How to Take Your (Automated) Trading to the Next Level
This week we’re taking a break from our Deep Dive series to address a question we see quite often from new algorithmic traders:
I’ve tried numerous automated strategies with unimpressive results. How do I take my trading to the next level?
Here at ArcTaurus, we’ve been trading cryptocurrency and building automated strategies for years, so we’ve made many of the same mistakes that you’ve made as well. In this blog post, we’ll touch upon some of the most common mistakes and misconceptions, and give you a few tools to take your trading to the next level.
No, we’re not going to give you some “advice” that you could easily figure out yourself — read on for some actual tips on improving your trading performance.
Note: nothing in this blog post should be construed as financial advice. Cryptocurrency is volatile. Do not trade with more than you’re comfortably willing to lose.
Amateurs vs. Professionals
Many major trading firms have battalions of data scientists, mathematicians, and quantitative analysts working 60+ hours a week or more, but that doesn’t mean you can’t be profitable as a lone trader with a computer and a Binance account.
What do you need is some time, patience, and a bit of knowledge — some of which we’re about to bestow upon you here. First things first, let’s address the elephant in the room: most new traders build strategies revolving around indicators you can find on TradingView. If this sounds like you, our first item of advice is to STOP.
That’s right — moving averages, the RSI, MACD, Bollinger Bands… all of these are lagging indicators, meaning they are indicators of the past, not of the future. These can help you visualize certain patterns when staring at charts, but by the time the next candle closes your opportunity could already be lost.
Technical analysis — that is, the act of looking for patterns on charts — is not really something that the professionals do. When they do look deeply at the charts, it’s often just to see if they can find ways to trap people into thinking a specific pattern is playing out. If you don’t believe me, here’s Alameda Research admitting as much:
Furthermore, and more importantly, if there was a single indicator or indicator-based strategy out there that was consistently profitable, EVERYONE would know about it.
The fact of the matter is, trading strategies based purely off of lagging indicators are almost universally unprofitable in the long run. In order to take your trading to the next level, you need to graduate beyond the “moving average crossover” mentality.
Don’t believe me? Here’s the median result from a series of backtests of 25 major MA crossovers on the 1m, 1h, and 1d time-frames:
You can see that they all underperform and lose money. Stop wasting time trying to figure out some combination of crossover conditions that will somehow be profitable — it won’t. You will need to graduate to the next school of thought, which is that you will need to combine a number of strategies and heuristics in order to trade more profitably.
For those that aren’t familiar, many machine learning systems in use today are utilizing a structure known as an “ensemble,” which uses a series of various algorithms to iteratively constrain the problem into a more manageable or understandable size.
Usually, these ensemble methods begin with a relatively basic method — clustering, for example. Once the results have been categorized into clusters, a different method may be applied, such as regression or boosted trees.
By breaking the problem down into smaller pieces and using the results from each iteration to inform the following process, data scientists are able to better predict, classify, or organize the data.
Similar to how AI ensemble methods improve upon the performance of the underlying models, strategies that build on top of each other often perform better than their standalone parts.
They also lead you to trade less frequently and define more stringent conditions for entering a trade, ensuring that your trades are higher-conviction and risk less.
Start with basic conditions — what is the current prevailing trend? Are we in a bull or bear market? How do you determine this? Even a simple indicator such as a fast MA over a slow MA on a high time frame may be a good enough start (just don’t use it as your only indicator).
Once you’ve defined the current trend, you can start diving deeper into shorter time frames. How to identify reversals, take advantage of volatility, manage position size, continuously take profits — these are all tools that the professionals encapsulate in their automated strategies.
Position Sizing is Key
Ask any professional trader and they’ll tell you that position sizing and risk management are among the most important tools a trader can have. You will dramatically reduce the risk to your overall portfolio while still ensuring you have proper exposure to the markets such that you can profit from strong directional moves.
When building an automated trading system it may be tempting to utilize 100% of your trading balance for each trade, but even with proper risk management this can be risky. A bug in your code or a failed order can mean a blown account.
Instead of using X% of your portfolio, spend some time devising a way to scale into and out of trades in an automated manner. We have previously written at length on this topic, so be sure to check out those other blog posts.
When you automate the process of scaling into and out of trades, you can increase your exposure as more and more positive conditions become true, and slowly scale out as the conditions unwind, allowing you to gradually reduce your risk throughout the entire process.
Visualize Your System
Humans are visually oriented. Walls of text and spreadsheets full of numbers are harder to decipher than a good visualization of the data.
Utilizing a backtesting library (such as backtesting.py) that has a built-in visualization system for your trades (or manually going through the charts on TradingView) is an old-fashioned but highly intuitive way to investigate your strategy. By seeing your trades on a chart you can more easily determine if your logic is flawed, or if your strategy is doing something you didn’t expect.
Let’s take a look at a very basic example:
This strategy enters a short position whenever the price goes above the upper band, a long position whenever the price goes below the lower band, and closes whatever position is open if the price goes 1% lower than the entry price.
This is just an example strategy (remember what we said earlier about strategies based on lagging indicators) but it is easy to visualize how this strategy works. By seeing exactly when your strategy is entering positions, you can not only evaluate the strategy itself but you can also potentially devise new heuristics or conditions that will further augment your strategy.
If you’re poring through spreadsheets or command line output, it can be exhausting to filter through all the data. Spend some time visualizing it!
If you implement all of the above tips and tricks, we feel confident that you will continue to gradually improve your performance without blindly backtesting random indicators hoping that you find the “magic settings” for that one indicator you saw on a recent TradingView blog post.
However, if this seems like a lot of work to you, that’s because it is! Again, there’s a reason why trading firms have hundreds of people working around the clock on these sorts of things.
Do you want to be building and testing powerful trading strategies without having to write thousands of lines of code? Try ArcTaurus instead! Our no-code platform allows you to devise any kind of trading bot you could imagine, with a powerful build-in backtesting tool to ensure that your strategies are truly profitable!