Slippage and friction modeling

Definition

Friction is everything that separates paper returns from realized returns: spread cost, market impact (the price moving because the order itself trades), delay cost, fees, and financing. Slippage is the realized difference between the decision price and the achieved fill. Perold (1988) named the total gap implementation shortfall — the platform’s simulation engine charges an explicit friction model so no strategy is graded on paper prices.

How it works / structure

• Cost stack: half-spread per marketable fill (ms-bid-ask-spread) + market impact (size-dependent) + fees and commissions + borrow/financing where applicable + delay cost between signal and execution.
• Impact modeling: impact grows sublinearly with order size — empirical work is broadly consistent with square-root-of-size behavior (Gatheral 2010 discusses the model class and its no-arbitrage constraints); temporary vs permanent components are modeled separately in the Almgren-Chriss (2000) execution framework.
• Execution scheduling: trading faster costs more impact, trading slower bears more price risk — Almgren-Chriss formalizes the tradeoff; participation-rate limits cap the modeled speed.
• Simulation parameters (platform defaults): per-instrument spread model, impact coefficient scaled by volatility and size/ADV, per-trade fees, and a delay assumption; stress runs multiply the friction model to test edge robustness.

When it applies

Every simulated fill and every strategy grade. Friction dominates the viability question for high-turnover strategies, small-cap and thin-options strategies, and stop-heavy management styles (mgmt-stop-loss fills are marketable orders in fast markets — the costliest kind).

Risk profile & failure modes

• Underestimated friction = fake edges: the most common backtest failure on record; edges smaller than realistic costs survive paper and die live.
• Calm-calibrated models: impact and spread calibrated on average days understate stressed costs, when exits actually happen (ms-liquidity).
• Crowding externality: several accounts running the same signal experience each other’s impact; per-account models understate shared-signal costs.
• Gap-through events: stops and market orders across session boundaries or halts fill far from trigger prices — friction models bound normal conditions, not gaps (ms-sessions-auctions).

Evidence & limits

Implementation shortfall (Perold 1988) and optimal execution (Almgren-Chriss 2000) are standard, widely used frameworks; square-root-class impact has broad empirical support (Gatheral 2010). Exact impact coefficients are instrument- and period-specific and not publicly settled — the platform states its calibration and treats it as a model, exposing stressed-friction replays rather than claiming precision.

Falsifiable-thesis examples

Illustrations only, not signals:

• “Strategy S’s live-to-simulated return gap will stay within the modeled friction budget of F basis points per trade over the next quarter” — falsified by realized shortfall exceeding the budget.
• “Doubling modeled friction leaves strategy S profitable in replay” — falsified by the stressed run’s P&L.

Cross-references

• Components: ms-bid-ask-spread, ms-liquidity, ms-sessions-auctions
• Consumers: every pillar-4 strategy entry, mgmt-stop-loss, mgmt-scaling
• Method: lens-quantitative (cost-inclusive backtesting rule)