How to Model Slippage, Fees and Market Impact in Python

Simulate fills relative to a timestamped reference price. For an immediately marketable order, account for crossing the spread, temporary impact and any delay between decision and execution. Use the gap between the reference and the simulated fill to measure slippage. Debit explicit fees separately. Estimate the model from past fills or market data, and rerun the backtest with different liquidity and cost assumptions to see how much they affect the result.

Build the fill from a quoted midpoint

A small illustrative model uses arrival midpoint

M

, absolute bid–ask spread

s

, and a temporary impact fraction

i

. For signed quantity

q

P_{\mathrm{fill}}=M+\operatorname{sign}(q)\left(\frac{s}{2}+Mi\right).

An order sized at a fraction $|q|/V$ of typical daily volume can use a scenario rule $i=\eta\sqrt{|q|/V}$ . This is a modelling choice. Estimate $\eta$ in consistent units, segment by asset and liquidity, and avoid extrapolation far beyond observed sizes. A separate participation limit can cap the shares filled within each bar. Add market movement between decision and arrival when your reference timestamp precedes the simulated execution window.

Update cash by $-qP_{\mathrm{fill}}-f(q)$ , with $f$ set by the brokerage schedule, including per-order minimums and per-share charges if relevant. Choose whether fees apply per fill or to a parent order. For longer-horizon cost models, specify how lasting impact changes subsequent marks and fills.

A buy for 5,000 shares arrives at a $40.00 midpoint, with a $0.04 spread and typical volume of 500,000 shares. With $\eta=0.004$ , the impact fraction is $0.004\sqrt{0.01}=0.0004$ . The simulated fill is $40.036. Spread costs $100, impact costs $80, and fees of half a cent per share cost 25 dollars. The total shortfall from the arrival midpoint is $205, or 10.25 basis points of $200,000 reference notional.

On the sell side, check that the same spread and impact assumptions push the fill below the arrival midpoint. Record execution cost as signed quantity times fill price minus midpoint, with the fee added separately. If P&L already uses the fill price, spread and temporary impact have already entered it. Deducting those amounts as a second charge would count them twice. Use the price bridge to explain the result, and use explicit fees for cash deductions that the fill price does not capture.

Implement signed fills and fees

from math import copysign, sqrt

def market_fill(mid, spread, shares, daily_volume, impact_scale=0.004):
    impact = impact_scale * sqrt(abs(shares) / daily_volume)
    price = mid + copysign(spread / 2 + mid * impact, shares)
    commission = max(1.0, 0.005 * abs(shares))
    cash_change = -shares * price - commission
    shortfall = shares * (price - mid) + commission
    return price, cash_change, shortfall

print(market_fill(40.0, 0.04, 5_000, 500_000))

This function prices a nonzero order assumed to fill in full. In a simulator, apply it only to the quantity the current execution window can fill.

Walk the book with different market-order sizes in the order book tool to see the price differences your fill model is approximating.

References and next steps

QuantConnect, slippage models shows practical fill-model assumptions.
How to Test Whether a Significant Signal Survives Costs

Build the fill from a quoted midpoint

Reconcile every component on a 5,000-share buy

Implement signed fills and fees

Find the execution cost that consumes a trading edge

References and next steps