Recently there is renewed interest in counter-trend strategies, especially for trading stock index ETFs. These strategies became popular about 5 years ago when they appeared in books and magazines. Their purported simplicity and profitability based on historical backtests have made them very attractive, especially to retail short-term and position traders of ETFs and stocks. The managed futures crowd has shown interest in such strategies recently because of the lackluster performance of trend-following models during the last few years and increasing fears that this style of trading is dead. In this article I present performance results of a portfolio of three very popular ETFs, SPY (NYSEARCA:SPY), QQQ (NASDAQ:QQQ) and IWM (NYSEARCA:IWM), for three different counter-trend trading models, and I compare them to those of a very simple trend-following model applied to the same portfolio. The results provide strong indication that the counter-trend trading models considered in this article are not as robust as the simple trend-following model and I list possible reasons for their performance degradation in recent years as compared to historical performance results before they gained popularity.
The basic idea behind counter-trend trading
Most known counter-trend trading systems attempt to capture profits from corrections generated by overbought markets on an uptrend, by establishing short positions, or by oversold markets on a downtrend, by establishing long positions. This is a trading method based on gambler's fallacy and the (false) notion that after a series of new highs or new lows, the probability that a correction will occur is higher. If the correction does not occur within the limits defined in the exit logic of such models, a loss is generated. Thus, it is immediately seen that for such models to perform well, the market must exhibit sufficient swings above and below the underline trend. Note, however, that buying weakness along the direction of an uptrend or selling strength along the direction of a downtrend do not amount to counter-trend trading and this fact is not always clarified by authors of such systems. This is equivalent to buying short-term dips and selling short-term tops along the underline trend. In that case, the trading models are better described as "counter-momentum" but the basic idea is the same, i.e. these models facilitate short-term position trading that is not related to trend-following.
Portfolio backtesting results of three counter-trend strategies
The performance of each strategy is evaluated on a portfolio of 3 popular ETFs, SPY, QQQ and IWM, in two distinct but adjacent time periods. The first period is from 01/04/2000 to 12/31/2007 and the second the period from 01/02/2008 to 11/08/2012. The former time period represents when these strategies were identified and possibly optimized. The latter plays the roll of an out-of-sample test period, i.e. a period after the strategies were already public and extensively used. The main objective of the article is to study the performance of each strategy in the two time periods and compare the results.
The starting equity value of the portfolio of the three ETFs is $100,000. No commission or slippage are included in the tests as the intent here is to evaluate the basic idea behind counter-trend trading. Every ETF is allocated one third of the available equity when a position is opened and no multiple open positions are allowed, i.e. an open position in each ETF must be closed before another one for the particular ticker can be opened. The number of shares is calculated by dividing the allocated equity by the entry price.
1. The double 7s strategy
This is a long-only counter-momentum strategy defined as follows:
- Price is above its 200-day moving average
- If price closes at a 7-day low, then buy.
- If price closes at a 7-day high, then sell the position.
Below is a graph of the portfolio equity from 01/04/2000 to 12/31/2007 :
Below is a table of the monthly returns for the same time period:
Next, is a graph of the portfolio equity from 01/02/2008 to 11/08/2012:
Below is a table of the monthly returns for the same time period:
A performance summary for this strategy is shown in Table 1 below:
Parameter |
01/04/2000-12/31/2007 |
01/02/2008-11/08/2012 |
CAR |
7.20% |
1.31% |
Win rate |
76.35% |
69.11% |
Trades |
203 |
123 |
Profit factor |
2.41 |
1.14 |
Sharpe ratio |
1.82 |
0.22 |
Max. Drawdown |
-10.81% |
-15.22% |
Table 1. Performance summary for the double 7s strategy
The degradation of the performance of the strategy after 2007 is not only evident from the equity graph and the monthly returns table but also from the sharp drop in the CAR, profit factor and Sharpe ratio values. Although during the first time period, from 01/2000 to 12/2007, there is only one down year out of 8 in total and the equity curve rises fast after 2004, in the second time period, from 01/2008 to 11/08/2012, there are 2 losing years out of 5 and the equity curves becomes erratic. These facts point to the possibility that the parameters of this system are suitable for market conditions encountered before 2008 and that the system itself is an artifact of selection bias and curve-fitting. However, there are other possibilities that are discussed at the end of this article.
