How much trading friction can your gold backtest actually survive?

The important missing denominator

A strategy can generate an attractive gross result by trading frequently for small gains. Another can reach the same total with fewer, larger moves. Their capacity to absorb transaction costs is different. My starting question is therefore how much friction the recorded activity can bear before its gross surplus disappears. This is an accounting question before it becomes a prediction about implementation.

TradingView’s current strategy documentation provides configurable commissions and fixed-tick slippage, while describing the resulting trades as simulated. It also notes that actual slippage varies with market conditions. Those features allow useful sensitivity checks; they do not supply the broker-specific execution history needed to verify a live result.

Source notes: [1] TradingView — Pine Script strategies: simulated trading costs

Define the cost unit before dividing

For a fixed list of completed trades with identical unit size, gross profit divided by the number of round trips gives a break-even total cost per round trip. The arithmetic assumes that all relevant costs can be expressed in the same currency and allocated to those completed trades. It does not apply unchanged to a list with varying sizes, partial exits or unresolved positions.

For unequal sizes, define activity consistently: total contracts traded on each side, total notional turnover, or another unit matching the actual fee schedule. Fees charged per order require an order count; a per-contract fee requires contract quantity. A percentage-of-notional charge changes with notional value. A correct-looking division with the wrong denominator produces a misleading budget.

Start with a frozen-path accounting stress

I would export a versioned trade ledger and hold its entry decisions, exit decisions, quantities and order sequence fixed. Subtract several clearly labelled hypothetical cost assumptions from that same ledger. Preserve gross profit, total friction and net profit separately. The scenarios are analyst choices, not assertions about a particular broker’s current charges.

This first pass reveals sensitivity to a known amount of accounting drag. It also exposes double counting. If simulated fills already include the bid–ask spread, subtracting a second full spread can overstate cost. If the ledger uses midpoint prices, spread may still be absent. Each component needs an explicit statement of where it enters the calculation.

Then identify where the path stops being fixed

A fixed ledger cannot capture every consequence of worse execution. A delayed or missed fill can remove a trade, change its exit, or leave capital unavailable for another order. If position size depends on account equity, additional costs can alter later quantities. Margin constraints may also change which orders are feasible.

Those cases need the strategy and execution model rerun under the altered assumption. Keep this second experiment separate from the initial subtraction exercise. Otherwise a reader cannot tell whether the deterioration came from higher charges on the same activity or from a different sequence of positions. Neither result should silently inherit the other experiment’s label.

Average cost can conceal the dangerous tail

An average friction allowance can look comfortable while a handful of event-driven fills consume most of it. I would therefore examine cost by order type, time window and position size, using group definitions chosen before searching for a favourable result. Genuine execution records should retain adverse and favourable differences from the stated benchmark.

Where such records are unavailable, publish a scenario range and explain its basis. Do not present a guessed spread or fixed tick penalty as a measured distribution. The relevant question is whether the model remains useful across plausible stresses, and what new evidence is needed to narrow those stresses. A polished equity curve cannot answer that question by itself.

What a defensible conclusion looks like

The final research note should identify the gross ledger, denominator, cost units, fill convention, financing treatment and unresolved exposures. It should distinguish an accounting break-even budget from a simulated rerun and from observed broker execution. Readers can then reproduce the arithmetic and challenge the assumptions separately.

This edition contains no gold strategy return, estimated trading advantage or CY66 execution result. Its contribution is a procedure for making future claims more precise. A strategy that survives a chosen cost assumption has passed that particular stress; it still requires independent evaluation of its signal, operational behaviour and evidence quality.