Robotics and AI Trading Strategies (BOTZ ETF Backtest – Rules, Settings) Last Updated on July 4, 2023 The advent of Artificial Intelligence is here, and is forecast to combine with and revolutionize many industries in the years to come. One such combined industry is ‘Robotics and AI’, and traders are increasingly seeking ways to position themselves to make use of this trend as adoption rates of robotics and AI increase. What can we do, if we want to trade robotics and AI trading strategies? In this article, we use optimization to find profitable robotics and AI trading strategies (for the AI industry). Table of contents: Global X Robotics and Artificial Intelligence ETF (BOTZ) As of yet, there are few good trading vehicles available for traders wishing to participate in the robotics and AI industry. Trading the stocks of startup companies is possible, but these usually have a very short price history and are very volatile, making them unsuitable for trading and besides, most IPOs fail. In order to find a viable trading edge, we need to find a way to trade the industry as a whole. This is where exchange-traded funds (ETFs) come into play. An ETF contains a variety of assets, and can thus track an entire industry by including many different assets from that industry. The Global X Robotics & Artificial Intelligence ETF (ticker name BOTZ) is an ETF that invests in companies that stand to benefit from increased adoption of robotics and artificial intelligence. The BOTZ ETF was founded in 2016. Contrary to many new robotics and AI startup companies, BOTZ now has enough price data and trading volume for us to start designing a functioning trading strategy around it. Donchian Channels Strategy Overview The BOTZ trading strategy in this article uses the indicator from our Donchian Channels article. We recommend you read that article for further details on the strategy itself. Donchian Channels (DC) is normally a trend following indicator, with a length setting of 20 (for example). In this article, we experiment with using short-term settings that are more appropriate for the price fluctuations in BOTZ, turning it into something closer to a momentum indicator. Exponential Moving Average (EMA) The second component of our BOTZ trading strategy is to use an Exponential Moving Average to reduce the number of false signals we encounter. Our previous article on the EMA goes into details on its calculation and use cases. In this article we experiment with various long-term settings, and use the EMA as a complement to filter bad trades from our short-term Donchian Channels strategy. BOTZ Strategy Trading Rules The trading logic and rules for the BOTZ Donchian Channel/EMA Strategy is as follows: When the intra-day price crosses above the upper band of the Donchian Channels, and the price is above the EMA, we go long (buy) immediately. When the intra-day price crosses below the lower band of the Donchian Channels, we close our position (sell) immediately. This is a long only strategy, where we only take long (bullish) trades. Displayed on a chart the trading strategy will look something like this: The green and red lines are the Donchian Channels. The blue line is the EMA. Blue arrows show where long positions are opened (‘ChBrkLE’), while red arrows show where positions are closed (‘Exit Long’). BOTZ trading strategy optimization Optimizing a trading strategy is an important step in trading strategy development, but not for the reason you might think. Optimization in trading, which is the process of testing different input values for each of our variables, is not about finding the perfect value. Instead, it is mainly about testing the robustness of our strategy. What we’re looking for is a strategy that holds up well even when we change the values and settings across a fairly broad range, without our results taking a big hit. If our backtesting results change significantly when the number is only varied slightly, it is likely that our strategy is curve fitted. If it holds up well under changing variables, it indicates that our strategy is more robust. In the BOTZ Donchian Channel/EMA Strategy, we have only two variables: The length of the EMA, and the length of the Donchian Channels lookback period. Thus, this a rather simple model. Let’s start by comparing the results for various Donchian Channels length inputs, without any EMA filter. We have backtested on BOTZ (on Nasdaq), on a daily timeframe, from September 13, 2016 (the furthest back the BOTZ ETF goes) until today: DC length No. of trades Win Ratio Profit Factor Max Drawdown Avg. trade 2 217 49.31% 1.618 22.24% +0.60% 3 157 49.04% 1.527 38.58% +0.70% 4 116 46.55% 1.710 30.26% +0.92% 5 100 46.00% 1.453 34.91% +0.76% 6 88 46.59% 1.272 44.41% +0.65% 7 76 42.11% 1.220 42.08% +0.64% 8 62 45.16% 1.680 30.30% +1.51% 9 54 46.30% 1.863 23.64% +1.86% 10 47 51.06% 2.348 26.70% +2.29% There’s no need to test with higher inputs right now, as the number of trades would be too small to provide us with meaningful results. This is because BOTZ has only been around since 2016. As time goes on, we’ll accumulate enough data to backtest these higher length settings as well. The key observation here is that nearly all settings lead to results that are at least profitable on paper (Profit Factor 1.25 or above), indicating that this part of our