Write a strategy that can be checked

A trading strategy is a set of rules describing what to observe, when an action is eligible, how exposure is limited and how the action ends. “Buy when it looks good” is an opinion, not a specification someone else can apply. A useful strategy document tells another learner how to classify a new date using the same available information.

The document is a hypothesis: you think particular conditions may produce a useful result under stated costs and constraints. Writing the conditions clearly makes them testable; it does not establish a profitable edge. Keep observation-only work, simulated historical results and demo order rehearsal distinct. None requires you to deposit money or open a live position.

This lesson uses invented rules and prices to teach specification, not to recommend a breakout or moving-average method. The next two lessons examine how to test such a document and describe its results honestly. If a term is unclear, clarify it before treating a chart as an eligible trade.

Scalping usually describes very short intraday holding periods seeking small movements. Day trading generally closes exposure within the defined day or session. Swing trading may hold across several days, and position trading may hold longer. These labels vary between educators; define the actual holding and session rules rather than relying only on the name.

Screen time is not the same as holding time. A daily signal can create a position needing handling while the learner is unavailable. Specify who or what monitors required conditions, what instructions exist on the platform and what happens when the connection fails. Automated orders have product-specific limitations and do not remove all gaps or operational risk.

There is no ladder in which a shorter or longer style automatically means higher skill or profit. Choose a scope that you can observe and document, then investigate its costs and risks. If it does not fit life, an observation-only project or no active trading can be a reasonable course outcome.

Name the pair, product, provider, date range and eligibility window before searching for attractive examples. EUR/CAD, EUR/USD and USD/JPY have different quote currencies and may have different spreads, activity and product terms. A chart from one provider is not automatically an execution record for another. Specify the universe of observations from which candidates may be selected.

State whether the study includes only long candidates, both directions or observation events without orders. Define permitted weekdays or sessions, how holidays and missing observations are handled and whether a position already open blocks another entry. Avoid switching to the pair that happened to produce the best result after seeing the evaluation data.

Record the timeframe, session timezone, completed-bar rule, source and price side. A bid candle, ask candle and midpoint candle are different inputs. A daily bar’s boundary may differ across providers. Specify the latest information that exists when a condition is checked, and do not use the later completed close of a bar still forming.

A historical chart may have revised data, missing quotes or retrospective markers. If the rule uses an indicator, state its settings, input field, smoothing, warm-up and when its value is available. A confirmed swing may require later bars. The chart timestamp and the time when a signal becomes usable are not always the same.

For the classroom example below, the source is a named teaching series of completed daily bid bars. Five earlier completed highs define the reference. The sixth bar’s completed bid close is checked only after it exists. A later bid/ask snapshot illustrates a possible entry comparison. All prices are invented; the snapshot cannot prove available liquidity or a real fill.

A setup describes context; a trigger is the precise condition that completes eligibility. Here the reference is the highest high among exactly five completed daily bid bars immediately before the current bar. The current bar is excluded from that lookback. The trigger asks whether its completed bid close is strictly greater than the stored reference.

The maximum is 1.4620. Store the five observations and the reference before classifying D6. If D6 later closes at 1.4630, the strict comparison is true. If it closes at exactly 1.4620, it is false. A rule allowing equality or using the current high would be a different rule version.

A complete rule specifies when the setup starts, how long its trigger remains usable and when it expires. A close-based trigger cannot be acted on earlier in the same bar. “Breakout” should not mean either a wick or a completed close depending on which later result looks better. Fix the crossing, equality and confirmation rules beforehand.

In this teaching specification, D6’s qualifying close produces one candidate for the first recorded bid/ask snapshot after that close. If the required snapshot is missing or an entry gate fails, the candidate is skipped; no later quote is substituted to make it work. This is a deliberately narrow research rule, not a claim that one-snapshot validity is best for trading.

If an earlier position is still open, the stated one-position rule blocks a new entry. A later re-entry needs a fresh candidate under the same definition; a previous gain or loss does not authorise it. Specify whether the reference rolls on the next completed bar and how the study avoids duplicating the same event in several records.

The entry rule checks a quote available after the completed signal. An order type then determines the instruction sent to the product or platform. For a conventional long, the ask is the relevant purchase side. A market order requests execution at available prices; the observed ask is not a promise that it will be filled there. A limit order adds a price constraint but may remain unfilled.

