Every cost number has a reference price
Slippage is the difference between an execution price and a specified reference. That reference might be the decision-time midpoint, the arrival midpoint or a scheduled benchmark. These choices answer different questions. A decision-to-arrival move is included under the first benchmark but excluded under the second. Name the benchmark before interpreting a positive or negative execution-cost number.
The SEC's Trading Basics bulletin distinguishes market orders from limit orders and explains that execution price and execution certainty are different considerations. A research engine should preserve that distinction. A visible quote does not prove that a large order would fill completely, and a limit price touched by a bar does not establish the queue position or traded volume available to the simulated order.
Reconcile a fictional buy order
Suppose a fictional stock is quoted at $49.98 bid and $50.02 ask, with midpoint $50. A 10,000-share buy executes at an average $50.035 and pays $10 in explicit fees. Relative to $500,000 of decision notional, the price difference costs $350, or seven basis points; fees add 0.2 basis points, making $360 or 7.2 basis points in total.
For illustration, the $350 can be decomposed into $200 for crossing from midpoint to the original ask and $150 beyond that ask. Those are four and three basis points, respectively. Calling the second component impact is only a modeling attribution: without richer evidence it may also contain market movement, latency or changing liquidity. The aggregate price difference is directly defined; its causal decomposition is not.
Avoid double counting and favorable omissions
If simulated execution prices already cross the spread, do not subtract another full spread from returns. If a model supplies all-in slippage relative to midpoint, adding half-spread again can duplicate part of its charge. Conversely, quoting commissions alone omits price effects. Keep a ledger listing each component, units, sign, benchmark and whether it is already embedded in the execution price.
Unfilled quantities also matter. Report intended size, filled size, cancellation time and residual position. A strategy that discards every unfilled losing order while retaining filled winners can appear excellent without being implementable. Opportunity-cost calculations require a declared comparison horizon and should be reported separately from paid fees. Do not manufacture an execution merely to keep a backtest fully invested.
Use scenarios until execution evidence exists
Test costs under different spreads, delays and participation assumptions, including stressed conditions. A daily closing-price dataset cannot validate intraday queue mechanics or a precise impact curve. It can support a clearly labelled scenario analysis with explicit limitations. Calibrate more detailed assumptions only against suitable observed executions, and preserve the distinction between the calibration sample and the strategy evaluation sample.
Negative measured slippage can occur when prices move favorably; it is not proof of execution skill from one order. Aggregate costs with consistent notional weights and inspect the distribution, not only its mean. This guide's quote, execution and fee inputs are hypothetical. They demonstrate a reconciled ledger and do not assert that any broker, venue or strategy achieved the stated prices.
Inspect the example
| Illustrative cost component | Dollar cost | Basis points of $500,000 |
|---|---|---|
| Midpoint to original ask | $200 | 4.0 |
| Execution beyond original ask | $150 | 3.0 |
| Explicit fees | $10 | 0.2 |
| Total decision-midpoint cost | $360 | 7.2 |
Sources & reproduction
Primary references support the definitions and method. Numerical examples are illustrative unless a source record is explicitly identified.