BDS / Intraday usability
Replaying a Day-Trading Signal Without Using the Future
Reconstruct only what the declared decision could know. A finished chart is not a record of delivered instructions.
Freeze the instruction and information clock
Save the first instruction, its missing fields, stated publication time and actual observation time before inspecting later prices. Record the venue, time zone and the meaning of candle labels. A bar can be labelled by its start or end while data becomes available on a different schedule.
Keep the research observation-window convention explicit. Do not silently switch an established research fill rule to first-available execution, or extend its window with later prices. A subscriber-execution study is a distinct evidence class with its own records.
Describe the data resolution
Name the price source, interval, coverage and availability. An end-of-bar high and low can describe the range without identifying the path after a valid entry. A later amended data set may not reproduce the information visible at the decision.
Keep any approximation labelled. Missing observations should not be interpolated into favorable trades merely to complete a result table. An unresolved case is a legitimate research output.
Worked OHLC case
Invented teaching bar: open 100, high 104, low 97, close 102. Suppose the declared entry is near 100, stop 98 and target 103. The bar alone does not establish which exit occurred first after an eligible entry. This is not a real provider observation.
Do not award the favorable sequence. State the ambiguity policy and ask what additional evidence could resolve it. Actual broker records, finer observations and simulated order assumptions answer different questions.
Keep missed and unfilled instructions
Messages received after expiry or outside the monitored session should remain in the instruction inventory. They need not be executed trades, but they matter to the product's operational usability. Their exclusion from a trade result requires a declared population rule.
Preserve duplicate, cancelled and incomplete objects too. A replay that retains only easily reconstructable winners changes its denominator after seeing the outcomes.
Separate result layers
A raw instruction percentage, a simulated cash result and a funded account return are different layers. Record sizes, simultaneous exposure and cost assumptions only where evidence supports them. Do not convert isolated percentage additions into a portfolio growth curve.
Changing assumptions should produce another version of the replay, not an invisible correction to the old one. Software correctness and an orderly chronology do not independently validate a strategy.
Deliver an inspectable replay
Retain instruction version, information availability, source, resolution, eligible window, ambiguity policy and output class. The blank replay worksheet supports these fields; it contains no measured provider results.
No paid feed or broker execution was tested to write this chapter. It is a reconstruction procedure designed to prevent hindsight from supplying information the original decision lacked.
Causal availability is more than sorting rows
Putting timestamps in order is necessary for a chronology, but does not show when the underlying input was published or received. A data series can contain later corrections or values formed over an interval. Keep the availability convention and source revision policy with the replay rather than assuming that chronological formatting eliminates hindsight.
Do not patch unknown prices favorably
An unavailable quote or gap can remain unresolved under the frozen method. A best hypothetical research fill has its own established observation window; a subscriber study needs actual account evidence. Neither permits looking beyond the declared boundary simply because a later price would complete the trade. Preserve the policy identity when comparing versions.
Retain a replay bundle
The bundle should contain the instruction version, source reference, data resolution, available-input convention, fill-policy description and ambiguity rule. Keep the output class attached when a summary is shared. A plot without this bundle can illustrate a path, but cannot demonstrate what an earlier subscriber knew or executed.
Evidence Packet
| Evidence object | Record to retain |
|---|---|
| Information boundary | Inputs available by the declared decision |
| Fill convention | Frozen research policy or separate subscriber record |
| Output class | Hypothetical, simulated, actual or unresolved |
Related Session Questions
Primary References
The sources support documented mechanisms and source definitions. Our procedure and teaching examples are not completed paid-feed, latency or account tests.