Profitable Crypto trading strategies part 10: Kratos 1.0
Cryptocurrencies are well known for their volatility and uncertainty. But they are so easily manipulated by crypto whales which have the resources to manipulate the markets as they please, a concept called pump & dump. Ordinary traders like you and me don’t have access to these insider trading groups, but we can use our own intellect to develop trading algorithms that detect these pump & dump scenarios.
Machine learning and A.I. is extremely useful to model complex problems, detect patterns and so forth. But they are not able to help us predict pump & dumps (P&D) —because these are rare and orchestrated events, it makes them highly unpredictable.
Also read: The Best Crypto Trading Bots
Instead of trying to predict the next P&D, we can instead try to detect it. If we can detect a pump early on, we can then initiate a buy order. The pump can move the price quite a lot in just a matter of minutes (or hours), usually anywhere between 1% and 5%, sometimes even more.
The detecting of pump & dump events can be categorized as anomaly detection (AD). Anomalies in statistics are referred to as outliers in the data, and there is a whole science around AD systems. If you want to know more about the technical specifications of AD systems, how they work and how you can make your own: Pavel Tiunov’s has a few good articles and tutorials to help you get started.
One can use basic mathematical/statistical formulas to create AD systems for simple problems, but these won’t work for complex problems. For the latter case one can use neural networks, or similar A.I. systems to detect anomalies in highly complex datasets.
Anomaly detection for crypto isn’t the most complex problem, but it’s far from the easiest one. For quite some time we have been trying to come up with trading algorithms that make use of P&D scenarios. The challenge is to detect them in a very early stage, because if it’s too late then you won’t make any profit. Detecting P&D’s is child’s play in terms of formulas, but the challenge is to categorize and react to them accordingly. Sometimes a small pump can evolve into a bullish uptrend market; or result in a long-term bearish downtrend market. Pumps also occur during bearish conditions, but these are very risky and we should try to avoid them. These additional complexities are just a handful of the total, in reality there are way more possible scenarios to take into account.
Kratos 1.0
As I was developing new trading algos, I stumbled upon a method for detecting P&Ds quite well. Out of this code the Kratos trading strategy was born, it remains an early stage strategy but has great potential. Matter of fact it outperforms many of our other strategies in the ETH/USDT market. On the chart below you can see a few of its buy and sell signals.
By looking at the signals on the chart above we learn two things:
- The Buy signals are pretty good, they occur pretty early during a pump. Maybe we can improve them even more in future revisions.
- The sell signals are sub-optimal, and this is an important point. The ROI from this strategy can be greatly improved by improving the Sell signals. In this version we make the algorithm sell when a 3% profit margin is guaranteed. In practice, especially if you’re trading manually, you may want to follow the price until it reaches a peak. Peak detection is yet another complex problem, since we can never know for certain whether the price has reached a local maximum or not. In this case you’ll have to either speculate and take the risk of waiting longer, or play safe and exit after making a certain profit margin (e.g. 3% as in our case).
Backtesting “Kratos 1.0”
In the simulations below we’ll run this algorithm over a 60 day period (1 April to 30 May, 2018) for three different cryptocurrencies from Binance exchange.
Keep in mind that our ROIs incorporates the exchange trading fees and take additional slippage into account. Finally we run the backtesting simulation 100 times whereby the buy and sell prices are randomly selected from the interval’s [Low, High] range. As a result the final average ROI is very realistic and serves as a lower-bound expected return.
BTC/USDT yields an average ROI: 14.33% (±8.56)
LTC/USDT yields an average ROI: 16.04% (±9.35)
ETH/USDT yields an average ROI: 75.80% (±15.59)
Analysis and discussion
These are some great ROIs, and by improving the “sell” positions we could improve these by another 30%. But what’s fascinating is that this strategy works way better for ETH than for BTC and LTC combined.
There are many reasons why this algo resonates so well with the ETH market, but its explanation is very mathematical and out of the scope of this article. But to visually illustrate why it performed relatively poor for BTC/USDT have a look at the next chart:
From the chart above we see that the algorithm generated quite a few unfavorable “buy” signals, most of these were at the end of a pump’s lifetime (thus too late). While some of these buy signals were pretty “okay”, in the end it had to sell at a loss (to reduce even more losses). This is a problem because we use a very basic heuristic: sell when a 3% profit is made or upon a 2% loss. So in BTC and LTC markets a pump rarely goes beyond 3%, as a result we rarely have the opportunity to sell at 3% profit. We can improve these algorithms by optimizing their hyper-parameters, as a result their ROIs could potentially double.
Conclusion
We learned that these buy signals are very good positioned. This means you can make a great profit by following the signals. But to improve our returns we shouldn’t put all our faith into the “sell” signals. We can make even better profits by manually monitoring the price ourselves, and increase our chances of selling at a profit.
If you enjoy our work and free articles, make sure to subscribe and follow. Have a great day and stay tuned for the next.
- Ilya Nevolin
Also, Read
- The Best Crypto Trading Bots
- The Best Bitcoin Hardware wallet
- The Best Crypto Tax Software
- Best Crypto Trading Platforms
- Best Wallet for Uniswap
- Best Crypto Lending Platforms
- Bitsgap review — A Crypto Trading Bot That Makes Easy Money
- Quadency Review- A Crypto Trading Bot Made For Professionals
- 3commas Review | An Excellent Crypto Trading Bot
- 3Commas vs Cryptohopper
- The Idiots Guide to Margin Trading on Bitmex
- The Definitive Guide to Crypto Swing Trading
- Bitmex Advanced Margin Trading Guide
- Best Crypto APIs for Developers
- Crypto arbitrage guide: How to make money as a beginner
- Top Bitcoin Node Providers
- Best Crypto Charting Tool
