Last month, I shared some insights on One Click Crypto Hub about our AI bots’ daily performance, raising the question of how good a trading bot is at beating the market daily.
Our further research concluded that the topic is more sophisticated than it first looked at, and a more in-depth data analysis is required. In the upcoming October report, we will share in detail the final results, including how we calculated everything.
For now, here is what the hypothesis looks like:
‘One Button Capital AI trading bots are not just capable of outperforming the underlying asset over the long run. The bots also increase their efficiency in time, essentially increasing its daily dominance over the market in an uptrend fashion.’
Market outperformance of the most profitable Performer v2 bots (ETH, BNB, CAKE, UNI, CRV, ETC, FTM, LTC) indicates an obvious uptrend over time.
The day-to-day analysis of Performer v2 against Ethereum shows that the AI is outperforming the asset on more days than it is not. That is not raw daily profit/loss but how much the bot outperforms (or underperforms) the market on any given day.
A scattered plot table with the same data can help us indicate the outliers more easily. Traditionally we can ignore the noise or data that is too far from the median, but in finance, this is not the case.
The two +15% outperformance days of the AI that were back-to-back in January can be the difference between having an investor’s portfolio go down by -27.75% in two days or not. The highly concentrated data between -0.5% and +0.5%, on the hand, can be hypothesized as less relevant, but more research is needed.
In order to find out if there is an increase in daily outperformance, we used several statistical methods, including a daily running average that is shown in the table below:
The market outperformance for all these pairs ranges between +0.11–0.36% daily. For the longest-running bots, we can see that the daily outperformance is consistent in the 0.21–0.24% range.
Furthermore, we tried to indicate a runtime relationship to daily outperformance using the trading data by utilizing two different models for the average performance.
While the results are fascinating, the R-squared values of these two tables are way below the needed 0.7 for financial data. Also, there is one fundamental mistake in this type of work with percentages that almost everyone gets wrong. For this reason, we are working on the development of a dynamic model that catches the ‘real’ amplitude in daily outperformance.
The best way to prove the hypothesis is pretty simple and can be made by just looking at a chart of an AI bot’s and the market’s performances. The area or distance between the two lines is the actual indicator of how proficient a bot is, compared to the asset (or set of assets).
Despite the overall downtrend of the market, it is evident that the bot has gained more edge over time. Therefore it is logical to hypothesize that this does not just come from a static daily outperformance rate but rather an increasing one.
That is what we described as shadowing/mimicking in our research as one of the key behavioral patterns behind an AI’s success in outperforming the market.
A more reliable way of working with data (with an R-squared value of 0.8121) is the observation of this area and its trend. The test on Performer v2 versus CAKE shows a positive trendline in favor of our hypothesis,
It is of utmost importance to be very cautious when working with data and financial numbers. Many companies have misled their investors and followers by presenting wrong analyses, and we do not want to do that.
We regularly prepare insightful reports and case studies about crypto trading and the blockchain industry.
We sent you a link to complete your sign-up.
Check your inbox, verify your email, and unlock all functionalities of your OB Trader account.
You were added to our waitlist. You will get an email within 3-5 days If you are shortlisted.