sifting/io
Quant Research & Backtesting
15 min readSiftingIO Team

Slippage vs Spread: Calculate Trading Costs in Basis Points

Calculate slippage, spread and fees without double-counting. Worked trade examples, basis-point formulas and an Excel worksheet for execution-cost analysis.

Slippage vs Spread: Calculate Trading Costs in Basis Points

Slippage is the difference between an execution price and a chosen benchmark. The bid–ask spread is the gap between the bid and ask at a particular time. They measure different things. A trade can cross a spread without slipping beyond the displayed ask, and a slippage calculation can already include the cost of crossing that spread.

That distinction matters when a trading dashboard reports costs, a researcher evaluates a backtest, or a team compares an expected price with a broker’s fill. Add the spread to a midpoint-based slippage figure without checking the definition and the same cost can appear twice.

This guide works through a complete example, expresses the result in basis points, and gives Excel formulas you can reproduce. All prices, quantities, fees and timestamps below are invented teaching examples, not observed trades or performance claims.

Try the data workflow: Create a free SiftingIO account and get an API key to explore market observations alongside your own execution records. The arithmetic in this article needs no account. API access and permitted use follow the current plan terms.

Slippage vs spread: the practical difference#

A quote describes a market observation. A fill records what actually happened to an order. Keep those records separate, even when their prices happen to match.

MeasureWhat it comparesWhat you need
Quoted spreadAsk minus bid at one observation timeA compatible bid and ask, with their timestamp and source
Slippage against the ask or bidA buy fill against the earlier ask, or a sell fill against the earlier bidFills and the specified bid or ask
Slippage against the midpointA fill against the earlier bid–ask midpointFills plus both sides of that quote
Explicit feesAmounts charged separately for the tradeBroker, venue or other applicable charge records
Implementation shortfallThe cost of implementing a decision relative to a defined paper portfolioDecision benchmark, intended quantity, fills, explicit costs and treatment of unfilled quantity

The word “slippage” alone leaves the benchmark unspecified. Two reports can show different numbers for the same trade because one uses the ask and the other uses the midpoint. Neither number can be interpreted properly until that choice is visible.

For a closer introduction to the quote itself, see our guide to reading a bid–ask spread. Here, the focus is the next step: reconciling an execution with that observation.

Start with a benchmark you can defend#

For a buy order, the ask is the natural comparison when asking, “Did the fill differ from the price offered at that moment?” The midpoint answers a different question: “How much did execution cost relative to the centre of that quote?” A decision price asks how much changed after the investment decision was made.

BenchmarkUseful questionImportant limitation
Decision-time priceWhat was the cost from the time the decision was made?Requires an actual decision timestamp and a stated price definition
Order-arrival midpointHow did fills compare with a neutral quote reference at arrival?The midpoint may not have been executable
Order-arrival ask for a buy, bid for a sellHow did the execution compare with the relevant quoted side?Available quantity, quote conditions and routing still matter
Last traded priceHow did this fill compare with a recorded transaction?That transaction may be old, elsewhere, or for a different size

Investor.gov’s explanation of order types notes that a market order’s execution price is not guaranteed and that the last traded price need not be the price at which it executes. A limit order specifies a price or better, but that price constraint does not ensure the order will fill.

Choose the benchmark before reviewing the result. Switching from midpoint to ask only when a trade looks expensive makes a comparison unreliable. Save both if they answer useful questions, but give each a separate name.

Worked example: a 500-share buy#

Assume a hypothetical stock has the following quote when an order arrives:

ObservationValue
BidUSD 99.98
AskUSD 100.02
MidpointUSD 100.00
Quoted spreadUSD 0.04
Intended buy quantity500 shares

The spread relative to the midpoint is 0.04 / 100 × 10,000 = 4 basis points. The ask is two basis points above that midpoint. Those are different measurements: four basis points is the full quoted width; two basis points is the distance from midpoint to ask.

The order then receives three fills:

FillSharesPrice in USDConsideration in USD
1100100.0210,002.00
2200100.0520,010.00
3200100.0820,016.00
Total500—50,028.00

The quantity-weighted average fill price is:

50,028 / 500 = USD 100.056

An unweighted average of the three prices would give every fill equal importance even though their sizes differ. Use actual filled quantities. Keep the underlying fills as well as the average so that partial fills, corrections and fee allocation can be audited later.

Against the arrival midpoint, the price cost is:

(100.056 − 100.00) × 500 = USD 28

Against the arrival ask, the price difference is:

(100.056 − 100.02) × 500 = USD 18

The USD 10 difference is the midpoint-to-ask distance multiplied by 500 shares. It has already entered the midpoint-based result.

Express trading costs in basis points#

One basis point is 0.01%, or 0.0001 as a decimal fraction. A cost of six basis points means six units of cost per 10,000 units of the stated benchmark notional. It does not mean 6%.