The gates pass for this snapshot. For a simple historical calculation, the study may explicitly assume entry at 1.4632 and label the result simulated. For a demo market-order rehearsal, record the actual accepted status and fill; it may differ. If the intention is a price-capped limit order, write its price, lifetime and unfilled/cancellation handling as a separate execution model.

A quote can change between the check and submission. Do not describe a pre-order ceiling check as a guarantee of the final market-order price. If the simulation lacks size/liquidity or sequence data, state that limitation. Changing order type after viewing outcomes blends different strategies and execution assumptions.

State the invalidation idea, protective order instruction and any time exit separately. A chart-close invalidation can occur at a different time from an intrabar stop trigger. For the teaching long entered at ask 1.4632, a hypothetical bid-side stop level of 1.4612 is 20 conventional pips away, and a bid-side target of 1.4672 is 40 pips away.

Define which event exits first: stop, target, maximum holding period or another specified condition. Our teaching example ends at the first applicable stop/target event, or at the bid close of the third completed daily bar after the entry if neither event is identified sooner. A bar touching both levels without sequence information is ambiguous and handled under the disclosed test policy in the next lesson.

Ordinary stop orders can fill worse than the trigger price. Limit exits may not fill merely because a chart touches their price. A stop-limit can remain unfilled beyond its limit. Trailing or partial-close rules require additional definitions about activation, quote side, size and fees. Do not insert them into a losing historical case just to improve its score.

Size belongs in the specification before the result. Name the account currency, balance/equity reference, budget, stop distance, contract, account conversion, volume increment and minimum. Then include charges, financing and worse-fill scenarios consistently. A planned price-risk calculation is not a guaranteed maximum loss.

If an invented separate C$0.50 fee is charged on that completed case, a stop filled exactly at 1.4612 makes the net loss C$10.50, before any other costs. If the same exit fills at 1.4607, the 25-pip price loss is C$12.50 and the separate fee makes C$13. The worksheet must state which scenario its gate actually checks.

Several positions can share a currency, and pending orders can add exposure. A per-case calculation does not replace daily/weekly gates or a combined account worksheet. If rounding or minimum size makes the action exceed the stated gate, record no entry under the policy rather than changing the budget until the platform allows it.

A filter excludes circumstances outside the intended scope, such as a spread above a declared threshold, an unavailable quote or a specified scheduled-news window. Write the measurement, boundary, timing and action for each filter. “Avoid bad conditions” is too vague for another learner to reproduce.

A news window is a predeclared teaching-policy choice, not a universally safe interval. Unscheduled events remain possible. A spread filter can discard an otherwise valid chart candidate, so keep both candidate eligibility and the execution skip reason. Do not erase the event from the denominator simply because the quote gate failed.

Each additional filter is another decision that might be fitted to old data. Keep the number of settings and tried variants in the development log. A simpler rule with documented limitations is easier to evaluate than a complicated rule whose exclusions were chosen after every loss.

This card joins the conditions into one document. It still needs the named data file, date range, calendar-filter choice and exact account-gate definitions before a complete study can run. Those inputs belong beside the card; do not leave them to be chosen after seeing results. A compact card and an attached assumptions sheet can work together.

Another example can use an invented five-period and twenty-period moving-average cross on completed EUR/USD daily closes. It must define the average type, warm-up, crossing/equality rule, next available entry, exits, costs and gates just as carefully. An indicator name alone does not replace the specification. Neither example is an endorsement.

Give the rule and a dated input set to another learner without showing the results. Compare their eligible dates and skip reasons with yours. A disagreement may reveal an unclear equality rule, timezone, warm-up period, quote side or handling of missing data. Correct definitions before drawing conclusions from a score.

Keep a version log with the change, reason, date and information already viewed. A factual correction and a new trading rule are different. If you change a filter after seeing July, that July sample has influenced development of the new version. Do not report it as untouched evidence for that version.

The first milestone is reproducibility: another person can apply the instructions and account for unresolved cases. Reproducibility does not imply profitability. The next lesson adds a fair test sequence, including a reserved period and new demo observations, so a clear recipe is not mistaken for a proven method.