How to Lock TradingView Strategy Properties for Comparable Backtests

Only results with a fixed premise and data context are comparable

Strategy Properties define the premise of a backtest experiment. Initial capital and currency set the reporting and capital base; order size and pyramiding shape exposure; margin changes capital requirements and margin-call behavior; commission and slippage represent assumed friction. Saving only net profit leaves the experiment impossible to reconstruct.

Save the settings record first, execute Strategy Tester second, then bind the result file and record with the same version ID. A statement such as “version B performed better” is interpretable only when symbol, timeframe, date range, session, script hash, inputs and properties are equal except for the one intended change. A different provider or accessible history is a different experiment too.

Separate capital, exposure, friction and calculation behavior

Treating the Properties panel as one long list hides why a field changed. Grouping fields into capital, exposure, friction and calculation/fill layers makes potential causes traceable. This article records the first three; TV15 covers recalculation timing and TV12 covers the broker emulator and fills.

Interface labels and grouping can change. Prefer the current official page and the values shown in the actual run.

Ten percent of equity and a fixed quantity of 10 are not interchangeable. Likewise, a 100% versus 20% margin setting changes capital requirements; it is not a direct maximum-loss control. Reconcile actual position risk separately with stop distance, quantity, point value, gaps and contract terms.

Bind a settings fingerprint to every exported result

A fingerprint here is a compact human-readable identifier, not a cryptographic proof. The complete record stores value, unit, entry point, rationale, whether code fixed it or the UI overrode it, and review date. The short ID belongs in filenames and comparison tables. Preserve text as well as screenshots so differences remain searchable.

Values illustrate the recording method and are not recommendations.

Reconcile strategy() defaults with the live Properties values

Pine v6 lets a strategy declaration state many default properties. The sample exists only to show explicit settings. pyramiding = 1 records an intent to permit one same-direction entry at a time. Its moving-average cross is a short fictional rule, not evidence of an edge, a recommended length or a directional instruction. The commission and slippage values are fictional too.

//@version=6
strategy(
    "Settings fingerprint demo",
    overlay = true,
    initial_capital = 100000,
    currency = currency.USD,
    default_qty_type = strategy.percent_of_equity,
    default_qty_value = 10,
    pyramiding = 1,
    commission_type = strategy.commission.percent,
    commission_value = 0.05,
    slippage = 2,
    margin_long = 100,
    margin_short = 100
)

int fastLength = input.int(20, "Fast length", minval = 2)
int slowLength = input.int(50, "Slow length", minval = 3)
float fast = ta.sma(close, fastLength)
float slow = ta.sma(close, slowLength)

if ta.crossover(fast, slow) and barstate.isconfirmed
    strategy.entry("L", strategy.long)
if ta.crossunder(fast, slow) and barstate.isconfirmed
    strategy.close("L")

plot(fast, "Fast", color.teal)
plot(slow, "Slow", color.orange)

Source defaults make a shared starting point easier to identify, but a runnable configuration can still differ if the corresponding Properties field is changed. Reconcile source, the live panel and the exported settings information. Assign a new fingerprint whenever an override matters.

Change one variable so a result difference has one named cause

Clone a frozen baseline and change one target field. Changing capital and sizing together makes it impossible to separate effects on amount, percentage and trade eligibility. When testing pyramiding, hold other properties, inputs, symbol and period constant, then retain trade count, exposure and margin events alongside the report.

1. 01Freeze a baseline

   Save script hash, inputs, properties, symbol, timeframe and date range.

2. 02Change one field

   Record the reason, old value, new value, unit, operator and time.

3. 03Rerun the Tester

   Export the same CSV/XLSX report type instead of comparing screenshots alone.

4. 04Reconcile differences

   Check the settings delta, trades, quantity, capital constraints and warnings in order.

5. 05Freeze the version

   Keep failed and unattractive variants so outcome does not rewrite the audit trail.

Varying commission, spread or slippage across scenarios asks a different question. Do not recreate that analysis here; continue to BT09 cost sensitivity and stress testing. TV11 identifies the exact baseline passed into that exercise.

Pass the CSV/XLSX downstream with its premise attached

After exporting a compatible Strategy Tester CSV/XLSX, store it beside the matching fingerprint. SG Group’s Backtest & Robustness Lab can then provide descriptive diagnostics such as KPIs, equity, drawdown, stress and OOS from the loaded results. When saved analyses and repeatable version comparison become important, the paid workflow is a natural next step.

The free TradingView indicator collection is useful for observation and hypothesis formation, but an indicator plot is not a Strategy Tester strategy export. Convert a precise rule into a separate `strategy()` specification, validate data, fills and execution, then send suitable results to the Lab. Do not relabel an attractive plot as tested performance.

Audit missing fields and overrides before sharing a result

• Record the strategy title, script version, Pine version and source hash.
• Record initial capital, currency, order-size type/value, pyramiding and margin by direction.
• Record commission and slippage value, unit and entry point; keep cost modeling in BT09.
• Save symbol ID, provider, timeframe, session, date range and adjustment context.
• Reconcile strategy() defaults against the values used in the Properties panel.
• Change one field at a time and preserve unsuccessful versions.
• Apply one fingerprint ID to the CSV/XLSX, settings record and run log.

Fixed settings do not guarantee future performance. Reproducibility is only the minimum condition for asking the same question again. Validate data quality, fills, execution, sample size, costs, OOS and robustness separately. Continue with TV12 for fill assumptions and TV15 for runtime timing.