Entry, exit and abstention as one testable procedure

A two-gate test for a mechanical-system

Editorial reading: evaluate a mechanical-system in two gates. First write entry, exit and skip so they can fire without a later human edit. Then resample the completed trades so one historical order of fills cannot stand in for the procedure.

The archive facts below describe a historical workflow. They do not establish present-day results.

Write the full procedure before scoring it

The archive reduced a completed position to the entry price, the loss exit, and the profit exit.

A reversal entry required a pin-shaped bar whose open and close sat in the top or bottom of the range, with a long tail, and only when that bar lined up with a support or resistance level.

That rule set was read from unmarked price charts across several time frames, without overlaying moving averages or other indicators.

Holding-style selects which setup can fire

Editorial reading: holding-style is the planned time in the trade that selects which setup conditions apply, such as a short swing versus a longer trend.

A separate mechanical service issued its own entry and exit alerts for intended holdings of days to a couple of weeks, so trade-class statistics could be collected without adding a discretionary overlay.

The mechanical search covered listed stocks and funds. Each candidate could be read as a swing, momentum, or trend setup, and the published alerts were limited to long entries.

A swing long waited for a cycle trough together with a momentum low. The alert was withheld until momentum reached an extreme negative reading, and the write-up wanted that momentum to turn up within a few days.

Count abstention on the same tape

Extra data-mined conditions discarded most otherwise valid setups before a paid entry alert was shown, so most first-pass patterns were recorded as skips.

Editorial reading: those skips are abstention. They belong on the unedited-result-tape as part of the same procedure, not as missing trades.

Resample so one fill order is not the procedure

Evaluation used two mechanical views of the unedited-result-tape. The first was a month-by-month profit-and-loss average. The second was a Monte Carlo draw of completed-trade outcomes from recent years to stand in for a year of trading, shown as a distribution that could finish higher or lower.

Editorial reading: that second view is monte-carlo-resampling. It draws completed-trade outcomes at random from a stored set to build many one-year paths instead of reading only the original sequence.