Assumptions in Backtesting
A guide on the common unrealistic assumptions made during quant backtesting and simulation, and the risks they pose.
TL; DR
Unaccounted Market Impact: Ignoring the effect of large orders on market prices can lead to overestimated strategy performance.
Ignoring Transaction Costs: Overlooking fees, spreads, or slippage can lead to an overestimation of profitability.
Underestimating the Impact of Liquidity: Failing to consider liquidity constraints may render a strategy non-viable in live markets.
Insights
Quantitative backtesting and simulation are powerful tools for evaluating the potential performance of trading strategies based on historical data. However, these methods can lead to misleading results if they rely on unrealistic assumptions. Below are examples of such assumptions and the risks they pose.
Unaccounted Market Impact
Assumption:
When a large order is placed, it does not significantly affect the market price.
Risk:
Large orders can move the market, especially in less liquid assets.
The strategy may appear profitable in simulation but fail in live trading.
Ignoring Transaction Costs
Assumption:
Trading does not incur any costs such as fees, spreads, or slippage.
Risk:
The profitability of a strategy is overestimated.
In reality, transaction costs can significantly erode returns.
Underestimating the Impact of Liquidity
Assumption:
Assets can always be bought or sold at the current market price regardless of the quantity.
Risk:
The strategy may not be executable in live markets due to insufficient liquidity.
The model's trades could themselves affect liquidity and pricing.
To mitigate these risks, it is essential to use realistic assumptions in quant backtesting and simulation. This includes accounting for transaction costs, market impact and liquidity.
Last updated