Historical Paper Replay and Forward Testing: Different Evidence

Historical Paper processes observations already saved in the past. Advancing a simulated account one day at a time does not establish that new information was collected on those historical dates. Record replay dates separately from the actual execution timestamp.

Forward testing fixes a rule before recording newly arriving information and decisions as events unfold. Its evidence differs from streaming already-known history in the same order. Neither approach, by itself, proves that real orders filled.

Suppose fictional prices for Monday through Wednesday are saved as 100,102,101. Processing them one session at a time on Friday does not prove that the Monday observation and decision were recorded on Monday. Describe the operation as a historical replay executed on Friday.

For a forward-observation example, Monday's available inputs and decision would be saved on Monday, followed by newly available information on Tuesday and Wednesday. Later rule or data revisions need their own history. This is a fictional illustration, not a report that forward testing was performed.

Stepping through history helps inspect the date of a decision or journal event. Run-all processes that history in a batch. Processing speed and button presses do not change the fact that the observations were already known.

Public Demo 4 uses a new, isolated research account for historical replay. Reopening it read-only queries state without advancing its cursor. A new query timestamp does not mean newly arriving future observations were added.

For replay, inspect price conventions, decision timing, fill assumptions and account calculations. When the information was historically available is a separate question. A replay clock is not evidence of the data's available_at timestamp.

Forward simulations can also use hypothetical fills and quantities, leaving liquidity, latency and rejected orders untested. Actual trading would need execution records, but this article and its public lessons do not conduct it. Limit claims to the type of evidence available.

State whether an output is replay or forward observation, along with execution time, input period and the point at which its rule was fixed. Paper succeeded does not identify what was checked. Read the public courses as RESEARCH-ONLY with pit_guarantee=false.

A useful deliverable says something concrete, such as replayed Monday-to-Wednesday saved observations on Friday and reconciled the records. Do not describe it as forward or actual-trading evidence if those activities did not occur. Favorable historical results do not guarantee future returns.

Read What is backtesting? · Lookahead bias · Out-of-sample testing to place this topic alongside other research checks. Confirm documented operations and limits in the official GUI research courses. Use documentation for your selected build; if public main differs from your environment, defer unverified claims.

This article teaches research procedures. Numerical examples are fictional teaching illustrations, not actual investment performance. It does not recommend securities or trades or guarantee future returns. Historical public courses do not establish complete historical information availability. Disclaimer

Explore the backtesting learning path

• What Is Backtesting? What Historical Tests Can and Cannot Tell You
• What are transaction costs? Why fees change backtest results
• What is lookahead bias? Using information that was not yet known