Trading Strategy Replication: Resolve Conflicting Exits

Trading strategy replication: Convert conflicting exit descriptions into separate explicit fixtures before comparing results.

A strategy replication starts with an unambiguous exit rule. Two implementations can disagree because they translated different sentences, even when both engines are functioning as designed. Resolve that ambiguity before comparing win rates, holding periods, or returns.

The Reddit discussion used conflicting high-versus-close exit descriptions. Author performance and replication claims remain unverified, and some replies were inaccessible. The following fixture is original and synthetic; it is designed to reveal a specification difference.

Trading strategy replication: write the decision contract

Separate the threshold definition, observation used for the decision, and first permitted execution event. Yesterday's high is not yesterday's close. Today's intraday crossing is not today's completed closing observation. A completed daily close cannot support a decision earlier that same day unless a different, genuinely available input was used.

Specify strict or inclusive comparisons. A price equal to the threshold might satisfy greater-than-or-equal but not greater-than. Declare how adjustments, missing bars, duplicate timestamps, and session boundaries are handled. For indicators containing a high-minus-low denominator, a zero-range bar needs an explicit abstention or other documented policy.

A synthetic disagreement

Suppose yesterday's high is $105 and its close is $100. Today's closing price is $102. The condition close above yesterday's close is true; close above yesterday's high is false. If tomorrow's opening price is $99, a next-open rule exits at its declared execution price only in the first variant. The second retains the position until another exit condition is met.

Those differences affect holding age, capital availability, and subsequent entries. One extra exit can change many later trades if positions share limited capital. Therefore a matched comparison must preserve the entire chronological account path, not simply calculate two prices for the same preselected list of trades.

Make execution costs directional

Create a second fixture with a synthetic long entry at a $100 reference and adverse entry execution at $100.10. Its exit reference is $101, with adverse exit execution at $100.90. Before fees, the realized difference is $0.80 per share. This checks the sign on both sides; subtracting a cost percentage from every price would improve one side incorrectly.

SEC order guidance distinguishes price protection from execution behavior. Your backtest still needs its own event and fill assumptions. Declare whether costs are absolute currency, basis points of price, or per-order charges. Preserve the raw reference and modeled fill separately so reviewers can audit them.

Reconcile the first disagreement

Export the indicator inputs, threshold, signal timestamp, order intent, execution timestamp, quantity, fees, and resulting position after each event. Compare two reconstructions at the earliest differing row. Aggregate return differences are a poor starting point because later discrepancies may all descend from that first difference.

Use Tradewink Advisory Research to locate the offline replay guide. Begin with the three-price fixture, freeze both rule variants, and confirm their expected exit decisions before adding historical data. Keep discrepancies labeled unresolved until the specification is settled. A successful fixture validates that contract; it does not reproduce the community author's results or demonstrate a trading edge.

Continue the research

Limit Order Backtests: Use Counterfactual Fill Ranges and Adaptive Backtest Overfitting: Log Research Decisions develop adjacent checks. Use a small synthetic fixture first; these educational diagnostics do not validate a trading edge or change account settings.