Forex Effective vs Realised Spread: Why a Negative Figure Is Not a Cash Refund

A forex execution report can show a negative realised spread without paying the customer a refund. The statistic compares a fill with a later reference midpoint. It is not the cash result of buying and then selling currency. In the worked example below, a realised spread of −4 pips accompanies a separate completed trade earning USD 4 after commission. Multiplying that statistic by pip value would produce the wrong account result.

Three questions keep the report useful: which midpoint and timestamp were used, whether the price distance was doubled, and whether the figure describes a fill benchmark or actual cash. Reconstruct those choices before comparing providers or interpreting a negative sign. The illustrative observations are invented, not market prices or a trading signal.

Define the quote, the fill and the unit separately

Use EUR/USD in USD per EUR and define one pip as 0.0001 USD per EUR. A bid of 1.09990 and ask of 1.10010 have midpoint 1.10000 and quoted spread 2 pips. Buying at 1.10010 is a one-pip distance above that midpoint. Under the doubled-distance convention used here, its effective spread is 2 pips. That does not turn one purchase into two cash transactions.

Let P be the actual fill price, M0 the latest valid midpoint from the designated quote source immediately before that fill, and Mh the same source’s midpoint 60 seconds after the fill. Set s=+1 when the client buys EUR and s=−1 when the client sells EUR. These directions refer to the base currency; reversing them reverses the signed calculation.

Our chosen pip definitions are E=2s(P−M0)/0.0001, R=2s(P−Mh)/0.0001 and D=2s(Mh−M0)/0.0001. E is effective spread, R the later-midpoint realised-spread measure and D matched signed midpoint movement. With these identical references, E=R+D. D is an accounting decomposition of prices, not proof that this trade caused the subsequent market move.

A report using one-sided distance, percentages, basis points or another horizon must be normalised first. The 60-second horizon and doubled pip convention are explicit assumptions, not requirements imposed on all forex reports. A relative measure divides by a reference exchange rate; a pip measure does not.

Reproduce three independent fills before taking an average

The following observations occur independently at 10:00, 10:05 and 10:10 UTC. Each has its own immediate pre-fill quote and its own 60-second follow-up. All initial midpoints are 1.10000; assume the same initial bid and ask given above, sufficient available quantity and a matching value date. Observation C receives a fill inside that quoted spread.

For A, the purchase is one pip above M0 but two pips below the later midpoint. Thus E=2, R=−4 and D=6 pips. For B, retain s=−1: the sale is one pip below the initial midpoint and two pips above the later midpoint, giving the same signed results. C instead gives E=1, R=2 and D=−1.

Weight by gross EUR filled quantity: 100000, 50000 and 50000, a total of 200000. The resulting E is 1.75 pips, R −2.50 and D 4.25; 1.75=−2.50+4.25. Do not net purchases against sales to create a smaller denominator. These observations are not a sequence forming one closed position.

This average also differs from how the weighting of quoted forex spreads changes an average. A quote-history average need not contain actual fills at all. An effective-spread average requires fill prices and directions, while a later-midpoint measure additionally requires a defined follow-up price for each observation.

Use an actual exit to calculate account profit

Now consider a separate completion of A: buy EUR 100000 at 1.10010 and actually sell all of it 60 seconds later at the executable bid of 1.10020. A matching ask of 1.10040 gives the assumed later midpoint 1.10030. The gross cash gain is 100000×(1.10020−1.10010)=USD 10.

Assume USD 6 of total posted commission and no financing, tax or currency conversion. Net profit is USD 4. The −4-pip R statistic is neither a USD 40 rebate nor the customer’s realised profit. Its midpoint is not the sale price; its doubled normalisation is not a cash amount.

Likewise, E=2 pips should not be multiplied by this quantity’s USD 10 pip value and booked as USD 20 of one-way midpoint cost. The actual purchase’s one-sided midpoint deviation is one pip, or USD 10. When calculating profit from actual entry and exit prices, the bid–ask effects are already in those prices; do not subtract them again.

Without an actual exit, only a benchmark comparison has been calculated. Nor does R reveal the counterparty’s actual profit: its hedging time, offset price, inventory and fees are unknown. A negative value alone proves neither a rebate nor misconduct.

Change the clock and the statistic can change sign

Keep A’s fill fixed. If the midpoint after 30 seconds is 1.10020, R is −2 and D 4. At 60 seconds, R is −4 and D 6. If the midpoint after 300 seconds is 1.09980, R becomes 6 and D −4. E stays 2 in all three cases. Choosing a later snapshot changes the statistic without recording another execution.

A stale initial reference causes a different problem. Suppose a 09:59:59 UTC midpoint of 1.09980 remains in one data stream although the valid immediate pre-fill midpoint at 10:00:00 is 1.10000. Applying that old value to the unchanged 1.10010 purchase produces E=6 rather than 2 pips. The fill and cash have not changed; the benchmark has.

Match source, timezone, timestamp precision and quote age. Do not take a future quote as the pre-fill reference. Order receipt, order executability and execution are different events; subtracting metrics anchored to different events need not recover the matched D above. A daily reference fixing or a forward quote for another value date is not a substitute for the stated spot quote stream.

Know when a comparison or identity stops being valid

The identity requires the same observations, initial references, horizon, directional signs, units and weights in all three averages. If a later quote is missing, mark it missing. Do not enter zero as R, silently carry forward an old quote or compare its reduced sample with the complete E sample. Publish coverage counts alongside any comparable subset.

Keep signed observations when the definition calls for them. Taking absolute values removes price-improvement information and changes the measure. A negative E may reflect a fill better than the chosen midpoint, but stale data, wrong direction or a mismatched quote can create the same appearance. Inspect the underlying observations before interpreting the sign.

A filled-only sample says nothing by itself about rejected requests or unfilled currency requirements. Nor can a midpoint comparison prove the reference quote was executable for the customer’s size, credit arrangement or product. Different horizons and currency pairs should not be pooled merely because every output is called spread.

Audit the report and the cash ledger in two linked passes

First obtain the exact method: currency quotation, pip unit, client direction, quote source, initial event, follow-up horizon, doubling factor, weighting and missing-data policy. Retain execution IDs, quantities, timestamps and the quote records used. If those fields cannot be obtained, the displayed statistic cannot yet be reproduced.

Then calculate E, R and D for each eligible fill, check their identity and aggregate the same gross quantities. Investigate a mismatch before rounding: different samples, clocks or units are more informative explanations than a provider ranking built from incompatible averages.

In a separate cash pass, match actual entry and exit quantities and prices, commissions, financing and any conversion to statement entries. Only that pass establishes the completed trade’s net account result. Keep the two outputs side by side: benchmark distance diagnoses a price comparison; the cash ledger records what was paid or received.