Last Updated on January 3, 2020 by admin
This week, our crypto trading machine learning models have been upgraded to release 3.5, which adds two key features: trailing stops and synchronization between the various “sell” tools and the Trade Meter.
Giving Machine Learning More Tools
As our machine learning models continue to evolve, our goal is to provide them with more tools to leverage. These tools should allow our models to better handle a wider variety of market conditions.
Our design principle is the models should be allowed to adapt these tools as they see fit. We provide the tool and allow the models to determine how to configure and when to use the tool. That is, these are intelligent tools that can be adjusted and adapted by our machine learning.
In July, we went live with Release 3.0 of our models, which added intelligent stop losses. Since that time, we have seen the models tighten up the stop losses as the market weakened. Seeing these tighten provided us additional insight and confirmation on the market in general. Recently, though, the stop losses have relaxed again, which may be a bullish indication.
Intelligent Trailing Stops
With 3.5, we have provided the models an additional tool: intelligent trailing stops.
“A trailing stop (often referred to as a trailing stop-loss) is a special type of trade order where the stop-loss price is not set at a single, absolute dollar amount, but instead is set at a certain percentage or a certain dollar amount below the market price.” –InvestingAnswers.com
The idea of trailing stops is they better protect your profits. With Release 3.5, the trailing stops will be intelligently set prior to opening a trade. But during the trade, the value will remain fixed.
However, for future releases, our team is analyzing dynamic trailing stops, which would have the ability to change over the life of a trade. For example, they may shrink as the trade ages or adjust as certain profit milestones are hit. More to come here.
When it comes to Auto Trade, similar to the stop losses, these trailing stops will not be entered as orders on the exchange. Rather, they will just be tracked by our models. When a stop loss or trailing stop point are hit, a market sell order will be initiated.
Boost in Trade Results
By applying the 3.5 release to historical data, we find trailing stops provide a significant performance boost in terms of rate of return:
- BTC +15%
- BCH +115%
- LTC +58%
- ETH +27%
Bitcoin Cash is an interesting standout. During the trial period, it saw the largest overall decline in terms of “buy and hold” performance. As such, we may infer that trailing stops particularly help when the general price is not moving in your favor.
Trade Meter Synchronization
In addition to intelligent trailing stops, we have addressed another point of confusion, which has to do with the Trade Meter and when a “SELL” signal is issued.
While the Trade Meter value must cross up above 0 to initiate a “BUY” call, the same is not true for a “SELL” signal. SELL signals are much more complex and dynamic. They can occur in any of the following situations:
- When the Trade Meter crosses zero
- When the machine learning model forecasts weakness (Models 2.0)
- When a stop loss is hit
- When a trailing stop is hit
This means that, in the past, the Trade Meter could stay above zero even if a SELL was generated. Now, after a SELL is generated, the Trade Meter will drop below zero. This will make the visualization simpler to understand and give us a better idea of where the models stand in terms of preparing for the next BUY.
On top of incrementally improving the trailing stops, the next major project our machine learning team is preparing for our next major release.
The 4.0 models will gain the ability to integrate and coordinate with the Market Index. This means the Trade Alerts may choose to behave differently if the market is weakening or strengthening. They may also behave differently in bull and bear markets.
Opening the sort of communication between our systems begins achieving a much bigger vision, where our models not only feed off of traditional data sources but also other machine learning models. Our tools will effectively be able to get in a conference room and talk jointly about how they want to build solutions. Exciting stuff ahead.
Crypto Trade Signals
If you’re not yet getting our trade signals, you can try the Trader membership free for 30 days.
Our check out our other membership plans here including the Free Membership.
And as always, if you have any questions or comments, let us know on our Community Forums.
Crypto-ML provides machine learning for crypto traders and investors. Gain crystal-clear signals and deep market insights. Add predictive capabilities to your toolbox. Learn more and join for free.