Greetings from
NeuroDimension!
The World Leader in Neural
Network Software
This issue of the
TradingSolutions newsletter shows you how to combine the strengths of multiple
profitable signals using committees and discusses the power of the optimal
signal.
In this issue you’ll find:
Feature Spotlight
* The Optimal Signal
- What is the Optimal Signal?
- Why Should I Use It For Modeling?
- Is Using Future Knowledge Legal?
* Utilizing Multiple Models for One Security
Products and Events
of Interest
* Neural 101
Note: You are receiving this newsletter because you requested to stay informed concerning information about TradingSolutions from NeuroDimension. If you would like to stop receiving these newsletters, please see the bottom of this newsletter for removal instructions.
=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=
This edition of the feature spotlight looks at one of the most powerful features in TradingSolutions, the Optimal Signal. Since the optimal signal is featured prominently in the tutorials, many of you may be familiar with its advantages. Here is an overview of this feature along with some answers to frequently asked questions.
The optimal signal is an entry/exit signal that is generated by looking at future prices to determine the most profitable current position. In addition to other information, it takes two parameters – the number of bars to look into the future, and the minimum profit required for a trade to be considered worthwhile. At a basic level, this means that if the minimum profit can be achieved within the specified number of days, an entry signal will be generated.
How is this useful? Visually, it can be used to highlight places in a chart where profitable trades could have been made. More importantly, it can also be fed into a neural network as an example of what a profitable signal should look like. In other words, it can be used as the “desired output” when training a neural network.
Since modeling the optimal signal differs from the way many neural network practitioners approach trading, it can take some getting used to. The conventional approach is to predict if the price will increase or decrease and trade based on that information. Unfortunately, this often doesn’t concentrate on whether the resulting trades would be profitable. However, by having the neural network model the optimal signal, you are training it to specifically identify profitable situations. And, by setting a minimum profit requirement, you are filtering out many of the small price fluctuations that can be attributed simply to “noise”.
Some newer TradingSolutions users have been concerned about using “future knowledge” with a neural network. As a general rule, you should never use future data as an input to a neural network. This is because when it comes time to use the neural network for trading, you wouldn’t have the future information to feed it in order to get a useful output.
However, in this case, we are using the future knowledge as a desired output, which is only used when the neural network is being trained. Because of this, the current value of the desired output is not needed. Technically, this is no different than predicting the change in prices five days in advance. The optimal signal just adds additional preprocessing to make the desired information more specific to how it will be used.
What would not be correct is to use an optimal signal as an input to a neural network. Since the optimal signal requires future prices to be generated, its value is not available for the current time. This means that a neural network wouldn’t have that input available and, therefore, not be able to produce a current value.
For more information on the optimal signal, see “Generating an Optimal Signal” in the “Creating and Analyzing Entry/Exit Signals” chapter of the online help.
=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=
As you use TradingSolutions to mine for profitable models, there will be times when you come across two models for the same security that both perform well. These models may have different strengths and weaknesses. For example, one may have a more smoothly ascending equity curve while the other may have a higher total profit.
The first thing to try when you have two good models is to form a third model by creating a committee of the two. This is simply a collection of the two models in which each model “votes” for each day’s market action (buy/sell/hold). The simplest case is to assign each model a single vote; however, you may want to experiment with other voting schemes.
In many cases, the committee that you create will combine the strengths of the two models to either increase your profit, decrease your risk, or both. If this is the case, then you will likely want to use this committee as your trading model. Otherwise, you may want to monitor the performance of both models over time. The easiest way to do this is to create a Signal Analysis field for both models that calculates the return since a specified date. These two calculations can then be added to the Display Fields. This screen shot is an actual trading portfolio that demonstrates this concept:
http://www.tradingsolutions.com/newsletters/images/altreturn.gif
The “6M Return” is the performance of the primary trading model (the one that is currently being traded) for the given stock since 7/1/00 (roughly 6 months). The “Alt Return” is the performance of the alternate trading model for the same time period. There will be times when the alternate model will begin outperforming the primary model. If this happens, you will want to analyze the performance of both models in more detail. If the alternate model looks better, then you may want to make it your primary model and start trading from its entry/exit signal instead.
For more information on committees of signals, see “Combining Multiple Signals in a Committee” in the “Creating and Analyzing Entry/Exit Signals” chapter of the online help.
=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=
Neural 101 is a free educational web site on trading stock and futures markets with neural networks. It includes message areas, links, trading indicators, systems and scans. The message areas include special forums on building trading systems, using neural networks, and even discussing tips, tricks, and ideas for using TradingSolutions.
NeuroDimension encourages the use of this free resource to foster discussions between TradingSolutions users. You can visit Neural 101 at http://www.neural101.com. You can go directly to the TradingSolutions message forum by selecting “Message Forums” from the top menu, then selecting “TradingSolutions” from the list of Specific Software Conferences.
Have an event or product you would like to announce? Contact us at submissions@nd.com.
=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=
We appreciate your feedback! Please send us your comments or
suggestions concerning this newsletter, our web site, or part of the
NeuroDimension product line. Write to us at: feedback@nd.com
Have questions about TradingSolutions? Send your questions to: info@tradingsolutions.com
Some mail readers may display an attachment at the bottom of this newsletter. This is typically caused when the mail reader is set to display text only. The original, formatted version is sometimes included as an attachment.
If you would prefer not to receive these newsletters or subsequent product updates from NeuroDimension, please reply to this letter with the subject heading changed to the word REMOVE.
Thank you again for your support of NeuroDimension products!