Crypto exposure adjusted returns divide a return measure by a stated exposure fraction, but the denominator may mean time in the market or capital deployed. Report whole-account return first, then define the adjustment, including idle cash, overlapping positions and sampling rules. A return divided by time exposure is a comparison diagnostic; it does not show what continuous investment would have earned or correct deposits and withdrawals.
A strategy that earns 6% while holding positions for half a month has not demonstrated a 12% monthly investment return. The division can describe how a result relates to participation, but it cannot create extra signals, fills or capital. Before comparing two reports, establish what each exposure field actually counts.
Two exposure denominators answer different questions
Time exposure measures the fraction of the evaluation window during which at least one qualifying position exists. For equal intervals, define an indicator I(t) as one when any position is open and zero otherwise:
U = sum(I(t)) / number of intervals
For irregular intervals, weight each indicator by its elapsed duration. With fill timestamps, measure the total duration of the union of holding intervals and divide by the full evaluation duration. State the position threshold and interval boundaries; a tiny residual holding can otherwise count as continuous participation.
This is an implementation choice, not a universal industry definition. The primary Backtesting.py statistics implementation sets occupied entry-through-exit bars from the supplied trade records to one and averages the resulting array. Its source explicitly measures bars rather than timestamp duration, and includes both boundary bars. Check whether ongoing trades entered those records before interpreting the result. That can differ from elapsed fill-to-fill time, especially for brief holdings or irregular data. Record the software version and convention used.
Capital exposure asks how much was deployed while those positions existed. For a disclosed gross-notional convention, define g(t) as the sum of absolute marked position notionals divided by positive account equity. Its duration-weighted average is G. A position using 10% of equity and one using 90% produce the same binary time indicator but different capital exposure.
AmiBroker's report documentation defines its exposure using bar-level position value relative to portfolio equity, then describes return measures divided by that exposure. Thus a familiar exposure label can refer to a capital measure rather than binary time. Do not transfer either package's definition to another report without checking its calculation.
Overlapping holdings: count the union once
The following original illustration covers exactly 30 consecutive 24-hour days. Day zero is the start timestamp; every interval includes its start and excludes its end. Holdings begin and end at the stated fill times. These are fictional inputs, not market history, a trading recommendation or KLINEO results.
BTC is held from day 0 to day 8, ETH from day 5 to day 12, and SOL from day 20 to day 23. For illustration, the marked ledger also supplies the interval-average gross notional/equity ratios below; these ratios are stipulated inputs, not quantities inferred from the holding dates.
| Day interval | Duration | Open holdings | Any-position indicator | Average gross notional/equity |
|---|---|---|---|---|
| 0–5 | 5 days | BTC | 1 | 40% |
| 5–8 | 3 days | BTC and ETH | 1 | 70% |
| 8–12 | 4 days | ETH | 1 | 30% |
| 12–20 | 8 days | None | 0 | 0% |
| 20–23 | 3 days | SOL | 1 | 20% |
| 23–30 | 7 days | None | 0 | 0% |
Adding the individual holding durations gives 8 + 7 + 3 = 18 asset-days. That sum counts the three-day BTC/ETH overlap twice. It is useful as an asset-duration quantity, but dividing it by 30 does not give the fraction of time the account held any position.
The union contains days 0–12 and 20–23:
U = (12 + 3) / 30 = 0.50 = 50%
The average capital exposure is a separate calculation:
G = (5 × 0.40 + 3 × 0.70 + 4 × 0.30 + 3 × 0.20) / 30
G = 5.90 / 30 = 0.196666… = 19.67%
The account held something for half the window, while its average gross position size was about one fifth of equity. Neither figure contradicts the other.
