Modeling Strategy Capacity and Slippage in Backtests

Summary

The document discusses how a strategy’s performance can change as invested capital grows. Trading costs, including slippage and commissions, may rise nonlinearly, so a strategy can reach a capacity beyond which adding capital no longer improves its results. It proposes evaluating performance at that capacity and using a minimum backtest length to obtain more reliable risk-adjusted statistics.

It asks for references and standard slippage models when only daily OHLC data are available, and whether this evaluation approach is sound. The text does not provide a specific model, empirical results, or a settled answer. Its central lesson is that backtests should account for both trading capacity and the data limits of cost estimation; capacity estimates and OHLC-based slippage assumptions require careful validation.

Key ideas

• Trading costs can increase nonlinearly as strategy capital grows.
• A strategy has finite capacity beyond which additional capital may not improve performance.
• The document proposes assessing results at estimated capacity and over a minimum backtest length.
• Daily OHLC data pose a modeling challenge for estimating realistic slippage.
• The proposal is a question for discussion and supplies no validated slippage model or empirical evidence.

Tags

# Realistic Modeling of Capacity in Backtesting a trading strategy


# Realistic Modeling of Capacity in Backtesting a trading strategy












Typically the backtest of a quantitative trading strategy assumes a fixed period and fixed capital at the start to backtest a strategy. However, each strategy has a capacity (due to non-linear trading costs/slippage/commissions) beyond which adding any money to strategy doesn't improve performance and I think it is correct to measure the performance of strategy exactly with the capacity of the strategy as the initial amount and for a Minimum Backtest Length (see Lopez de Prado https://papers.ssrn.com/sol3/papers.cfm?abstract_id=2308682) so that the risk-adjusted statistics are accurate.

Are there any references for this kind of approach? What are the standard models for slippage used in backtesting strategies daily OHLC data? Also, any thoughts on the correctness of this approach to backtesting the strategy are appreciated.

Shown in full with attribution under the source's licence. Licence: CC BY-SA 4.0 (Stack Exchange)

This summary was written by Stratmill's research agent from the original; it is not a copy of the source.