For the simple positive-price, linear examples in this article, define:

  • s = +1 for a buy and s = −1 for a sell.
  • Q as the positive filled quantity.
  • P as the quantity-weighted average fill price.
  • B as the benchmark price.
  • F as explicit costs in the same currency as Q × B.

The signed price cost is s × (P − B) × Q. Positive means adverse execution relative to that benchmark; negative means improvement. The all-in cost for this completed order is that price cost plus F.

All-in cost in bps = [s × (P − B) × Q + F] / (Q × B) × 10,000

Assume the broker charges a separate USD 5 fee. Using the arrival midpoint, the result reconciles as follows:

ComponentUSDBasis points of USD 50,000
Midpoint to arrival ask10.002.00
Average fill above arrival ask18.003.60
Total price difference from midpoint28.005.60
Explicit fee5.001.00
Total cost against arrival midpoint33.006.60

The first two rows reconcile to the third; they are not additional costs to add to it. The total is USD 28 plus USD 5, not the sum of every row in the table.

For a sell at an average of USD 99.944 against a USD 100 benchmark, the same sign convention gives −1 × (99.944 − 100) × 500 = USD 28. A sale below the benchmark is adverse. Using the buy formula unchanged for a sale would reverse that interpretation.

These formulas need adaptation for contract multipliers, inverse contracts or non-positive prices. For a futures contract, use its actual contract specification to establish notional and tick value. Margin posted is not a substitute for trade notional.

Why adding the spread can double-count costs#

Suppose a backtest buys at a midpoint and then applies a slippage allowance of 5.6 basis points, calibrated from the midpoint-to-fill result above. That allowance already contains the observed two-basis-point half-spread contribution in this example.

Adding another half-spread would produce 7.6 basis points before fees, even though the measured midpoint-to-fill difference was 5.6. Adding the full four-basis-point spread would overstate it further.

There is a legitimate alternative convention: start a simulated buy at the ask, then model an additional execution difference from that ask. Under that convention, the USD 18 residual is the relevant observed comparison. The starting price and the cost model must use compatible definitions.

An immediate buy at the ask followed by a sell at the bid would cross the full spread if the quote remained unchanged, sufficient executable quantity existed, and no other costs or movements intervened. That round-trip illustration is not the cost of every one-way trade and is not a promised fill model.

Price movement is not automatically market impact#

The example’s USD 18 difference above the arrival ask is observable. Its cause is not established by the calculation. It might involve market movement, changing liquidity, order size, routing delay, or several effects together. Calling the whole difference “our order’s market impact” would claim a causal explanation that the records do not provide.

Top-of-book prices also say little about the quantity available beyond the best level. CME Group’s liquidity methodology illustrates the size problem: when a specified quantity cannot be obtained at the first level, deeper prices enter its cost-to-trade calculation. A narrow displayed spread alone does not describe the cost of a large order.

For analysis, retain order size, fill sequence, source and timing. A reference-price snapshot can describe context. Establishing what caused an execution outcome requires more evidence than one before-and-after price comparison.

Arrival cost is not the whole implementation shortfall#

Now suppose the decision to buy those 500 shares was made at USD 99.90, before the order arrived at the USD 100 midpoint. The completed order’s price difference against the decision benchmark is:

(100.056 − 99.90) × 500 = USD 78

That consists of USD 50 between the decision benchmark and the arrival midpoint, plus the USD 28 arrival-to-fill difference already calculated. With the USD 5 fee, the total becomes USD 83.

The decision benchmark notional is 500 × 99.90 = USD 49,950. The cost is therefore 83 / 49,950 × 10,000, or approximately 16.62 basis points. The earlier 6.60-basis-point arrival figure used a different benchmark and denominator. They describe different windows.

CFA Institute’s overview of trading costs describes implementation shortfall as encompassing explicit and implicit costs, including delay and opportunity costs associated with unfilled trades. Our completed-order examples isolate price differences and a fee. An analysis of an incompletely executed decision needs additional records.

If 1,000 shares were intended but only 500 filled, reporting the 500-share average is useful, but it does not settle the full decision’s cost. Record the unfilled amount, cancellation or expiry time, and the evaluation price used for the unfilled portion. State how that opportunity component is measured instead of quietly dropping the unfinished quantity.

Calculate slippage in Excel#

Put the three fill quantities in D2:D4 and their prices in E2:E4. Calculate the weighted average with:

=SUMPRODUCT(D2:D4,E2:E4)/SUM(D2:D4)

For the invented buy example, enter these inputs and formulas in a separate two-column block:

CellMeaningValue or formula
B2Side: buy +1, sell −11
B3Filled shares500
B4Benchmark price100
B5Weighted average fill100.056
B6Explicit fee in USD5
B7Signed price cost in USD=B2*(B5-B4)*B3
B8Total cost in USD=B7+B6
B9Total cost in basis points=B8/(B3*B4)*10000

The outputs are USD 28, USD 33 and 6.60 basis points. Leave B9 as an ordinary number labelled “bps”; applying percentage formatting would display a different scale. Regional Excel settings may require semicolons instead of commas in the weighted-average formula.