strategy is reasonably robust. For now, we’ll set our Donchian Channels length to 3. This is not the optimal result we found, but is still profitable, and has a decent win ratio. It also gives us a decent number of trades (157) to work with before the EMA filter reduces the number further. Let’s continue by comparing the results of Donchian Channels length 3 with various EMA inputs: EMA No. of trades Win Ratio Profit Factor Max Drawdown Avg. trade No EMA filter 157 49.04% 1.527 38.58% +0.70% 25 112 49.11% 2.228 11.98% +0.94% 50 103 51.46% 2.310 10.89% +0.99% 75 99 51.52% 1.915 18.20% +0.86% 100 95 53.68% 1.954 17.85% +0.87% 150 93 53.76% 2.134 11.41% +0.94% 200 87 52.87% 2.069 11.93% +0.89% 250 87 54.02% 1.716 13.28% +0.66% 300 83 46.99% 1.381 21.47% +0.44% If we compare this to the results of using no EMA filter in the top of the table, we can clearly see that using the EMA as a filter improves our results across the board using almost any length setting. This in turn indicates that this part of our strategy is also reasonably robust. While it is possible to scan through every input number to find the optimal one (we found it at EMA145), this is a mostly pointless exercise, as all you’re doing is curve fitting. It is highly unlikely that this exact number will continue to be the optimal input in the future. We’ll stick with EMA200 for now, as this is a common setting used by many traders. This is again not the optimal result. Choosing relatively underperforming inputs in this way helps increase the realism of our backtesting results. BOTZ Donchian Channel Strategy Backtest Equity Curve We can now view the equity curve for the BOTZ DC/EMA Strategy we just created, using a Donchian Channel setting of 3 and EMA 200 as inputs. Overall, the results look good. Our strategy also beats Buy and Hold (blue line) by a good margin, and with much less volatility. The BOTZ DC/EMA Strategy exhibits many of the traits we value as traders: Good Win Ratio (Percent Profitable), which is above our 50% target. Good Profit Factor. Should preferably be above 1.75, which it is. Good Max Drawdown. It should preferably be below 25%, which it is. It is even low enough to allow for some Leverage, if we are seeking ways to increase Net Profits. The strategy so far has a relatively low total number of trades (87). This is again due to BOTZ only being around since 2016. This makes our equity curve less statistically reliable for the time being, but we could also potentially have found an early entry point into a profitable strategy. This is important, as all strategies stop working eventually. The small number of trades increases the chances that our strategy could be Curve Fitted. If you are considering adopting this strategy in some form for live trading, please remember that nothing on the website is investment advice. This article serves just as an example. We strongly encourage you to do Out-of-Sample Backtesting. The best way to do this, especially considering the small amount of data we have so far, is to conduct an Incubation Period. BOTZ Donchian Channel Short-Term Strategy? We will round off this article by simply noticing that our BOTZ DC/EMA Strategy also seems to have good results on lower time frames. The following are the results for the 2-hour time frame: Here we observe a very smooth equity curve, with results that are nearly as good as those on the daily time frame. Have we found a good Short-Term Trading Strategy in BOTZ? Maybe, but we should be wary of drawing any such conclusions too early. Short-term trading is difficult, and most good trading opportunities are found on long-term time frames. The above equity curve should serve mostly as food for thought, and something to pay attention to into the future. Only if it proves itself through out-of-sample backtesting during an incubation period, should it be considered as a viable short-term strategy.
%% MOST likely a very good top trender ,assume volume stays 7 figures+/; even though my don channels look much different from yoursLOL YTD its better than QQQ, SPY; 3 years SPY + QQQ is better. 88% buy on barchart.com; but 50% sell short term. [Sell on PSAR daily] And its below 50% don channels monthly mid point which is short term bear. Looks related to semi liquid ROBO; + liquid QQQ, sqqq
%% It also tends to be sign of short term trading goals+ usually inferior lower profits, due to bid\ask spread + sometimes comissions, slippage+ lack of market ''prediction'' . Old time trend turtles had such a poor Win rate+ billions of bucks made; i also had to study other stuff. IN my current cash metals business + 44+ years of home + business improvement business i aimed for 100% wins[profit all contracts].NEVER happened all the time, even though i got about a 10% deposit or enough to cover materials/so aim for moon/ knowing may not get there LOL ONE clown, said after i got there with materials ''not going to give you deposit'' I said ''OK, i'm leaving in about 15 minutes if you dont'' He did. Some of best profits are made with drawdowns >32%\33% peak to valleyLOL
In theory, a low win rate is irrelevant, yes, but not in real trading. Human biases want to win frequently. A low win rate and thus risking many losers in a row, make traders abandon a strategy, even though it's a fantastic one in the long run. Very few can tolerate such pain.