Filters are designed with one purpose—to remove impurities. This is as true when applying filters to trading strategies as to your water supply.
What impurities are there in trading strategies, and what types of filters are used to remove them?
Filters in the trading sense need to be thought of as additional criteria or conditions that must be met before a trade can be taken. A filter could be as simple as applying a minimum volume threshold on a simple moving average crossover strategy.
Some are common and some more obscure. And it is often the obscure filters that can have the biggest impact. Two worth discussing are time and day of week. The basis for applying these filters is derived from the reality that no single strategy works in all environments, and if you can identify a way that eliminates more negative trades (in volume and scope) than positive trades, then you have improved your strategy.
A mean reversion strategy does badly in a trending market, as does a trend-following strategy in a choppy or sideways market. This begs the questions, “do trend days trend all day?” Or, “Do choppy days trade sideways all day?” The answer is “yes” and “no.” But when you take a strategy and break it down to more finite detail in terms of how it works, the answer becomes more black and white, for example, the size of your profit targets or stop losses.
Quantitative trading shuns generalities; it needs hard data, which is where time and day filters come in. Our analysis of the Treasury bond market showed that short positions in our mean reversion strategy underperformed on Fridays. Therefore we placed a filter on the strategy restricting us from taking short signals in Treasuries on Fridays.
Next comes time, which plays an even bigger role. At the equity cash open or the Treasury cash close, market volatility increases dramatically. False moves, for example, or whipsaws, happen quickly. Why risk capital under such adverse conditions? That is the simplest example of when to use a time filter.
Markets act differently between the U.S. equity cash open and the European close as compared to the European close to the U.S. equity cash close. Those in many ways are two different markets. Two different strategies can be used, one for each period of time. So why take one strategy and run it through two, often totally different, environments? That is where a time filter can be applied.
Trading can be seen as a covert, special ops mission: Get in and get out; limit risk by reducing exposure and wait for opportunities where probability is more in your favor.
The use of filters potentially can improve a strategy’s performance. The key is being able to use quantitative data to fine tune and construct those filters. “Improvement,” shows two returns for the same 30-year Treasury futures scalping program (one with a time filter). Rather than run the same strategy from the cash open to cash close, we looked at what time during the session performance started to drop off.
By applying a simple time filter, performance was greatly improved. Note how the filter more than doubles the gross per trader return. It is the same strategy, except when the filter is applied it is operating 30 minutes after the equity cash open and shortly after the European close.
These time filter parameters change by strategy. What works for the above example will differ for other strategies. Other filters can include days of week, data releases and more. The point is, filters can play an important role in improving performance. The key, though, is to use a long enough look-back period to get a true understanding versus a shorter look-back that would be more curve-fitting.