Backtesting is a method used by investors to develop trading systems. It involves testing a trading system on historical data to see how it would have performed in the past. Backtesting can be used to test a wide variety of trading systems, from simple trend-following systems to complex algorithmic trading strategies.
Backtesting relies on historical data to test a trading system. This data can be sourced from a variety of sources, including brokerages, data vendors, and exchanges. Backtesting software can be used to automate the process of testing a trading system on historical data.
Despite the advancement in software development, most backtesting software still cannot take slippage into account. Slippage is the difference between the price at which a trade is executed and the price that was anticipated. Slippage can have a significant impact on trading performance, especially when testing high-frequency trading strategies.
Reference [1] proposed a method for estimating the bid-ask spread. Specifically, it developed unbiased bid-ask spread estimators from several combinations of Open, High, Low, and Close (OHLC) prices. It then combined them to arrive at an efficient estimator, a so-called Efficient Discrete Generalized Estimator (EDGE). This bid-ask estimator can be applied at any frequency.
As an application, the author demonstrated the efficiency of the bid-ask estimator,
To demonstrate the potential benefits of EDGE, we provide three representative examples, illustrating how the choice of the bid-ask spread estimator can affect economic significance and statistical inference in empirical work.
The results are reported in Figure 7 where small, mid, and large caps are shown in Panel A, B, and C, respectively. From the recent sample (CRSP-TAQ), we conclude that EDGE closely follows the effective spread benchmark whenever it is not tiny. This is the case for small-cap stocks and for all stocks before the year 2000. CS and AR tend to underestimate the spread, particularly for small-cap stocks, mirroring the fact that these estimators are biased when trading is infrequent. In the arguably less liquid historical sample prior to 1993, we find that the gap between EDGE and the alternative estimators further widens. The unbiased EDGE is by a factor of two larger than AR, and the difference is even more pronounced compared to CS.
In short, the developed bid-ask estimator was effective and can be used in a variety of applications and at different frequencies.
We believe that this estimator should be used to determine the impact of slippage on trading performance. It should be incorporated into backtests in order to provide more accurate trading results.
References
[1] Ardia, David and Guidotti, Emanuele and Kroencke, Tim Alexander, Efficient Estimation of Bid-Ask Spreads from Open, High, Low, and Close Prices (2021). https://ssrn.com/abstract=3892335
Originally Published Here: How to Account for Slippage in Backtesting
No comments:
Post a Comment