Build a testable observation

A testable observation is a condition someone else can check using the same information you had at the same time. “It looks ready to rise” is not enough. Name the pair and product, price series, timeframe, cutoff, reference area, trigger, expiry and what result will be recorded. You can test an observation without placing an order.

Level 1 explained the market, Level 3 taught chart records, Level 4 showed calculations and Level 5 set cash boundaries. A price-action idea must fit all of those foundations. A rule that can be repeated is not automatically useful or profitable. This lesson first makes the observation measurable; the following lesson connects it to a demo exit and cost-aware checklist.

A supply or demand zone is usually an area inferred from earlier price behaviour. A demand label suggests an earlier upward reaction, and a supply label an earlier downward reaction, under the method being used. There is no universally agreed candle selection or boundary rule. Fix yours before testing. The rectangle is an annotated reference, not a visible store of guaranteed future orders.

This intentionally simple EUR/USD rule uses the full anchor bar’s high/low. Another method might use its body or several bars, but that is a different rule version. Do not switch to a narrower rectangle after seeing which boundary happened to work. Record the anchor-selection method and the creation timestamp as carefully as the boundary prices.

A touch rule can be expressed as interval overlap: a later bar’s low is at or below the zone’s upper edge and its high is at or above the lower edge. That only tells you the bar reached the area. A rejection description might also require a close away from it or a defined wick/body relationship. A close-only rule ignores a wick that returns before the interval ends.

The touch test is not a profitable entry rule. C1 and C2 both intersect the zone, but their closes differ. At C2’s completion, C3 and C4 are unknown. The later C4 does not retroactively erase the fact that the earlier condition occurred; it changes the later status. Keep event records separate from the eventual outcome.

Confirmation is additional evidence that fits an interpretation; it is not proof that the interpretation is true. An entry trigger is the exact condition that starts considering a demo action. You may decide that one completed close is enough, require another condition or record observations only. Each choice changes when the condition becomes available and which cases qualify.

Write exactly when evaluation occurs and how equality is handled. If the current close is 1.1000, this strict-above rule is false. If the right-hand confirming bar for a swing has not completed, that swing cannot be used yet. Multiple signs drawn from the same price data are not automatically independent confirmations.

An idea needs an expiry as well as a price condition. Without expiry, you can keep an old zone until a much later reaction makes it look successful. Choose a fixed count of completed bars, session end or another explicit time limit before collecting outcomes. Define whether a valid trigger is allowed only once per zone or repeatedly after new touches.

For the teaching zone, a close strictly below 1.0980 discards it. A touch below 1.0980 followed by a close above might not discard it under this close-based version; another rule could discard on a wick. That difference must be stated beforehand. Do not draw a replacement zone from the same losing event without labelling its own selection rule and creation time.

A larger timeframe can describe background while a smaller one gives detail. Choose the combination before examining results, such as the last completed daily structure with completed hourly observations. Avoid using the final high, low or close of the current daily candle while assessing an hour earlier in that day. Those final values did not yet exist.

A daily rise and an hourly fall can both describe their selected windows. Define whether that conflict makes you skip, merely annotate or use a different observation. If you inspect many timeframes until one agrees, the successful-looking example is being selected after the result. More views do not erase uncertainty.

A chart condition occurs on the price series you selected. A market order can be considered only after the condition is known; it is filled according to the provider’s execution rules and current price. The bar close used to recognise the condition is not necessarily an executable future entry. Record signal time, order time and fill time separately.

For a conventional long, an entry is bought at ask and its exit is sold at bid; the short direction reverses the transaction sides. A bid-chart close above a level does not automatically imply an ask-side order fills at that close. Pending orders can also activate without waiting for a candle close, so do not describe a buy-stop placed earlier as a close-confirmed entry unless the actual order logic enforces that condition.

Before a simulated order, set limits for acceptable spread, price deviation and delay under your demo policy. If they are exceeded, record a skipped candidate. A clean chart condition can remain true while the execution decision is no. This avoids quietly excluding awkward fills from the results.

An observation test needs a measurable outcome. You might record the next ten completed closes relative to the signal close, whether price reached a stated level within a window, or the complete simulated cash outcome under a specified order model. These are different questions. A ten-bar later close is not the same as a stop/target trade result.

Define outcome handling for incomplete windows, missing quotes and weekends before counting results. If a ten-bar outcome is unavailable at the sample end, label it incomplete instead of silently treating it as a winner or discarding it after reading the result.

Choose the sample dates and instruments before collecting cases. Log all observations that meet the rule, failed candidates, skipped execution decisions and relevant no-signal periods. Keep separate counts: charts examined, conditions triggered, demo actions accepted, and outcomes available. Mixing those denominators can make an ordinary result look more impressive.

A collection of ten memorable rebounds is not a test of how often rebounds occurred. If you marked only the reactions that later looked good, the failed or absent reactions are missing. A screenshot can illustrate an event; it cannot supply an unbiased sample of all eligible events. Keep a simple row-per-event record and retain the original chart cutoff.

Test a rule on unseen dates or use replay with later bars hidden. Have a second learner apply the same written rule. Differences reveal unclear inputs, ambiguous boundaries or availability mistakes. Clarify the rule, label a new version and reserve later observations for checking it. Do not rewrite the earlier version’s failures as if the clearer rule had always existed.

Backtesting examines historical observations; forward demo testing uses later conditions as they arrive. Neither guarantees live results. Choosing settings from the same sample used to advertise the result is an overfitting risk. Level 9 will examine strategy evidence in more depth. Here the basic requirement is that the observation can be evaluated without inventing its rules after seeing the outcome.

Once a candidate is identified, it still needs an invalidation, a possible exit, permitted size, conversion, costs and total account-risk check. Define these in a separate demo execution worksheet. A trigger can be recorded even if the worksheet says no trade. Do not increase size or move the invalidation merely because the chart observation seems convincing.

Read both the observation record and the cash worksheet. If the minimum size exceeds the budget, the quote is unavailable, or your weekly gate blocks new activity, the demo decision is to skip. Keep the candidate in the observation dataset with the reason for the execution skip. This makes the limitations reviewable rather than invisible.

Read the card to someone who has not seen the chart. If they must ask “which candle?” or “how far above?” the observation is still incomplete. Complete the definition before examining the later path. A compact written card can hold a detailed rule without needing a collection of vague labels.