Arithmetic vs Geometric Returns: Why Averages Mislead

A trading report can show a positive average monthly return while the account finishes below its starting value. No missing trade is required. The average may be adding percentages while the account is multiplying growth factors.

The arithmetic mean describes the average of the period returns. The geometric mean describes the constant compounded rate that would produce the same ending value. Before choosing an automated trading strategy from a performance table, find out which number the table calls “average return.”

In an AlgoWay workflow connecting TradingView alerts to execution, the relevant audit question is whether the backtest and the trading account use comparable sizing and return measurements. A sequence of signals alone does not specify how much account equity is exposed to each trade.

A positive average, a smaller account

Consider an invented two-month example with no deposits or withdrawals. Returns below are net account returns, measured against equity at the start of each month. They are not observed strategy results.

The arithmetic average is (30% - 25%) / 2 = 2.5% per month. Yet the total return is 1.30 × 0.75 - 1 = -2.5%. The second month's loss is $3,250 because its denominator is $13,000. It is not a $2,500 subtraction from the original capital.

The geometric monthly mean is (1.30 × 0.75)^(1/2) - 1, approximately -1.258% per month. Applying that constant rate twice takes $10,000 to $9,750, apart from rounding. It does not claim that either actual month earned that rate.

Calculate with growth factors, not signed percentages

For equally spaced observations, add the returns and divide by their count to obtain the arithmetic mean. For the geometric mean, convert each return to a growth factor, multiply the factors, take the nth root, then subtract one:

Geometric mean = [(1 + r1) × (1 + r2) × ... × (1 + rn)]^(1/n) - 1

Enter decimal returns: a 3% gain is 0.03 and its growth factor is 1.03. Do not multiply the signed returns themselves. OpenStax's finance text explains both averages and their different interpretations. With positive growth factors, the geometric mean cannot exceed the arithmetic mean; they are equal when all period returns are identical.

Keep the period label. A geometric mean calculated from monthly observations is a monthly rate. Its compounded annual equivalent is (1 + monthly rate)^12 - 1, not twelve times the monthly rate. Annualizing two unusually strong months produces a mathematical equivalent, not evidence that the next ten months will repeat them.

The ordinary calculation assumes positive equity and period returns above -100%. A complete loss takes the compounded product to zero; subsequent percentage returns from zero equity are undefined without new capital. Negative equity requires a different analysis rather than forcing it through the usual growth calculation.

Compounding is an accounting relationship, not a fee

The gap between the averages is often called volatility drag. It is not a separate charge that a broker deducts. A CFA Institute discussion of the two means emphasizes the mathematical relationship and the danger of attaching a causal story to a label.

In the two-month example, a hypothetical steady +2.5% each month would finish at $10,506.25. It has the same arithmetic mean as +30%, then -25%, but a different compounded result. That comparison holds the arithmetic mean fixed; it does not establish that reducing a real strategy's volatility will preserve its average return or improve its profitability.

Reversing the original two returns gives the same endpoint: 10,000 × 0.75 × 1.30 = 9,750. The intermediate equity is different. A drawdown limit, a withdrawal or a sizing rule can make that path matter in an actual account. The drawdown and recovery calculation answers a related question about the fall from a peak, not the average growth across periods.

Account returns are not average trade returns

A trade that earns 5% on its allocated capital does not necessarily add 5% to total account equity. For example, if $2,000 of a $10,000 unleveraged account is invested and gains 5%, the $100 profit adds 1% to the account, assuming the rest is unchanged and costs are already included.

Do not chain together overlapping trade returns as if every trade used the whole account in sequence. Build the periodic equity series instead. Use the same valuation times and include open-position gains and losses consistently. Otherwise the calculation may hide losses merely because positions remain open.

Fixed-quantity and equity-based sizing can also generate different return series from the same entries and exits. Compounding the measured account returns is still valid without cash flows; that identity does not mean the strategy automatically reinvests profits by increasing its order size. Verify the sizing model before treating backtest growth as an execution plan.

Reconcile the return series before trusting the headline

For a period with no external cash flow, calculate the account return as ending equity divided by beginning equity, minus one. As a check, the product of all the period growth factors should equal final equity divided by initial equity. Use unrounded figures for this reconciliation.

A deposit breaks that simple interpretation: moving from $10,000 to $11,000 solely because of a $1,000 deposit is not a 10% trading return. As described in the GIPS handbook discussion of time-weighted returns, for a time-weighted calculation, value the account around external flows, calculate the flow-free subperiod returns and link their growth factors. If the question is the return on the investor's dated deposits and withdrawals, that is a money-weighted measurement question instead. Do not label a raw balance ratio as trading performance.

In a strategy comparison, retain the start and end dates, periodic equity, cash flows, cost treatment and sizing rule alongside both averages. Then inspect drawdown and the sample's length separately. A perfectly calculated geometric mean can summarize a short, lucky or badly modeled backtest with impressive precision. Arithmetic will check the bookkeeping; it will not certify the strategy.