This article won't be to everyone's taste, especially those sitting over price charts of forex, stocks, or cryptocurrencies for years and listening to self-proclaimed trading "gurus". One little secret to start with: Technical analysis does not bring any competitive advantage to the trader. Its results are random. Don't believe it? Then read on.
We also started with technical analysis and charting before we understood it was bullshit. The great danger is human bias and greed, which is strongly manifested in technical analysis. Why? Because we want to see repeating patterns with evident predictive power. But that is just a wish.
We're not just talking to the wind. We're presenting a study here with a thorough analysis. You'll eventually get the message and start pursuing sophisticated investing in the form of a diversified buy-and-hold strategy of various underlying assets.
We will go through the most common technical analysis ("TA") strategies, using some oscillators, moving averages, price patterns, and so on. Many can justify TA similarly: “But if you combine this pattern and that oscillator with those two moving averages, and only on 48-minute bars, and only on those markets… Actually, that is just an overfit, nothing more. The strategy was half the bias and the beliefs. Can they prove that this kind of nonsense was working historically or better statistically? Can they explain why it should work? In the case of TA, your eyes are lying to you.
Your mind gives your eyes orders to see the patterns you believe are functional. But when the same pattern doesn’t work, your eyes don’t even see it because it looks totally different on the chart.
Moving Average Crossovers
Everyone who did something with technical analysis has heard about moving average [MA] crossovers. Simple logic for trend-following strategy. You have two MAs; one is more adaptive (calculated from fewer bars). When one crosses another, there is a change in a trend. The most common are 20-50 and 50-200. It can be applied to daily data, hourly, weekly, or anything. It looks like this.
An example of moving average crossover (50-200). Circles represent buy and sell signals on MA crossover, and triangles are the buy/sell price.
We tested this common strategy like no one showed before. We used more than 7000 different time series! We analyzed 2926 stocks and 686 ETFs on daily and hourly data. Daily data are calculated from 2006 and hourly from 2015.
Let’s look at the results on daily data. On all backtests, we applied a low cost of 0.15% per trade (open and close of position). In the following performance plot, you can see the performances of different stocks over time. Percentage values are calculated according to the initial capital. We don’t reinvest and always invest the same amount of capital. That’s why you can see even values lower than 100%. This gives you a better perspective than reinvesting. With reinvesting, you go for astronomical gains compared to the initial capital. You continually invest a lower amount with losses, so theoretically, you cannot go bankrupt and trade even with 0.1% of your capital. So we always invest the same amounts to see how the strategy works.
Results of moving average crossover for stocks. Returns are without reinvesting
Some basic statistics: the average Sharpe ratio is -0.02, the average annual return is 0.44%, and the median annual return is -0.87%, so more than 50% of stocks finished with a negative return. The average winning ratio (calculated on trades for each stock) is around 35%, which is typical for the trend-following strategy. Almost 46% of all stocks finished below the yellow line, so your trading account would have vanished. As you can see, the results are random. When you see that someone claims it works, it is just selection bias. If someone says you have a short MA of 23 days and a long MA of 78 days, he is just over-fitting. You can find examples that worked perfectly, but they are random.
Suppose you choose only the top stocks based on history. Let’s say up to 2014 [your in-sample data, IS], and from 2014 you have paper trading results [out-of-sample data, OOS]. Of the top 100 stocks selected on IS, only 11% had at least a ¾ Sharpe ratio of IS on OOS. The average overall return on IS was 230%, while on OOS was only 11%. Look at the top 15.
Top 15 stocks from the moving average crossover example selected on the IS (until 2014, red) to see the results on the OOS (blue)
For intraday traders, the results are similar (as expected). No more words are needed here. We will not show the results on ETFs since they have identical conclusions. The results on ETFs are similar, only less volatile because most of them consist of many stocks, and their volatility is much lower.
RSI Strategy (Relative Strength Index) with Moving Average
In this case, we will look at the results quickly because it is not worth your time to bother you with the details. Relative Strength Index, according to its believers, shows oversold and overbought stock price levels. Yes, you have read it correctly! Actually, the RSI is just a non-linear combination of positive and negative price differences over a period of time, respectively, the ratio of the moving average of price differences. RSI indicator takes values between 0 to 100, and values over 70 are considered overbought and less than 30 oversold. So the primary strategy is mean-reverting. When the RSI indicator is in extreme value, it is expected to go back to a value of 50. The funny thing is that strategy can also be momentum – when we reach the overbought limit, we expect the momentum to continue. So some traders use it as momentum, others as mean-reversion, yes, two opposite strategies. And many of them claim on YouTube how greatly it works.
Many trading gurus claim you have to mix it with other indicators or price patterns so it functions better. We developed a simple strategy: we enter a long position only if the RSI shows oversold and the stock price is below the 50-day EMA (exponential moving average). Similarly, we go to the short position when the price is overbought according to RSI and over 50-day EMA. We used RSI constructed on the last 21 trading days.
Similarly, as in the previous analysis, the performance is calculated according to starting capital without reinvesting, so we continually invest the same capital. That's why the value can go below -100%. In this case, we show results on ETFs. There are no statistics needed, and the performance says it all.
Results of RSI+MA strategy for ETFs
So what do we conclude? If you use technical analysis, ditch it. And if you are currently doing well with it? Ditch the random trading and do something useful that will add value to you.