Mechanical systems need fill-aware failure modes

Design the procedure around fills

A mechanical trading system is a single testable procedure that turns rule inputs, market state, and execution constraints into entry, exit, or abstention signals over the system's holding period.

Editorial reading: treat that procedure as a fill-and-failure design, not a signal factory. Entries, exits, and abstention should be specified together so a dead rule can degrade into small, frequent, roughly break-even activity instead of buying into movement that cannot be filled.

When the breakout is already gone

A mechanical procedure that looked tradable on older, slower charts can stop working once prices jump so quickly that the breakout is already gone before an entry can be taken.

A system that buys into rapid movement can fail in live trading because fills are too costly. A weaker-looking system that works limit orders inside congestion can be the one that is actually executable.

Work the limit inside congestion

A limit order is a working order placed at a chosen price so the trader can seek a fill in congestion instead of paying for rapid movement, then account for partial fills and implementation cost over the order's life.

Semi-automated execution can keep moving a limit order toward a support or resistance level until the trader cancels or a fill arrives, instead of sending a single static order.

Prefer a small failure to a chase

A preferred failure mode for a mechanical system is that a dead rule degrades into roughly break-even churn rather than a large loss. Limit-order systems that insist on good fills often approximate that path.

Breakout-style systems that buy into movement do not fail gracefully. When the edge disappears, the same movement-chasing entry becomes the source of damage.

Robustness testing checks whether that same procedure still behaves acceptably when a rule fades, when live fills differ from the backtest, or when one model in a set of similar systems dies. Running several similar-horizon, differently specified systems at once is a robustness tactic so that if one procedure dies, the others can keep operating while a replacement is built.

Probe small, then code what you saw

Small live probes and historical tests are used together. One-lot or 100-share trades show how quotes and noise respond, and those observations are then coded and checked on past data.

Pattern detectors should be judged by whether they find the intended formations on charts. If they do not, the detector is rewritten rather than assumed to be correct.

Survival sizing is framed as trading small and frequently rather than placing oversized bets, so a developing procedure can be observed without ending the account first.