Preserve precision until reporting. Rounding the average fill from 100.056 to 100.06 before multiplying by 500 changes the estimated price cost from USD 28 to USD 30. Displaying fewer decimals is different from overwriting the underlying calculation.

Forex: keep pips, basis points and cash separate#

Consider a separate hypothetical purchase of EUR 100,000 against USD. The benchmark is 1.08000 USD per EUR and the fill is 1.08012. Before fees, the price difference costs 100,000 × 0.00012 = USD 12.

With a pip size of 0.0001, the difference is 1.2 pips. Relative to the benchmark notional of USD 108,000, it is approximately 1.11 basis points. The pip count and basis-point figure differ because their denominators differ.

Record whether quantity is expressed in base-currency units, lots or contracts. For an account denominated in another currency, keep the original cost currency and identify the conversion rate and time used for reporting. Our guide to base and quote currencies explains the pair convention.

A quoted gold reference, a broker’s leveraged gold product and an exchange-traded gold futures contract also need not represent identical instruments. Match the instrument and contract terms before comparing their prices. Our spot-versus-futures guide covers that distinction.

Build a cost record that survives review#

A useful trading-cost dataset preserves evidence rather than just the final percentage. Keep the original records and make the calculation reproducible from them.

RecordFields worth retainingCheck before calculation
OrderIdentifier, side, intended quantity, instrument, decision and submission timesDo quantities and contract units match?
BenchmarkPrice, definition, source, event time, receipt timeWas the observation available when the benchmark was chosen?
QuoteBid, ask, source and any available size or conditionsAre both sides comparable and valid?
FillsExecution identifiers, quantities, prices, timestamps, correctionsDo fills reconcile to the broker’s confirmed executed quantity?
ChargesFee amounts, currencies, allocation and applicable rebatesAre charges already embedded in another input?
OutputCost currency, denominator, benchmark label and calculation versionCan someone reproduce both the cash cost and basis-point result?

When joining observations to an order time, use a documented rule, such as the latest eligible observation at or before that time, with a maximum permitted age. A later observation creates look-ahead in the comparison. If none qualifies, mark the benchmark missing instead of borrowing the next price without disclosure.

Store event and receipt times separately. A calculation that compares clocks from different systems can produce a misleading age or delay estimate if those clocks are not aligned. Do not label that difference network latency without establishing what the timestamps measure.

For grouped reports, keep currencies and benchmark conventions consistent. A simple average of per-order basis points gives a small order the same weight as a large order. A notional-weighted result answers a different question. Name the weighting method, reconcile its included costs and avoid mixing arrival-based and decision-based figures in one unlabeled total.

Where SiftingIO fits in an execution-cost workflow#

SiftingIO provides market data for the context and research side of this workflow. Depending on the market and dataset, its API documentation describes live observations, historical bars and streaming access. Broker or venue execution records remain the source for actual fills and charges.

Keep a SiftingIO observation’s source definition attached to the comparison. As explained in our data methodology, a composite or reference value has its own construction. It should not silently become a broker’s executable quote, an official consolidated best quote or evidence of depth available for a particular order.

For a first prototype, reproduce the workbook above, import your own permitted execution records, and add clearly labelled market context. The Market Data Export tool can support historical price studies; an OHLCV bar does not reconstruct every bid and ask or reveal the path that filled an order. If the analysis requires historical quote depth, verify that specific dataset rather than assuming price-bar coverage includes it.

The resulting report should make the distinction visible: an actual fill, a benchmark observation, and a cost calculation with an explicit definition. That is more useful to a customer than a “slippage” number whose source and denominator are hidden.

Frequently asked questions#

Can slippage be negative?#

Yes. Under this article’s sign convention, negative price cost means improvement relative to the selected benchmark. A buy below the benchmark or a sale above it produces a negative value. Fees can still make the all-in cost positive.

Does a limit order remove slippage?#

A limit order constrains execution to its limit price or better, but does not guarantee a fill. The result may still differ from an earlier midpoint, decision price or quote. The benchmark and order limit are separate inputs.

Can daily candles measure the slippage of a specific trade?#

They cannot establish it precisely. A daily bar summarises a period and does not preserve the quote, available quantity or event sequence at order arrival. It can provide context, but a trade-specific measurement requires a suitably timed benchmark and actual fills.

Is there a normal slippage percentage for every market?#

There is no single percentage that defines acceptable execution across instruments, sizes, sessions and order types. Compare like-for-like observations, report the benchmark and investigate unusual outcomes with the underlying records. A fixed backtest allowance is an assumption to test, not evidence of future execution quality.

This article is educational and is not investment or trading advice. Its examples are hypothetical. SiftingIO supplies market data and does not execute trades or guarantee execution prices.

Keep reading

Related posts