Last Updated on 3 November, 2022 by Samuelsson
The 3-day mean-reversion strategy is one of the strategies in the book, High Probability ETF Trading, published by Larry Connors and Cesar Alvarez in 2009. Let’s take a look at the strategy and the backtesting results from Quantified Strategies.
What is the 3-day mean-reversion strategy?
The 3-day mean-reversion strategy is an ETF trading strategy that uses the highs and lows over 3 consecutive days to determine a trade setup. The basic principle of the strategy is to buy when an ETF has made lower highs and lower lows for three consecutive days.
The rules of the 3-day high/low method
Here are the rules of the strategy:
- The day’s close must be higher than the 200-day moving average.
- The day’s close must be lower than the 5-day moving average.
- Both the high and low of the preceding two days were lower than those of the day before.
- The preceding day’s High and Low were lower than those of the day before.
- The day’s High and Low are lower than those of the preceding day.
- If conditions 1-5 are true, then buy at the close of the current day.
- Exit at the close when the close is above the 5-day moving average.
As you can see, the strategy has seven criteria.
The Amibroker code for the strategy
In Amibroker, the code for the strategy looks like this:
Buy= C>MA(C,200) AND C<MA(C,5) AND H<Ref(H,-1) AND L<Ref(L,-1) AND Ref(H,-1)<Ref(H,-2) AND Ref(L,-1)<Ref(L,-2) AND Ref(H,-2)<Ref(H,-3) AND Ref(L,-2)<Ref(L,-3);
The results of the 3-day mean-reversion strategy
Here’s a table that summarizes the result and compares them with Connors’ results:
Result by Connors Average gain since Profit
Average gain in % inception to Nov.2020 factor Difference:
DIA 0.27 0.23 1.42 -0.04
EEM 0.91 0.41 1.6 -0.5
EFA 0.63 0.37 1.68 -0.26
EWH 0.72 0.08 1.11 -0.64
EWJ 0.46 -0.03 0.89 -0.49
EWT 0.42 0.48 1.81 0.06
EWZ 1.66 1.37 4.03 -0.29
FXI 0.89 0.49 1.58 -0.4
GLD 0.79 0.58 2.24 -0.21
ILF 1.79 0.76 1.73 -1.03
IWM 0.5 0.5 1.91 0
IYR 0.01 0.12 1.1 0.11
QQQ 0.88 0.7 2.47 -0.18
SPY 0.91 0.72 2.93 -0.19
XHB 0.76 0.99 4.27 0.23
XLB -0.29 -0.01 0.97 0.28
XLE 0.66 0.39 1.42 -0.27
XLF 0.38 0.6 2.77 0.22
XLI 0.41 0.37 1.62 -0.04
XLV 0.36 0.42 2.27 0.06
ETFs not included:
GDX -0.3 0.8
GDXJ 0.9 1.48
TLT 0.29 1.72
XLP 0.2 1.49
XME 0.76 1.7
The table shows that 13 of 20 ETFs could not perform well since the original test 12 years ago. This poor performance in the last 12 years can be clearly seen in the compounded equity curve for ILF below:
Here’s how the equity curve for the S&P 500 (SPY), the most-traded ETF, looks:
See the equity curve for XHB, the ETF with the best profit factor:
How the strategy performs on a portfolio
To simulate how the strategy performs in a portfolio of ETFs, let’s assume you traded the strategy on all 25 ETFs (the original 20 tested by Connors and the additional five included by Quantified Strategies) from the year 2000 until December 2020. If you set the system to make a maximum of five open positions at any time and risk no more than 20% of equity on each position, the equity curve (compounded) would look like this:
The results are as follows:
- There are 1616 trades from the year 2000 until November 2020.
- The average gain per trade is 0.38%.
- The CAGR is 5.74%.
- The win rate is 72.3%.
- The profit factor is 1.53.
- The maximum drawdown is 16.4%, which is much lower than the 55% for the SPY during the period: the worst drawdown came at the end of July 2011, at the peak of the EU debt crisis.
While the strategy seems not to perform as it used to, it still shows some promise. However, it might perform better if some of the variables are tweaked a little bit.