How to estimate slippage from data

Spread, depth, and volatility are conceptually relevant inputs for bounding execution cost at the decision timestamp, with spread representing immediate crossing cost, depth representing liquidity consumption, and volatility representing timing risk.

Slippage is the gap between a decision price and the price actually obtained when an order is executed. For transaction cost modeling, a useful estimate starts at the decision timestamp and asks what the market state implied about execution difficulty at that moment.

A practical bound can be built from three contemporaneous inputs: the bid ask spread, available depth, and short horizon volatility. The spread measures the immediate cost of crossing from one side of the market to the other. Depth shows how much size is available near the top of book before the order would need to consume worse prices. Volatility captures the risk that prices move during the interval between the trading decision and the completed fill.

The mechanism is straightforward. A small order in a deep and stable book may incur a cost close to the quoted spread. A larger order relative to displayed liquidity may need to walk the book, so the expected execution price depends on depth across multiple price levels, not only on the best bid or ask. If volatility is elevated, the uncertainty around that execution also widens because the market can move while the order is being routed and filled.

In that sense, spread, depth, and volatility represent different components of execution cost. Spread is the immediate crossing cost. Depth is the market impact implied by the available resting liquidity. Volatility is the adverse price movement risk during execution. Combining them at the decision timestamp produces a more realistic cost bound than a spread only estimate because it reflects both liquidity and timing risk.

The backtest overfitting audit source supports the need to test transaction cost assumptions with strict out of sample discipline. It explains that backtest research can overstate model quality when design choices are tuned to historical data, and it recommends controls such as validation splits and multiple testing awareness. Applied here, a slippage bound should be calibrated on one sample and checked on another, and improvements over a naive baseline should be measured by forecast error rather than accepted from in sample fit alone.

The deflated Sharpe ratio glossary source is relevant only as a caution about overstating statistical evidence after repeated testing. It explains why naive significance measures can be inflated when many variants are tried. For slippage estimation, the implication is that a better fit from adding spread, depth, and volatility should be judged with methods that account for selection effects if many specifications were explored.

Spread, order book depth, and venue specific market conditions are relevant inputs for execution cost modeling, and any evaluation should use robust out of sample testing to avoid overfitting.

Drafted with AI assistance from cited sources. Reviewed and approved by Sonar Sciences Quant & Research Team.