Using machine learning for predicting movements in the stock market has been a hot topic for years. Novices and professionals alike have attempted to create reliable models that can turn their crypto portfolio into profit machines.
Whether or not there are teams that have successfully implemented such ML strategies over the long term is heavily debated. Certainly, there are aspects of portfolio strategies that can be automated with ML, but a completely automated strategy may be more difficult to conceive.
One study from Eurekahedge suggests AI and ML funds outperformed quants and other traditional hedge funds.
Machine Learning for Trading - Topic Overview - Sigmoidal
Artificial Intelligence (AI) and Machine Learning (ML) are quietly revolutionizing nearly all areas of our lives. Did…
Although AI and ML strategies have become popular in the traditional financial markets, I still don’t see a significant volume of discussion on the topic in the crypto space.
Maybe the volatility and unpredictable market movements have scared people off. Whatever the reason, it’s time to spur some discussion.
On the price prediction front, Nomics took the reigns when they launched their 7-day price predictions. These price predictions use a machine learning model known as “long short-term memory” (LSTM). By training the models with daily OHLCV candlestick data, the model predicts the price 7 days later.
I was pumped to see this topic being explored and was ready to give the predictions a try. However, I didn’t know exactly how to test them out. With only a few data points, it would be difficult to backtest with any real confidence. That meant the only way to test would be to do it live.
So, I got started with writing the rules for a case study.
Since I thought just picking one or two assets to buy and hold each week would be risky and not give the best picture across all the predictions, I decided to use 10 assets each week.
That meant at the start of each week, I would take the top ten assets that were predicted to increase the most by Nomics and put them into a portfolio.
Note: Only assets that were available on Binance were considered. I also had a restriction that there needed to be enough liquidity and minimal spread on the order books.
Using machine learning for your portfolio is easy. If you wanted to run the same study, you could do it with these 3 easy steps.
1. Go to Nomics
Nomics: Crypto Market Caps - Prices, All-Time Highs, Charts
Real-time crypto market cap rankings, historical prices, charts, all-time highs, supply data & more for top…
2. Sort by 7D Prediction
This will arrange the assets in order from the highest predicted change for the next 7 days to the lowest.
The prediction for each individual asset can be found by hovering over the information buttons.
3. Add the top ten assets to a portfolio
With this information, you can go to the exchange and buy each of the assets that were predicted to perform the best.
The whole process takes only a few minutes, and since the predictions are over a 7 day period, the portfolio only needs to be updated on a weekly basis.
If you want to automatically copy this strategy without doing any of the work yourself, you can do that by following this leader on Shrimpy.
Shrimpy | MLCaseStudy — Machine Learning Case Study
Join the study on Shrimpy to automate your portfolio strategy!
Knowing these details is rather useless if the strategy doesn’t work. So, how has this portfolio strategy been working out so far?
Like a legend.
That’s right. At the time of writing, the portfolio that uses the ML predictions each week has seen a 104% increase in value over 9 weeks. This is all while Bitcoin has only seen a +7% change in value over the same period.
It has only been 9 weeks so far, but the study has shown that machine learning is worth investigating.
This doesn’t say anything about future performance, but I would be missing out if I didn’t at least consider this data as fascinating.
I hope you will also find this fascinating!
There are countless ways this data could be used to select assets, make trading decisions, or even just factor into a larger strategy.
Machine learning is just getting started in the cryptocurrency market. The data I’ve seen so far has blown me away. Hopefully, this is just the beginning of a growing field of research.
I publish weekly updates about the study, so I would encourage you to follow along. The original case study article can be found here:
Case Study: Using Machine Learning for Portfolio Management
In a recent blog post, Nomics announced the release of its 7-day crypto price predictions . Their predictions use a…
The update from last week can be found here:
Machine Learning for Crypto Portfolio Management Case Study: Week 10
The Nomics ML strategy has maintained a steady lead over the other strategies since the second week of this study. We…
Our Social Trading Platform
Shrimpy is a social trading platform for cryptocurrency. It is designed for both professional and novice traders to learn about the growing crypto industry. On Shrimpy, users can copy the portfolios and trading strategies of other traders.