In this post, we ask the question “Does technical analysis work?”. Practitioners swear by it, but academics deride it. I have to confess, I too am not a believer, and as we tackled this piece, we started our approach through the lens of the skeptics. Did I change my mind in the end? Read on to find out.
Technical analysis (TA) is a price movement forecasting method that uses past prices, volume, and open interest (for the future markets). Clearly, practitioners of technical analysis do not believe the warning labels of “Past performance is not indicative of future returns.” They believe that price and volume movement embeds all other information about sentiment, interest rates, macroeconomic conditions, company cash flow, and other factors.
If that sounds too implausible, recall that this is the same belief as the efficient market hypothesis (EMH). The irony is proponents of EMH typically look down on technical analysts. In his seminal book, A Random Walk Down Wall Street, Burton Malkiel (a leading proponent of EMH) compared the practice to alchemy. He wrote:
“Chart derived from random coin tossings looks remarkably like a normal stock price chart and even appears to display cycles.” — Burton Malkiel
But really, who could blame Malkiel? After all, one of the fathers of technical analysis, William Stanley Jevons, a British economist from the late 1800s, believed that commodity cycles (typically lasting 10 years or so) were attributed to sunspot activities ?…
Jevons’ work inspired Charles H. Dow (of the Dow Jones Index and Wall Street Journal) to develop his own theory that looks at market cycles, the Dow theory. This was the beginning of the adoption of technical analysis in the US1.
Despite derision, the efficacy of technical analysis has been the subject of much academic study. So, in order to find our answer, we began our quest with academic literature.
Ok – we will not bore you with the complete survey of the literature. But if you’re interested, you can read this and this for a more complete review. The key takeaway is that TA works – sort of, with several important caveats:
The first is the diversity of analyses approach. TA includes a variety of forecasting techniques such as chart analysis, pattern recognition, and various computerized trading methods. Some of the chart analysis and pattern recognition methods are subjective. What is heads and shoulders (not the shampoo mind you) to one person, may not apply to another. This subjectivity is the source of much academic criticism (as Malkiel stated above). For what it’s worth, a study done by Hasanhodzic et al in 2010 shows that human subjects can indeed differentiate between actual market returns vs. randomly generated charts – contrary to Malkiel’s belief.
Therefore, in order to minimize the impact of this pattern of subjectivity, most academic research on Technical Analysis is limited to what can be objectively defined mathematically.
The second is that the approach can be applied to various markets, with varying degrees of success. In 1995, the Federal Reserve Bank of New York published a paper titled “Head and Shoulders: Not Just a Flaky Pattern” (even though the Fed comprises serious economists, they have jokes too!). It studied the predictive power of the head-and-shoulders in the FX market from March 1973 to June 1994. Here’s the chart from the Fed paper (Figure 1).
Results were mixed. The approach has predictive power for the German Mark and the Japanese Yen, but not for the Canadian Dollar, Swiss Franc, or the UK Pound. The reason for this discrepancy is unknown. One could argue that, at the time, the Mark and the Yen were developing market currencies, and therefore there was more information asymmetry. Another potential explanation could be that it’s a self fulfilling prophecy – if a lot of people were using the same signal, they would buy and sell at similar times.
Now, before you go out and use your heads-and-shoulders and start trading FX, know that the vast majority of retail FX traders lose money.
The third is diminishing alpha. As you would expect, any advantage that one might have from certain techniques would decay over time as more people become aware of them. Numerous studies indicate that getting alpha (an excess return over the market) from technical strategies alone is getting increasingly difficult. One study found techniques that may have worked for US stocks until the late 1980s seemed to have stopped working after. Others also observed technical trading profits having declined or disappeared in recent years for the FX market.
In summary, the general theme of the survey is that using the TA approach as the sole tool is unlikely to work, especially for more complex markets such as equity markets. As such, professional practitioners overlay TA strategies with others (fundamental or value), creating fusion strategies.
Ok – it’s time to take our heads out of our books and time to put this to test with our own methods. For the next section of this article, we will be using AlphaScreener, a screening and backtesting tool we just launched (try it free here!).
Can we create our own fusion strategy? And can we show, via backtest, that adding a technical signal to the strategy improves the hypothetical returns? Let’s find out.
This hypothetical strategy was developed by a Vested team member who used to be a commodity trader (please note that this is not financial advice – the results presented here are hypothetical). The thinking is to develop and backtest a strategy that focuses on quality, using the following filters:
In other words, profits and cash, which are especially important in the current higher interest rate regime (something that we discussed at length here).
