Write a claim that a test can reject
Separate a possible market mechanism from the rule that represents it. For example, the idea that unusually active breakouts attract follow-through is a hypothesis. A volume filter is one measurable proxy, and it can fail even if the explanation sounds plausible. The following research card is illustrative, not a tested strategy.
- This card needs the execution fields filled in before it becomes an executable strategy. Choosing exits after seeing the outcome changes the test.
- A missing or zero volume feed cannot support the example filter. Mark those observations unavailable under a rule chosen in advance.
- Use the same rule definitions after losses. Removing an inconvenient trade after the fact changes the population being studied.
| Research field | Example to freeze before testing |
|---|
| Hypothesis | Adding a volume filter to a fixed breakout rule improves net returns without exceeding the same drawdown budget. |
| Observable condition | On a completed daily candle, close exceeds the highest high of the previous 20 candles. Current volume exceeds the average volume of those same 20 prior candles. |
| Execution specification | Record the exact instrument and venue, session, data adjustments, next-open entry assumption, exit rule, size, maximum concurrent positions, and gap handling before running either version. |
| Control | The identical breakout strategy without the volume condition, over the same calendar dates, instruments, capital, and risk limits. |
| Decision rule | Predeclare the minimum economically useful net improvement, acceptable drawdown, cost stress, uncertainty method, and review date. |
| Version history | Save every attempted filter and parameter choice, including failures. Retuning after viewing validation data starts a new experiment. |
Choose a baseline that answers the claim
A positive result and an improvement over a control answer different questions. Compare the filtered and unfiltered versions on a common calendar and risk budget. Fewer trades can raise average profit per trade while reducing total opportunity or concentrating exposure. Keep both the trade-level and portfolio-level comparisons.
- Report net return, drawdown, exposure, trade count, and the comparison unit. A per-trade mean alone does not measure the return on an account.
- Keep a daily or session-level series for both versions when trades overlap. Their differences can then be studied over the same time blocks.
| Claim | Useful comparison | Remaining limitation |
|---|
| The filter improves the entry rule | Same strategy with the filter removed; inspect the excluded trades as well as the retained trades. | The filtered trades are a selected subset, not an independent control sample. |
| The strategy adds value over market exposure | A stated passive benchmark over the same dates, with comparable capital and an explanation of exposure differences. | Different leverage, holding time, or downside exposure can explain a return difference. |
| Execution discipline improves results | Record planned orders and actual fills for the same eligible signals, including skipped signals. | A behavioral association does not establish that journaling caused higher returns. |
Worked example: a positive gross edge disappears in costs
These are invented outcomes for an arithmetic exercise, not TradingView results. Assume 100 closed trades, constant initial planned risk of $100 per trade, no compounding, 45 gross winners of +1.5R, and 55 gross losers of -1R. One R is the initial planned dollar risk. The base round-trip cost is $10, or 0.10R, on every trade.
- The base case earns only $2.50 per trade above its assumed costs. A $5 increase in round-trip costs reverses the sign; this example fails that cost stress.
- Include commission, spread, slippage, and applicable financing or borrow costs. If spread or slippage is already included in fill prices, do not subtract it twice.
- Actual losses can exceed the planned 1R after gaps or poor fills. The fixed two-outcome exercise omits that risk, and its total R is not an account return percentage.
| Measure | Calculation | Result |
|---|
| Gross total | 45 × 1.5R - 55 × 1R | +12.5R ($1,250) |
| Base total costs | 100 × 0.10R | 10R ($1,000) |
| Base net total | 12.5R - 10R | +2.5R ($250) |
| Base net mean | 2.5R / 100 | +0.025R ($2.50) per trade |
| Cost stress at $15 round trip | 12.5R - 100 × 0.15R | -2.5R (-$250) |
| Break-even average cost | 12.5R / 100 | 0.125R ($12.50) per trade |
A trade count does not establish certainty
The base example has 45 net outcomes of +1.4R and 55 of -1.1R. Its sample standard deviation is 1.25R. Under an illustrative independent, identically distributed sampling assumption, the standard error of the mean is 1.25 / sqrt(100) = 0.125R. A rough normal interval is 0.025 ± 1.96 × 0.125, or -0.220R to +0.270R. It includes zero even under those favorable assumptions.
