A nice, little exercise to check whether MACD and RSI could be used in cryptocurrency markets to generate excess returns
Technical analysis is, especially in Web 3, a mainstream analytical tool for individual cryptocurrencies despite the apparent hatred on their existence. Decades of papers have been written on the effectiveness of the technical indicators, on whether they could be utilized to extract alpha; yet, their effectiveness is hotly debated until today.
“Hotly debated” does not mean that technical analysis is an unfounded belief, it means that technical analysis has duality — sometimes it works, sometimes it does not. It works in LSE, but it might not work in NYSE.
There has been an extensive research on the topic in the traditional finance sector, so I would like to go just one step further, and apply the same methods to two major cryptocurrencies to check whether the technical indicators are effective in Web 3 market.
1. Related Studies
Chong et al. (2014) showed that trading rules with MACD and RSI oscillators can generate excess returns to buy-and-hold strategies, and will be the foundation for this little exercise.
Just to be fair, let me mention that Yamawaki et al. (2007) showed that MACD and RSI are pretty much useless in predicting price movements, thus have no significance whatsoever.
I only use two coins, BTC and ETH, for this exercise, because they have the longest time series, and they are the two most important coins currently active. I have Closing price from 2014 till now for BTC, from 2017 for ETH.
The trading rules are the followings:
- MACD (12, 26, 0) : Buy when EWA(26)－EWA(12) crosses 0 from below, and vice versa
- MACD (12, 26, 9) : Buy when EWA(26)－EWA(12) crosses EWA(9) from below, and vice versa
- RSI (7, 50) : Buy when RSI(7) crosses RSI(50) from below, and vice versa
- RSI (14, 50) : Buy when RSI(14) crosses RSI(50) from below, and vice versa
- RSI (14, 30/70) : Buy when RSI(7) drops below 30, vice versa when it rises above 70
I will compare the 10-day returns for Buy & Hold strategy to 10-day returns for the above trading rules, and use t-statistics to check if 10-day returns of trading rules are greater than those of Buy & Hold (B&H) .
For BTC, all trading rules generate significantly less returns than B&H does.
For ETH, it is a bit different. Overall p-values are a lot smaller than BTC’s, and for some strategies — RSI (7, 50), RSI (14, 50) — technical indicators that look like it could generate some excess returns.
I subset BTC into BTC (2017+ or since 2017), as I figured a different timeline and the maturity of the market might have played a role.
For BTC (2017+), the result was a lot more similar to ETH’s; RSI (7,50) and RSI (14, 50) were the most significant ones, and overall had much better p-values than BTC (2014+).
However, basically no trading strategies suggest above generating meaningful returns under 5% significance level. In fact, most trading rules have significantly less mean returns than simple B&H returns.
BTC (2017+) and ETH data might show that RSI could be used as good technical indicators, but we need to remember I had to artificially massage the BTC data to get the results.
My findings are very coherent with the past studies on the TradeFi market. Technical analysis possess duality — sometimes it works, sometimes it does not.
Some technical analysis genius might have earned a fortune with her analytic skills, but it might be simply due to her being in the right time for technical analysis, or might be due to her animal spirit irrelevant of technical indicators.
No matter what, that genius would have never generated such returns based on the technical indicators. She needed the right timing or her own instinct.
The lesson here is, you should be really high in the following technical indicators, because well, you don’t know if it works this time or not.
New to trading? Try crypto trading bots or copy trading