Keep the account return above the adjustment
Assume the fictional account starts at $10,000 and ends at $10,600, with no outside deposits or withdrawals. Ending equity includes all marked holdings and liabilities after modeled trading costs. Assume idle cash earns nothing and there are no other income flows. The whole-account period return is:
R = 10,600 / 10,000 − 1 = 6%
Under an explicitly chosen time adjustment, R / U = 6% / 0.50 = 12%. Under the separate gross-capital adjustment, R / G = 6% / 0.196666… ≈ 30.51%. Use exposure fractions in the division: 50% means 0.50, not 50.
Publish these as “period return divided by time exposure” and “period return divided by average gross exposure.” The investor's account still gained $600 and 6%. Neither adjusted figure is an earned monthly return, an annualized return, or a return on a separately funded portfolio.
Continuous deployment would need additional entry opportunities, executable prices, available liquidity and a changed sequence of capital commitments. Dividing by exposure does not simulate those conditions. It also does not measure volatility, drawdown or loss severity. Keep the benchmark comparison on the same account boundary and dates, and use the annualized return guide for the separate calendar-time calculation.
A favorable adjusted result may also come from one brief outlier. The median versus mean trade-return comparison helps identify that concentration; low participation alone does not establish efficient or repeatable decisions.
Idle cash remains inside the economic result
If $8,000 remains available to support a strategy that deploys $2,000, removing the reserve after seeing the result changes the capital boundary. A reserve may be required for future entries, margin or operating constraints. Document it before comparing returns.
The GIPS handbook's cash-return discussion treats cash allocation and cash income as part of portfolio performance within its standards. That is a useful accounting reference, not a claim of GIPS compliance here. For a crypto account, value idle assets consistently in the reporting currency; a stablecoin holding is not automatically risk-free dollar cash.
When idle balances earn income, dividing the total account return by a small trading-exposure fraction also scales that income. Retain the whole-account result and separately reconcile trading outcomes, costs and idle-asset income before interpreting the quotient. Do not subtract a guessed cash yield and call the remainder observed trading performance.
Outside funding is another question. Depositing money while flat can increase ending equity without producing a gain. The time-weighted versus money-weighted returns guide addresses that cash-flow adjustment. Exposure division cannot repair an incorrectly measured numerator.
Failure cases worth recording
A midnight snapshot can miss a position opened and closed between snapshots. Conversely, counting every touched hourly bar as a full hour can overstate a two-minute holding. Preserve fills and choose either elapsed duration or a disclosed bar convention. Missing observations should be investigated, rather than silently classified as flat.
A matched long and short can have near-zero net notional while still carrying substantial gross positions, execution costs and basis risk. Do not use a near-zero net denominator as though the account took no risk. For derivatives, margin posted is also different from contract notional; state which quantity the report uses. Gross exposure is a descriptive allocation measure and does not make different instruments equally risky.
If U = 0, return divided by time exposure is undefined; if G = 0, the gross-exposure quotient is undefined too. Report the affected ratio as unavailable and explain any remaining account change, such as income, fees or valuation changes. Very small positive exposure makes either quotient unstable. If equity is zero or negative, the ordinary notional/equity convention also needs an exception rather than a reassuring percentage.
Finally, an exposure adjustment cannot validate a strategy selected after inspecting the same historical results. Where research labels overlap across evaluation periods, the purging and embargo guide addresses that distinct source of contaminated evidence.
A denominator record to attach to the run
Save the full window and time zone; position threshold; fill-to-holding mapping; entry/exit boundary rule; interval weights; gross versus net notional policy; equity valuation and cost treatment; idle-asset income; and outside cash-flow policy. Report R, U and G separately before any quotient, with unavailable values explained. That record makes a 12% diagnostic traceable to a 6% account result instead of leaving readers to infer an investment return.
Method and limits
This AI-assisted educational article combines the linked primary software documentation and performance-accounting guidance with an original fictional holding ledger. The interval union, weighted exposure and return divisions were recomputed. No historical market test, live account or KLINEO exposure-reporting feature was evaluated. These diagnostics describe chosen denominators; they do not establish future returns, comparable risk or an achievable fully invested strategy.