Is volatility drag an extra cost deducted from my account?

No. It describes a difference between arithmetic and geometric averages of the same returns. Fees, borrowing costs and slippage are separate inputs. Subtracting drag again from an already compounded equity result double counts the effect.

Does reversing the returns change volatility drag?

Reversing the same set of account returns leaves both averages and their product unchanged when there are no external cash flows. Deposits, withdrawals, liquidation or rules that change future exposure can make the sequence matter to final wealth.

Can a strategy have positive average returns and lose money?

Yes. A hypothetical +50% year followed by a -40% year has a +5% arithmetic average but a total return of 1.50 × 0.60 - 1 = -10%. Its geometric annual return is sqrt(0.90) - 1 = -5.13%.

Is half the variance always the exact volatility drag?

No. It is a small-return approximation in consistent units. For +80% and -80%, half the variance implies a -32% geometric mean while the exact value is -40%. Use the actual growth factors when they are available.

TradingView Backtesting

Volatility Drag in Backtests: Arithmetic vs Geometric Returns

Volatility drag describes the gap between the arithmetic average of periodic returns and their geometric, or compound, average. A 20% gain followed by a 20% loss has a 0% arithmetic average but leaves an account 4% lower. The percentages apply to different balances. This gap is a mathematical property of compounding, not a separate fee to subtract from an equity curve that already includes those returns.

See the decision guide

Calculate geometric vs arithmetic return from the same observations

For n consecutive, equally spaced account returns r₁ through rₙ, write percentages as decimals: 20% is 0.20. Assume positive equity throughout, no external cash flows, and returns measured after the costs you intend to include. The formulas below describe observed returns; they do not estimate a future expected return.

With positive growth factors, g cannot exceed a. They are equal when every periodic return is identical.
Trade percentages are usually measured against position value. Overlapping trades, different position sizes and idle cash prevent them from being chained as whole-account returns.
A geometric mean per month is not a yearly rate. Unequal time intervals also require a separate elapsed-time calculation.

Measure	Formula	Interpretation
Arithmetic mean a	(r₁ + … + rₙ) / n	Average simple return per observation period
Cumulative return R	(1 + r₁) × … × (1 + rₙ) - 1	Total growth over all n periods
Geometric mean g	(1 + R)^(1/n) - 1	Constant periodic return with the same ending value
Mean log return	[ln(1 + r₁) + … + ln(1 + rₙ)] / n	Equals ln(1 + g), not g itself
Arithmetic/geometric gap	a - g	Difference in percentage points per period

Equal average returns can produce different account growth

Each hypothetical path starts at $10,000 and has two one-year periods, retained gains, no deposits or withdrawals and no additional costs. In these examples the geometric mean is also the two-year CAGR. Amounts and percentages are rounded to two decimals.

For +20% then -20%: $10,000 × 1.20 = $12,000; $12,000 × 0.80 = $9,600. The first gain is $2,000 and the later loss is $2,400.
The annual geometric mean is sqrt(0.96) - 1 = -2.0204%. The annual mean gap is 2.0204 percentage points; the cumulative loss is 4%. These are different quantities.
For a symmetric pair +x and -x, the total growth factor is (1 + x)(1 - x) = 1 - x². After a 20% loss, a 25% gain is needed to recover: 0.80 × 1.25 = 1.
The +30%, -10% path still makes money. A positive arithmetic/geometric gap does not imply a negative compound return.

Annual returns	Arithmetic mean	Ending equity	Total return	Geometric mean
0%, 0%	0.00%	$10,000	0.00%	0.00%
+20%, -20%	0.00%	$9,600	-4.00%	-2.02%
+80%, -80%	0.00%	$3,600	-64.00%	-40.00%
+10%, +10%	10.00%	$12,100	21.00%	10.00%
+30%, -10%	10.00%	$11,700	17.00%	8.17%

When the half-variance approximation is useful

The common shortcut is g ≈ a - σ²/2, with a and σ expressed as decimals over the same period. For this observed-series comparison, σ² = Σ(rᵢ - a)² / n uses the population denominator n. It approximates compounding when simple returns are small and higher-order terms are negligible. Large losses, skewed returns and large mean returns can make it inaccurate.

The Taylor expansion ln(1 + r) ≈ r - r²/2 gives mean log return ≈ a - (σ² + a²)/2. Converting back with exp and retaining terms through second order gives the familiar g ≈ a - σ²/2. Treating the shortcut as exact drops terms that matter near a total loss.
Do not mix an annual mean with monthly variance, enter 20 instead of 0.20, or square an already calculated variance.
A sample variance with denominator n - 1 estimates a population quantity for a different purpose. It does not make this finite-series approximation exact.
When every return is identical, calculate g exactly: it equals a. When the full series is available, calculate the product or sum of logs instead of relying on the shortcut.

Return pair	Variance σ²	Approximate g	Exact g
+2%, -2%	0.0004	-0.0200%	-0.020002%
+20%, -20%	0.04	-2.00%	-2.0204%
+80%, -80%	0.64	-32.00%	-40.00%

Return order matters differently when cash enters or leaves

Reordering the same fully compounded account returns does not change the final product when there are no external flows: 1.20 × 0.80 equals 0.80 × 1.20. The interim balances and drawdowns differ. With cash flows, the same percentage return can apply to a different dollar amount, so the order can change final wealth.

