Crypto-ML’s trading system leverages multiple machine learning trading strategies to deliver trades to our Trader and Auto Trader customers. Below you’ll find information on how each trade strategy functions.
Crypto-ML trade alerts intelligently select from the best strategy for given market conditions. This means customers do not need to select a strategy. This information is purely to provide transparency and additional data for those interested in the details of Crypto-ML’s approach.
Note: video coming soon.
Matic performs a deep, novel, technical assessment of the market across numerous layers of market data. It utilizes this information to target extremely high-probability trades. These are trades that match patterns of highly successful trades in the past.
Matic performs deep technical assessments of the markets across multiple layers to deliver high-probability trades.
Matic performs complex equations but is not basing decisions on the output from our neural networks. Despite this, there are several areas machine learning contributes to this solution:
- Deep learning uncovers key patterns and relationships in the market.
- Optimization algorithms tune hundreds of parameters.
These findings and parameters are then curated into the Matic solution.
Strengths of Matic
Matic seeks to issue very high probability trades. Its goal is to rapidly compound your trading funds.
Since it doesn’t rely on neural networks for buy and sell decisions, it can execute trades extremely quickly with little lag. This reduces trade slippage.
Weaknesses of Matic
There may be extended periods that Matic stays dormant. If it identifies unfavorable market conditions, it will avoid trading altogether. This, however, is mitigated by Crypto-ML’s ability to intelligently switch between strategies.
Crypto-ML Standard (ML)
ML stands as the core strategy introduced in Release 3 of Crypto-ML. This approach uses the predictions generated by our neural networks to generate trades.
ML trades based on the predictions of highly-complex neural networks.
ML’s trade logic is as follows:
- If predictions are above a certain value, open a trade.
- If predictions are below a certain value, close the trade.
The predictions are generated by machine learning. But an additional layer of machine learning, the optimizers, determines the open and close thresholds. The optimizers also determine stop values and profit targets.
Read more about how Crypto-ML’s neural networks work.
Strengths of ML
The price predictions test on unseen data (data not trained on) to be accurate approximately 92% to 96% of the time, within a tight confidence range (actual values can be found on the Member Dashboard) typically between .31% to .37%. This means if our neural network predicts price will go up 2.50% in the next 6 hours, there is a 96% chance price will go to between 2.13% and 2.87%. That is strong and actionable.
Said simply, ML predicts normal market behavior very effectively.
Weaknesses of ML
While ML predicts normal market conditions effectively, in general, the non-normal conditions are highly interesting for traders. This means the 4-8% of the time that ML is generating predictions outside of the absolute error, critical market behavior may be occurring.
Specifically, these exception conditions tend to be strong rallies and strong drops.
Last, all data must pass through our machine learning pipeline before a trade can be issued. While we continue to shrink this processing time, any delay between input and trade execution may result in slippage.
Crypto-ML Manipulation Detection (Manip)
Manip is designed to trade based on the exceptions mentioned in the above ML section. The reality is machine learning will likely never predict 100% of market movements. There is always going to be a certain amount of chaos. And there will always be a certain amount of manipulation.
If one of our neural networks predictions are right 94% of the time, but we know that other 6% is highly interesting, then the Manip strategy is our solution.
Manip exploits the known limitations of machine learning and statistics to identify unusual price movements as they are unfolding.
The name “Manipulation Detection” comes from an understanding of the way it operates. Here are some examples:
- If the market normally would go up a small amount but it instead goes up a lot, this is a bullish anomaly. Something unusual is causing the market to go up.
- If the market normally would go down but instead goes up, this is also a bullish anomaly. Instead of pulling back like normal, some force is pushing the markets up.
- These situations can also be viewed through a bearish anomaly lens. For example, if the market would normally drift sideways but is pushing strongly down, there must be something causing this unusual movement.
As you can see in these examples, the market is behaving abnormally. It is likely being manipulated by some unusual activity or force. It is this force we want to identify and follow.
Read more about our manipulation detection.
Strengths of Manip
Whereas ML may miss out on strong rallies and poorly handle strong drops (because these are exceptional conditions), Manip will better capture rallies and better avoid drops.
Manip has the ability to generate large gains and constantly trade through major upswings.
Weaknesses of Manip
Under normal, non-extreme market conditions, Manip will tend to deliver choppy trades that result in numerous, small losses.
Manip is generally high-reward and high-risk by nature. As a result, it may incur larger losses than ML.
Last, as with ML, all data must pass through our machine learning pipeline before a trade can be issued. While we continue to shrink this processing time, any delay between input and trade execution may result in slippage.