This thesis research how technical analysis performs on cryptocurrencies. Some basicmethods were tested on multiple currencies in different categories. The tested currencieswere not the most popular ones because there were too much data on the most popularcurrencies for the scope of this research. Basic methods were tested on some most usedsettings. Every technique was tested alone. After testing, methods were compared witheach other, and also different currencies were compared.
The reason why only those three currencies were profitable is unknown. A closer investigation of profitable currencies revealed no good reason Bollinger bands made profitable trades with them. Later on in this research, those currencies are analysed more precisely. A common thing about those currencies was that their values rose and dropped on the researched timeline. Most of the currencies' values were much higher at the beginning of the timeline than at the end. Interestingly, Bollinger band made the best profit along all the tested technical analysis methods on those three currencies. On the Bollinger bands setting, the moving average type has three options: 0 for simple moving average (SMA), 1 for exponential moving average (EMA) and 2 for weighted moving average (WMA). Slight differences exist between moving average types, but differences are not significantly different. 6.2.3 EMA Period Average Median Max
id: c21bf24f235b5a88ec009e23d8359a57 - page: 37
Min 5 -42446,57 -47647,84 72340,53 -95647,66 37 10 -41178,43 -46098,41 53236,11 -94013,05 20 -42110,12 -46439,36 59937,08 -94287,17 50 -45834,52 -56944,43 30188,12 -95959,21 100 -46254,45 -49259,79 67140,54 -97292,69 200 -45508,95 -49906,20 66373,00 -95576,14 Table 4, EMA performance in different periods EMA was profitable with XMR, DRC and partly on lit. On EMA settings, there were no significant differences on different settings. On the median, period 50 worked much worse than others, but the average profit was as bad as in every other setting. The worst performances were relatively the same in all settings. Based on this research, there is no difference between the number of periods. Five to twenty periods seem to perform slightly better than a hundred or two hundred, but the difference is insignificant. 6.2.4 MACD fast slow signal average median max
id: 016c80c6af18c7692470962dec7b97e5 - page: 37
Table 5, MACD performance on different settings In this research, MACD was run by giving three numbers: fast period, slow period and signal period. The fast period is the number of periods used for calculating fast EMA. For more sensitive MACD lines, fast periods should be short. A slow period calculates long-term price trends. It would have been better to research also short-term price changes and signals alone in this research. That would push the scope of the research even further, and on this scope, both were researched simultaneously. That is not good for performance, but overall performance did not 38 differ much from other technical analysis methods. On MACD, there is a significant difference in performance on different settings: Average profit is -from -40% to -56%, and median profit is from 43% to -60%. MACD was profitable only on two settings, 21,55,8 and 18,35,6. There is no logic
id: 78fe0ddd6f9ed5eed1078950aee839a7 - page: 38
It is impossible to say what is the best setting. According to this result, luck affects the result more than the setting itself. It does not matter the value of fast, slow or signal setting. This research does not reveal which is the best MACD setting. 6.2.5 RSI period average median max min 7 -39149,99 -44802,93 28827,69 -90105,99 14 -31573,09 -37350,86 31714,22 -86443,27 21 -17623,23 -17354,58 44398,95 -77578,38
id: c78139ef66dece79df579c9553958289 - page: 39