Both paths have a -4% cumulative time-weighted return: split at the deposit and link 1.20 and 0.80. The $4,000 difference in final wealth comes from the added capital experiencing a different second-year return.
Ending equity divided by starting equity would misclassify the deposit as investment performance. Subtracting the deposit alone still ignores how long it was invested.
Time-weighted return isolates the linked investment returns. Money-weighted return uses the amounts and dates of cash flows to describe the investor experience. Keep a dated cash ledger and valuations at the flow boundaries.
Withdrawals create the same need to track timing. Margin calls, changes in position sizing and forced liquidation can also make a real trading strategy depend on the sequence of returns.

Hypothetical path	Balance after year 1	Deposit at year 1 end	Balance after year 2
Start $10,000; +20%, then -20%	$12,000	$10,000	$17,600
Start $10,000; -20%, then +20%	$8,000	$10,000	$21,600

Daily leverage changes the factors you compound

A daily reset exposure targeting L times the underlying daily return compounds factors of 1 + Lr each day, before financing, fees and tracking differences. It does not generally deliver L times the underlying cumulative return. The SEC investor bulletin explains this distinction for leveraged and inverse ETFs.

These are hypothetical arithmetic examples, not ETF performance forecasts. The trending path also shows why daily compounding does not always mean underperforming the stated multiple over a longer period.
Under the small-return approximation and ideal constant daily leverage, a becomes La and variance becomes L²σ². Approximate growth is La - L²σ²/2 before costs. The variance term grows quadratically, but the approximation fails for large leveraged moves.
If 1 + Lr reaches zero, the idealized account is wiped out. A negative factor is outside ordinary positive-wealth geometric compounding. Real margin rules, liquidation and product terms must be modeled separately.
Borrowing a fixed dollar amount without resetting exposure is a different strategy. Do not apply a daily reset ETF formula to it unchanged.

Two-day underlying path	Unlevered total	Ideal 2× daily total	Twice the unlevered total
+10%, -10%	1.10 × 0.90 - 1 = -1%	1.20 × 0.80 - 1 = -4%	-2%
+10%, +10%	1.10² - 1 = +21%	1.20² - 1 = +44%	+42%

Audit the return series before annualizing it

Save the TradingView symbol, date range, capital, sizing rule and cost settings. Build consistent daily or monthly whole-account returns from equity, with open positions valued on the same basis. Include flat periods. Compute a, g and a - g over that one frequency before making a yearly conversion.

For twelve observed monthly returns, the yearly result is the product of all twelve growth factors minus one. If each month is exactly +1%, that is 1.01^12 - 1 = 12.68%, not 12%. Multiplying an arithmetic monthly mean by twelve does not recover realized annual growth.
Annualizing a short sample extrapolates its pace; it does not create a year of evidence. The Annualized Return guide covers elapsed-time conventions and partial years.
For r = -100%, wealth reaches zero and ln(1 + r) is undefined. Report the total loss explicitly. Returns below -100%, zero starting equity and gaps in the valuation record require separate handling, not a plausible-looking average.
Do not subtract estimated drag from a backtest CAGR. Compounding is already included. Nor does lowering volatility alone establish improved returns: changing exposure also changes the return distribution and costs.

Use Pineify to reconcile supported report metrics

Backtest Deep Report accepts TradingView trade CSV files and calculates supported return and risk metrics. Its current CAGR uses initial capital plus net P&L from closed trades and the interval from first entry to latest exit. Those inputs are not automatically a complete daily, mark-to-market account return series.

Upload the trade export, check parsed trades and initial capital, then reconcile the report with the original test settings. Keep open-position equity and external cash flows separately when they are needed for your calculation.
This page calculates the arithmetic/geometric gap independently. It does not claim that Backtest Deep Report has a built-in volatility-drag statistic or cash-flow-adjusted performance calculation.
For a custom Pine statistic, specify the equity series, sampling frequency, costs and invalid-input behavior to the Pine Script AI Coding Agent. Review the generated code and compare its output with the exact examples before using it on a strategy.

Where Pineify fits

Backtest Deep Report

Import a TradingView trade CSV and inspect supported return and risk results. Keep your volatility-drag calculation separate until its inputs match the report.

Also useful: Pine Script AI Coding Agent. Define a custom account-return statistic and ask for help implementing it in Pine Script. Verify the formula and generated output against a manual calculation.

Explore Backtest Deep Report

Related resources

TradingView Backtesting Guides Annualized Return and CAGR Audit Backtest Accuracy Set a Backtest Date Range

Frequently asked questions

Educational information, not investment advice. All numerical paths are hypothetical calculations. They are not observed backtest or live results and do not predict future returns.

Sources and verification

Last verified: 2026-10-17

Published by Pineify. Pineify is an independent product and is not affiliated with TradingView. Sources support the technical explanations; examples do not establish trading performance.