Swing mean reversion strategies in 2022
A substantial portion of my systematic portfolio comes from short-term swing mean reversion strategies in equity markets. How did they do in 2022?
I continually compare my performance against benchmarks to get an answer as to whether my strategies are working even when they are going through drawdowns.
Personally, my mean reversion strategies have performed "around zero" this year, which I view as very positive given the moves we have seen in the markets this year. Mean reversion strategies basically trade against the market. That is, they buy what is falling, short what is rising. Like all strategies, they have better and worse years. This year was certainly one of the worse ones, mainly because many stock titles fell in very volatile moves that did not revert to the mean. However, I see it as very positive that the performance of the systems I trade in my bigger account has been much better than if I had classically held stocks (buy and hold any of the indices).
But with worse performance, how do you know if the system is working or maybe stopped working due to over-optimization? Personally, I compare its performance to the benchmarks I create. These represent the absolute simplest form of trading principle. For mean reversion, my benchmark looks like this:
I track stocks in a given index whose price is greater than $20 and average volume over 20 days is greater than 200,000. I follow stocks that have sufficient volatility and that are trading above their long-term moving average MA200.
For LONG, I follow stocks that close below MA5 - 1*ATR5. For these, I open trades the next day with a limit order at Close - 0.75*ATR5. I close the position on the first rising day or if the trade is open for 5 days.
For SHORT, I follow stocks that close higher than MA5 + 1*ATR5. For these, I open trades the next day with a limit order at Close + 0.75*ATR5. I close the position on the first down day or if the trade is open for 5 days.
Within the bechmark I trade all possible signals regardless of capital. I do not follow trading fees. I am interested in the main characteristics of the result when trading all signals.
Such simple benchmarks basically capture the basic principle of edge capture. If the benchmarks were rising but my live trading performance was falling, it would be time to look for a flaw in my own systems. If the benchmarks started to show some fundamental change relative to historical test runs, one would have to wonder if the alpha in question was still present in the market.
Right now, the situation looks like this.
Benchmark applied to stocks in the Russell 3000 index:
Benchmark applied to stocks in the S&P 500 index:
Benchmark applied to stocks in the Nasdaq 100 index:
Benchmark applied to stocks in the biotech stocks:
Benchmark applied to stocks in the Russell MicroCap:
Benchmark applied to stocks in the Russell MidCap:
Benchmark applied to Canadian stocks in the S&P/TSX Composite Index (here without regard to share price):
The red line shows the performance of the short part of the strategy, the blue the long part. Black line is the performance of the whole portfolio (long + short). Graphs are in logarithmic scale. Strategies are treated for survivorship bias - they only trade stocks that were in the index on the trade date.
I trade a wide range of US stocks and Canadian stocks within my portfolio. Thus, for the US, I am most interested in the benchmark stocks of the Russell 3000 index, whose performance has been going sideways for the last year and a half or so. Which is broadly in line with what is happening in my account. However, on the long term chart of the benchmark, you can see that the biggest gains from mean reversion come in cycles. So I'm not making any major changes to my strategies for now, and I'll be watching to see how the benchmarks perform over the next year when the markets settle down a bit again.
BTW - at first glance, the equity curve of the benchmark based on Canadian equities looks the most interesting (I only trade the long component in my live account for now). This goes in line with my belief that well performing algorithmic strategies from the US markets can be even more profitably traded in markets that are not yet so algorithmically heavy.