- This rough interval is an arithmetic illustration, not a valid confidence statement for a strategy selected from many backtests. It also does not estimate future drawdown.
- Several trades in one session or correlated symbols can share the same market shock. One hundred rows can contain far fewer independent observations.
- Keep development and later validation periods separate. Remove cross-boundary trades or apply a declared boundary rule so one position cannot contribute to both samples.
- For dependent observations, use a method that preserves relevant time blocks and document the block choice. No fixed trade count guarantees an edge.
- Bailey and coauthors show why selecting strategies from repeated backtests creates overfitting risk. Retain the number and scope of trials; an attractive final chart hides that search history.
Keep a TradingView validation log
TradingView strategy results use simulated orders. Its official strategy documentation describes the broker emulator, execution assumptions, and selection bias. Before interpreting the result, save the script version and the chart context with the properties used for that run. A screenshot of an equity curve cannot recover these inputs.
- Use standard candles for this exercise. TradingView warns that synthetic chart prices can produce unrealistic strategy results.
- In strategy properties, commission applies to entries and exits; slippage is specified in ticks for market and stop orders. Record the units instead of treating every cost input as a percentage.
- Reconcile a few orders against the chart, including gaps and a bar that touches both a stop and target. Historical bars do not necessarily reveal the actual intrabar order of events.
- Preserve results from unseen data even when they disappoint. Repeatedly inspecting and tuning to a holdout makes it part of development.
| Log before running | Append after running |
|---|
| Run ID, rule version, hypothesis, control, and all variants attempted | Decision: rejected, inconclusive, or eligible for a specified next test |
| Exchange-qualified symbol, interval, session, timezone, and date range | Available observations, missing data, excluded trades, and reasons |
| Capital, position sizing, commission, slippage, and fill settings | Net result, drawdown, exposure, and sensitivity to stated cost changes |
| Development/validation split and fixed review schedule | Both periods reported separately; date and reason for any later revision |
Record invalid evidence and possible decay separately
A broken test and a deteriorating result require different responses. The examples below are hypothetical failure records. They show what to preserve and what conclusion is justified, without diagnosing every loss as a vanished edge.
- Do not move a loss limit after reaching it. A risk limit controls exposure; crossing it does not by itself prove a statistical change in the market.
- Define review dates or an appropriate sequential procedure before monitoring. Repeatedly checking until a result looks significant changes the inference.
| Observation | Assessment | Action to record |
|---|
| A winning signal used the final daily close before that daily bar ended. | Lookahead makes the result invalid as a contemporaneous trading test. | Reject the run, fix information timing, and rerun all versions. Preserve the rejected record. |
| The 100-trade example changes from +2.5R to -2.5R at the declared higher cost. | The claimed advantage does not survive this cost assumption. | Record the cost failure; do not present the gross +12.5R as available profit. |
| A frozen rule previously averaged +0.20R over 200 observations; the next scheduled 50-observation review averages -0.08R. | Possible decay, noise, or a changed execution/data process. The means alone cannot distinguish them. | Compare costs, rule adherence, exposure, and the predeclared uncertainty analysis before drawing a conclusion. |
| The predeclared loss or data-quality limit is breached before the next review. | The operating condition for continuing the test has failed. | Pause under the written rule, retain all observations, and document a separate revision if one is proposed. |
Carry the research record into Pineify Trading Journal
Pineify Trading Journal supports manual trade records, setups, tags, session notes, and closing notes. Use a separate setup name for each frozen rule version, such as Breakout-v1, and keep paper and funded results in separate journals. These are user-maintained labels; they do not make a record an immutable audit trail.
- Review the supported setup performance summaries alongside the underlying trades. Retain the baseline calculation and uncertainty analysis separately where needed.
- This workflow uses manual recording. It does not assume automatic TradingView synchronization, automatic statistical validation, or a journal feature that discovers a profitable strategy.
| Journal field | Research use |
|---|
| Setup description and checklist | Save the hypothesis, rule version, eligibility conditions, and planned risk checks. |
| Trade entry, close records, and fees | Enter actual or simulated fills accurately and reconcile their net result. |
| Tags | Label the version and rule adherence consistently; preserve profitable violations and losing compliant trades. |
| Session notes | Keep the run settings, control summary, skipped opportunities, review date, and rejection reasons. |
| Closing notes | Explain deviations and append the outcome without rewriting the original hypothesis. |