When you set it up in AlphaScreener, the screener setup looks as shown in Figure 2 below.
Note: AlphaScreener is best accessed via a laptop/desktop. The user interface is not optimized for mobile devices.
We then take the screener and conduct a backtest. We do this by taking the screener criteria above, starting in the year 2000, filtering for companies that existed then and met the criteria, holding the portfolio for one year, and re-screening and rebalancing once a year from the year 2000 to date. In this hypothetical scenario, we choose only to invest in the top 25 companies that met the screening criteria. Note that AlphaScreener does all this work for you in the background, as part of the Evaluate (backtest) function.
In Figure 3 below, you can see the snapshot of the backtest results.
The results are not bad. A hypothetical portfolio started ~22 years ago would’ve handily beat the S&P 500. But if you look closer, almost all of the outperformance occurred in the 2000 – 2010 decade. In the past 10 years, the strategy underperformed the S&P 500 (you can see it clearly from the year-over-year chart, at the bottom of Figure 3).
There are several takeaways from the chart above:
Let’s see if we can improve the screen by adding another filter, a technical filter, to the quality screen above.
Let’s call the original quality screen above “quality-A” and the new one “quality-B”. In quality-B, we added a 50-day weighted moving average (WMA) to the screener, as shown in Figure 5 above. The result of the backtest is shown below (Figure 7).
As you can see above, the outcome of the hypothetical portfolio is improved.
A side note on base rates: During backtesting, AlphaScreener simulates multiple portfolios at different start dates and holding periods. The system aggregates the returns of these hypothetical portfolios, compares them with the performance of the S&P 500 benchmark over the same time period, and compares the two performances. It then counts the rate at which the different hypothetical portfolios beat the benchmark. This is called the Base Rates.
One can calculate base rates over different holding periods (1 year, 3 years, 5 years, 7 years, and 10 years). For illustration, let’s discuss the 3-year Base Rate.
The 3-Year Base Rate is the percentage where the hypothetical portfolio created by your screen beat the S&P 500 benchmark over the past 20 years. For example, if the 3-Year Base Rate is 80%, the odds of you beating the S&P 500 benchmark (had you invested in the strategy defined in the screener, starting on some random date in the past 20 years and holding for 3 years) over the same time period is 80%.
Figure 7 above compares the Base Rates of quality-A (dark blue) vs. quality-B (light blue). For quality-A, if you invest in the hypothetical strategy starting anytime in the past 20 years and hold for a year, the odds of you beating the S&P 500 is about a coin flip (the Single Year Return Base Rate is ~50%). Contrast this to quality-B, whose Single Year Return Base Rate is 64% (or 1 in 3 chances of outperforming the S&P 500 in any given year in the past 20 years). More importantly, quality-B’s hypothetical return robustness is maintained as the hypothetical holding period gets longer, unlike that of quality-A’s.
In quality-B, the addition of the 50-day WMA helps to pinpoint quality companies that are trending up. In this example, we used the 50-day WMA as an indicator of the underlying momentum trend. If the share prices remain above the 50-day WMA, it means the share prices of these quality companies might be trending up, and one can hope that the momentum will continue to build. This would lead to multiples expansion, which will then be accompanied by share price increase.
Notice that the WMA filter is used only at the end. In AlphaScreener, the filters are applied sequentially. In this example, we have applied three fundamental screeners (Net Profit Margin, ROIC and Dividend Yield – in that order) to limit our observable universe to quality companies. If you reverse the order and place the 50-day WMA first, the results wouldn’t be as favorable. You could be screening for companies that are bad (and therefore are justly priced below its momentum trend by the market) and trying to find quality companies within that list.
Maybe, but with a lot of caveats…
As with many things in life, the answer to the questions is more complex and nuanced than we originally anticipated. There are a lot of caveats to consider when considering the efficacy of TA. That said, one thing is for certain, the plethora of classes that promise to sell you technical techniques that consistently beat the market are probably lying to you.
1The modern practice of technical analysis stemmed from the works of Joseph de la Vega, in 17th century Amsterdam, and Homma Munehisa, in 18th century Japan (who developed the candlestick charting approach). But it was not until the late 1800s that these techniques were popularized in the US by Charles H. Dow (of the Dow Jones Index and Wall Street Journal). Dow introduced the Dow theory, although he did not use it for trading.