AI Dramatically Outperformed HODL This Summer [Cryptocurrency Research Analysis]
2018 has been a difficult year for crypto players, believers, and especially HODLers. Cryptocurrencies depreciated by a huge percentage since January this year. Ether for example, dropped by more than 60% since January 2018. But even in the worst down market, AI based crypto trading still led to profit.
Here we show the power of AI even in severe down markets and even if the signals are interpreted in the least risk-averse way of trading signals (as opposed to portfolio management). For that purpose we will take Ether prices and Ronin AI analysis between June 25, 2018 and August 30, 2018. Within that time frame, ETH dropped by -38%, representing one of the largest declines in the entire year of 2018.
The reason this timeframe is collected is due to the fact that it is very hard to see any opportunity to generate profit since the market has been sideways followed by a sharp decline as shown on the chart below. At the same time, the chart below also represents performance of Ronin AI. We posted 5 versions of Ronin AI performance based on the minimum sufficient probability of success.
For example, AI Above 50% portfolio shows that Ronin AI would only trade signals with above 50% probability of success. Similarly, AI Above 70% portfolio indicates that Ronin AI would go long ETH if the signal has a probability of 70% or above and would not enter the ETH market if the probability of success if less than 70%.
We can clearly see that Ronin AI easily outperforms HODL with the following results in the table below. It is also important to take into consideration risk-adjusted return, or Sharpe Ratio, also presented in the table. The higher the Sharpe Ratio the higher is the risk-adjusted return. It is pretty evident that AI Above 60% portfolio shows both highest return of 34.3% in just two months accommodated by the Sharpe Ratio of 8.8.
But how is it possible to make 34.3%, while ETH was down -38.0%? Is it even real?
We analyzed data and AI predictions for a total of 93,375 minutes between June 25, 2018 and August 30, 2018, representing 64 days and 20 hours of data. Turns out that the distribution of returns is actually slightly negatively skewed with more minutes finishing up than down. AI simply takes advantage of such distribution accurately predicting ETH direction.
The histogram below shows distribution of predictions by Ronin AI, indicating that AI correctly assigned probabilities. The total number of predictions with probabilities above 50% is higher than the total number of predictions with probabilities below 50%.
Referring back to the chart of HODL versus Ronin AI portfolios it is important to conduct necessary risk management analysis and decide how risky an investor wants to be. Higher returns are always associated with higher level of risk and vice versa. Despite lower returns, only trading signals with above 80% and 90% predictions provide a lot higher level of portfolio stability and minimum amount of drawdowns, thus having a positive effect on investor’s emotional well being.
All of these factors should go into consideration when deciding which strategy to follow based on Ronin AI signals. But the final conclusion is that Ronin AI models do work even in bearish markets. Within short timeframe AI is able to beat the market and gross positive returns even in the most bearish market stages.