2. The double 10s strategy with 1-day exit
This is a long/short counter-trend/momentum strategy, defined as follows:
- Buy at the close if price makes a 10-day low and exit the position at the close of the next day.
- Sell at the close if price makes a 10-day high and exit the position at the close of the next day.
Continuation and reversal signals are allowed, in the sense that when a long or short signal is closed another long or short signal can be opened if the conditions remain in effect and short signals are initiated at a long exit bar if conditions are right and vice versa.
Below is a graph of the portfolio equity from 01/04/2000 to 12/31/2007:
Below is a table of the monthly returns for the same time period:
Next is a graph of the portfolio equity from 01/02/2008 to 11/08/2012:
Below is a table of the monthly returns for the same time period:
A performance summary for this strategy is shown in Table 2 below:
Parameter |
01/04/2000-12/31/2007 |
01/02/2008-11/08/2012 |
CAR |
8.63% |
3.01% |
Win rate |
51.03% |
47.68% |
Trades |
1742 |
1034 |
Profit factor |
1.23 |
1.08 |
Sharpe ratio |
0.74 |
0.27 |
Max. Drawdown |
-11.32% |
-13.28% |
Table 2. Performance summary for the double 10s strategy with 1-day exit.
Although the equity performance in the first time period is attractive, with only one losing year out of 8 in total, in the second time period we observe degradation of performance with 2 losing years out of 5, a marginal profit factor, a very low Sharpe ratio and a much lower CAR. These observations also point to the possibility that the parameters of this system are suitable for market conditions encountered before 2008 and, although the overall return is positive after 2007, if more time is allowed the strategy may turn to a loser.
3. The Double 10s strategy with moving average exit
This is a long/short counter-trend/momentum strategy, defined as follows:
- Buy at the close if price makes a 10-day low and exit at the close when price crosses above its 10-day moving average or when a sell signal is triggered.
- Sell at the close if price makes a 10-day high and exit at the close when price crosses above its 10-day moving average or a buy signal is triggered.
Continuation and reversal signals are also allowed in this case, in the sense that when a long or short signal is closed another long or short signal can be opened if the conditions remain in effect and short signals are initiated at a long exit bar if conditions are right and vice versa.
Below is a graph of the portfolio equity from 01/04/2000 to 12/31/2007:
Below is a table of the monthly returns for the same time period:
Next is a graph of the portfolio equity from 01/02/2008 to 11/08/2012:
Below is a table of the monthly returns for the same time period:
A performance summary for this strategy is shown in Table 3 below:
Parameter |
01/04/2000-12/31/2007 |
01/02/2008-11/08/2012 |
CAR |
12.16% |
3.47% |
Win rate |
65.92% |
62.61% |
Trades |
625 |
353 |
Profit factor |
1.59 |
1.11 |
Sharpe ratio |
0.87 |
0.25 |
Max. Drawdown |
-23.49% |
-25.98% |
Table 3. Performance summary for the double 10s strategy with moving average exit.
Here we also have a case of excellent performance in the first time period before 2008, with only one losing year out of 8 but with high drawdown, that can be a random event, however. Also in this case, in the second time period after 2007, we observe degradation of performance with 2 losing years out of 5, a marginal profit factor, a very low Sharpe ratio and a much lower CAR. These observations also in this case point to the possibility that the parameters of this system are suitable for market conditions encountered before 2008 and although the overall return is positive after 2007, if more time is allowed the strategy may turn to a net loser.
