One of the most appealing aspects of quantitative trading is that there is no end to the number of different components that we can add to our strategies in order to improve performance and reduce risk factors. While most traders agree that simpler systems will be more robust over time, we still catch ourselves constantly looking for tweaks to improve our returns.

Is experimenting with strategy filters and stops a complete waste of time or is there evidence that supports the effort?
About a month ago, I covered an article about an attempt to salvage a bad strategy. その記事で, we looked at a stock rotation strategy that Cesar Alvarez had written about and how disappointing its performance had been. Despite the poor performance of the base strategy, Cesar received a number of comments from readers suggesting different ways that the strategy could be salvaged using different types of filters and stops.
で a recent update to that post, Cesar tested quite a few of these ideas and discussed the results. His backtesting results demonstrates that despite the large number of ways to combine stops and filters, most of them actually have a negative impact on the base strategy.
The Base Strategy
The base strategy that Cesar was working with ranks each of the stocks in the S&P 500 based on performance over the previous nine months. The strategy holds the top ten stocks and rebalances on the first trading day of each month.
This strategy posted an annual return of 7.3% 差出人 2004 を通じて 2013. The maximum drawdown during that time was 61.01%. This represents a slightly better annual return with slightly large drawdown than simply holding SPY would have produced.
Adding Stops & Filters
Cesar tested six different types of filters that were based on moving averages and new highs. These are the types of filters that are commonly suggested for improving strategies of this type.
意外にも, all but one of the filters actually lowered the annual return. Requiring a stock to be above its 50- or 200-day moving simple moving average made almost no difference in this case because stocks ranking highest in 9 month return tend to already be trading above their moving averages.
Adding stops to the strategy had a similar effect. While the stops did a good job of reducing the maximum drawdown, they also reduced the annual return quite significantly.
The Type of Filter That Did Work
The one type of filter that did work across the board was a market timing filter based on the S&P 500 closing above its 200-day simple moving average. Requiring the S&P 500 to be in an uptrend improved the returns on every version of the strategy that Cesar tested.
What this case study shows us is that it is important to integrate a level of critical thinking when adding filters and stops to our strategies. Filters that replicate the basics of the strategy are probably a waste of time, but there are probably other filters out there that would be worthwhile to experiment with.