Slippage Underestimation in Trading Execution

Issue: Execution Agent’s Pre-Trade Cost Model Systematically Understates Realized Slippage, Especially for Larger or Illiquid Orders

Frequency: Very Common

Symptoms

• Realized execution price consistently worse than pre-trade estimated price, beyond what bid-ask spread alone explains
• Slippage estimates derived from average daily volume but ignore intraday liquidity patterns (open/close auctions, lunch lull)
• Performance attribution shows persistent negative “execution alpha” that the model never learns to correct for
• Larger orders show disproportionately worse slippage than the model’s linear cost assumption predicts

Root Cause Many execution-cost models assume a linear or square-root relationship between order size and price impact calibrated on average historical conditions. They do not adapt to current order-book depth, time-of-day liquidity, or the order’s own footprint as it executes. The result is a pre-trade cost estimate that looks reasonable in isolation but is consistently optimistic versus realized fills, particularly for less liquid names or stressed markets.

Example

Scenario: Agent submits a buy order for 2% of average daily volume
Pre-trade estimate: Expected slippage = 8 bps vs. arrival price
Execution: Order worked over 2 hours during low-liquidity midday window
Realized slippage: 35 bps vs. arrival price
Cumulative impact: Over a quarter of similar trades, underestimated slippage erodes ~40bps of annual strategy alpha

Key Statistics

• Linear/static slippage models underestimate realized cost by 2-4x for orders >1% of ADV in studies of institutional execution data
• Time-of-day liquidity variation accounts for a large share of intraday slippage variance not captured by static volume-based models
• Persistent negative execution alpha attributable to slippage underestimation is a commonly cited gap between paper and live strategy performance

Mitigation Strategies

1. Order-Book-Aware Cost Models: Use real-time depth-of-book and recent trade-and-quote data to estimate impact, not just historical ADV
2. Time-of-Day Liquidity Curves: Model intraday liquidity seasonality explicitly (open/close auctions vs. midday lull)
3. Post-Trade Calibration Loop: Continuously compare realized vs. estimated slippage and recalibrate the cost model with a feedback loop
4. Adaptive Order Slicing: Size and pace child orders dynamically based on live liquidity signals rather than a fixed schedule

Metrics

• Realized vs. pre-trade estimated slippage (bps), tracked per order-size bucket
• Execution alpha (cumulative slippage cost vs. benchmark)
• Model calibration error trend over time

Alerts

• Realized slippage exceeds pre-trade estimate by >2x on >10% of orders in a session → P2
• Cumulative execution alpha drag exceeds strategy’s gross alpha budget → P1

References

• Agentic Trading: When LLM Agents Meet Financial Markets
• Toward Reliable Evaluation of LLM-Based Financial Multi-Agent Systems