A simple trend-following strategy
A simple, long-only, trend-following strategy is defined as follows:
- Buy when the 50-day moving average crosses above the 200-day moving average.
- Sell when the 50-day moving average crosses below the 200-day moving average.
Below is a graph of the portfolio equity from 01/04/2000 to 11/08/2012:
Next is a table of the monthly returns for the same time period:
A performance summary for this strategy is shown in Table 4 below:
Parameter |
01/04/2000-12/31/2007 |
01/02/2008-11/08/2012 |
CAR |
4.55% |
3.80% |
Win rate |
73.33% |
72.73% |
Trades |
15 |
11 |
Profit factor |
3.72 |
2.87 |
Sharpe ratio |
0.48 |
0.40 |
Max. Drawdown |
-13.36% |
-18.80% |
Table 4. Performance summary for the simple trend-following model
In contrast to the three counter-trend strategies analyzed before, it may be seen that the performance parameters of this trend-following strategy have been much more stable for the two time period, as shown in Table 4. More importantly, this is a trend-following strategy with high win rate, something that goes against the widespread (false) notion that this type of strategies must have low win rate, in general. I frankly do not know where that came from but this is another example of how vague statements can become rules if repeated many times without a sanity check. The reason that many trend-following strategies have low win rate is that they are optimized. This is also the reason that most managed futures funds suffer losses, i.e. the fact that they employ optimized strategies, often involving several parameters. The more parameters available for tweaking, the smoother the equity curve obtained during the design phase but the probability for a future failure will be higher. Also, a low the win rate increases the risk of ruin and contributes highly to performance degradation.
Discussion of the results
In this article I analyzed the performance of three simple counter-trend strategies for ETFs that became popular about 5 years ago. I showed that since the time these strategies started enjoying widespread use, mainly by retail traders, their performance has degraded, although they are still net marginal winners. In contrast, a simple trend-following strategy has shown robust performance without the application of any advanced filters and modifications. This is expected because trend-following is supported by common sense while counter-trend/momentum trading is based on gambler's fallacy. However, in the markets, if it takes a fallacy to make money, that is good enough provided performance is robust and does not hide surprises. I am hesitant to declare the three systems analyzed in this article curve-fitted and the optimization of their parameters as the source of performance degradation. There could be some other reasons in addition. Below are three possible reasons:
- The strategies were curve-fitted and optimized.
- Reflexivity (Soros): The strategies became very popular after they were published and actually caused their own performance degradation.
- The strategies were random, an outcome of selection bias via some (random) process that designs strategies (including the human brain).
I do not have an answer but I believe that in most cases a combination of the above reasons is the ultimate cause of the observed performance degradation.
Disclosure: I have no positions in any stocks mentioned, and no plans to initiate any positions within the next 72 hours. I wrote this article myself, and it expresses my own opinions. I am not receiving compensation for it. I have no business relationship with any company whose stock is mentioned in this article.
Additional disclosure: Charting and Backtesting program: Amibroker. CFTC RULE 4.41 – HYPOTHETICAL OR SIMULATED PERFORMANCE RESULTS HAVE CERTAIN LIMITATIONS. UNLIKE AN ACTUAL PERFORMANCE RECORD, SIMULATED RESULTS DO NOT REPRESENT ACTUAL TRADING. ALSO, SINCE THE TRADES HAVE NOT BEEN EXECUTED, THE RESULTS MAY HAVE UNDER-OR-OVER COMPENSATED FOR THE IMPACT, IF ANY, OF CERTAIN MARKET FACTORS, SUCH AS LACK OF LIQUIDITY. SIMULATED TRADING PROGRAMS IN GENERAL ARE ALSO SUBJECT TO THE FACT THAT THEY ARE DESIGNED WITH THE BENEFIT OF HINDSIGHT. NO REPRESENTATION IS BEING MADE THAT ANY ACCOUNT WILL OR IS LIKELY TO ACHIEVE PROFIT OR LOSSES SIMILAR TO THOSE SHOWN.