Last Updated: 2026-10-26
You can inspect a rule with a historical chart and a spreadsheet. An automated strategy report requires a program or configured rule object that actually implements that rule. A similarly named public script is not evidence that your own idea was tested. This walkthrough explains the record to keep and the discrepancies to investigate; it does not report a profitable backtest or promise a free plan's current limits.
Choose the kind of test you need
Scroll sideways to see every column. Keyboard: focus the table region and use the arrow keys.
| Method | Useful for | What it cannot establish alone |
|---|---|---|
| Manual historical replay | Recording decisions without looking at later bars | Complete repeatability if judgment remains undefined |
| Strategy script in a charting tool | Repeating its encoded rules over the loaded data | That the script matches your written rule or live fills |
| Bot-specific simulator | Inspecting the supported bot's parameters and capital use | Testing arbitrary logic outside that bot's model |
| Forward paper run | Observing rule decisions as new data arrives | Real order priority, market impact or account performance |
TradingView's strategy documentation describes hypothetical orders, strategy reports and broker-emulator assumptions. It also warns that non-standard charts can use synthetic prices. Check the current account's data, replay and report limits before paying; this guide does not rank providers or certify a subscription entitlement. The tool does not remove the need to understand the rule.
Write a falsifiable rule before opening the report
Here is an arbitrary teaching specification, not a recommended strategy: on a completed daily standard-price candle, enter a long signal when its close is at or below a 20-period, two-standard-deviation lower Bollinger Band and RSI(14) is below 30. Do not add to an existing position. An exit signal occurs when a completed candle closes at or above the 20-period middle band. The illustrative protective stop is 8% below the assumed entry fill.
That is still incomplete. Record the venue and instrument, timezone, indicator definitions and warm-up history, quantity rule, maximum concurrent positions, next eligible fill time, order types, costs and treatment of ambiguous intrabar events. The 8% value is an example parameter, not a guaranteed loss limit or suitable sizing rule. A signal at a close cannot use later data; a next-bar open is a simulation assumption, not a promised executable price.
If a public script permits repeated entries, uses an unfinished candle, calculates a different band or exits at the signal close, it implements a different experiment. Record the mismatch. Either find an exact implementation, adapt one with competent review, or perform a limited manual test. Do not hide an approximation under the name of your intended strategy.
Resolve the same-bar ambiguity explicitly
Suppose a simulated one-unit long enters at $100 with a $92 stop and a $110 target. The next historical candle has open $100, high $112, low $90 and close $105. Both thresholds were crossed, but those four prices alone do not reveal which was reached first.
One assumed path reaches the target first: +$10 before costs. Another reaches the stop first: −$8 before costs, assuming that stop fill. A gap or delayed execution can be worse. Record the emulator's path rule, obtain appropriate finer-grained evidence where available, or show both scenarios. Do not quietly select the profitable ordering. More intrabar detail does not prove that your real order would have filled at the simulated price.
Audit costs and the trade ledger
Synthetic purchase: 20 units at $100, all sold at $105. Gross profit is $100. At an assumed 0.1% fee on each fill, entry costs $2 and exit costs $2.10, leaving $95.90 before any other costs. The exit fee uses $2,100 turnover, not the $2,000 entry amount. If prices already represent slipped fills, do not subtract the same slippage a second time.
Keep these columns in the trade log: rule version; input availability time; signal time; submission/fill assumptions; quantity; entry and exit fills; fee asset and amount; funding or borrowing where applicable; open-position value; reason for exclusion; and source reference. A spot example does not cover leveraged perpetual funding, liquidation or margin mechanics.
Check several ordinary trades and every unusual state: missing data, a gap through the stop, a partial close, overlapping signals and a position left open at the end. A closed-trade profit report can conceal a large open loss. Keep the original export beside your calculations.
Read metrics together
In an invented 20-trade sample, 12 wins of $15 produce $180 and eight losses of $25 lose $200. Win rate is 60%, profit factor is 0.9 and net result is −$20 before costs. A high win rate is not a profitability test. Profit factor divides gross winning-trade profit by the magnitude of gross losing-trade loss; when the denominator is zero, a displayed infinity does not establish a reliable strategy.
For a mark-to-market equity path that falls from $2,000 to $1,600, drawdown is 20%. Recovering requires 25% from $1,600. State the equity sampling interval and whether unrealized results are included: a closed-trade-only curve may miss the deepest decline. No particular profit factor or drawdown threshold certifies future results.
Separate discovery from evaluation
Choose the historical universe and evaluation window before inspecting results. Include unavailable and delisted instruments where your stated experiment requires them; report missing history. Retain every parameter variant tried. Round-looking parameters do not prevent overfitting, and two years of data do not guarantee adequate independent observations.
Keep a later period out of parameter selection. Respect any overlap in outcome windows and fit transformations only on training data. Compare against an appropriate cash-flow-matched baseline. Report cost sensitivity, concentration in a few trades, sample size and failure cases without automatically declaring an unusual result either skill or luck.
The useful deliverable is a reproducible rule, data provenance, settings, trade ledger, discrepancy log and evaluation result. It may conclude that the evidence is insufficient or the idea should be retired. A passing simulation is not an instruction to connect real capital. For operational failure tests, see the Python system design; for interpreting contributions and benchmarks, use the monthly review example.
Educational simulations, not investment advice or actual trading results. Losses can exceed a model's assumptions. AITR may earn commissions from other site links; no purchase is required for this worksheet.