fitting noise.
Zone as full wick range beat body only on EURUSD London, but failed on GBPJPY where wicks are huge. I now fix body for majors and full range for yen crosses, written in the test config so I cannot switch mid run.
Validity after first touch is the rule that hurt my live results most. Keeping a zone live for three touches looked great in sample and terrible forward.
Stop beyond the far side plus 0.3 ATR was more stable than a fixed pip buffer. Fixed buffers ignored volatility regimes and inflated my stop out rate in quiet weeks.
Have you locked impulse size and validity before you trust a win rate, or are you still iterating those two live? I freeze both for a full 200 trade block before changing anything.
The bug that hurt my first order block test most was lookahead. The zone was marked on the candle before the impulse, but the impulse only qualified six bars later, and my code let price touch the zone during those six bars.
Once zones became valid only after the impulse finished, the win rate fell from 58 to 49 percent. Nothing else in the rules changed, so most of the original edge was the test peeking at the future.
The second thing I do now is a small parameter grid before trusting any single setting. I run impulse sizes of 1.25, 1.5 and 1.75 ATR against windows of 4, 6 and 8 bars, which gives nine cells.
If only one cell is profitable, the result is noise. If seven or eight are positive and the neighbours look similar, the definition is probably capturing something real rather than one lucky combination.
Overlapping zones need a rule as well. Do you merge zones that overlap, or treat each one as its own trade?
The gain is measured on a smaller balance. Start with $10,000, lose half, and $5,000 is left. A 50% gain on $5,000 is $7,500, still a quarter short. To get back to $10,000 the account needs +100%.
The general rule is recovery needed = 1 / (1 - loss) - 1. So a 10% loss needs about 11.1%. A 20% loss needs 25%. A third needs 50%. Half needs 100%. Three quarters needs 300%.
The curve is gentle early and then takes off. Going from a 10% hole to a 20% hole raises the needed gain from 11% to 25%, but going from 50% to 75% takes it from 100% to 300%.
For a scalper this matters because the profits are small and slow, and one bad week of oversized trades can cost months of grinding to repair.
Where do you draw the line on drawdown before cutting size or stopping completely, and what made you pick that number?