In this issue:
I. Commentary by Marge Sherald, CEO
II. Commentary by Denham Ward, Lead Developer of NeuroShell Trader
III. Save 20% on purchase of ChaosHunter and Adaptive TurboProp2 and Adaptive Net Indicators add-ons
IV. Noxa Analytics announces expert advisor for NeuroShell Trader
V. Ward Systems Group closes for Labor Day and returns to fall schedule
VI. Tech support questions
VII. One way to stop this newsletter
***********************************************************
I. Commentary by Marge Sherald, CEO
A. ChaosHunter review in STOCKS & COMMODITIES Magazine
The September 2012 issue of Technical Analysis of STOCKS & COMMODITIES Magazine includes a very thorough overview of ChaosHunter by writer Dennis Peterson. If you subscribe, take a look at page 52 to reader Peterson’s concise explanation of how to build trading models in ChaosHunter.
If you don’t subscribe, we recommend that you do because our experience in reading the magazine has shown it to be an excellent source of ideas for technical trading systems. Users of NeuroShell Trader can create the systems described in the articles and then use the optimizer to refine the system for their own portfolio. ChaosHunter users can look at the inputs for the trading systems and feed them into ChaosHunter to see what type of model develops.
B. When to retrain a neural net
We are often asked when you should retrain a neural net prediction. Our usual answer is that there is no right answer. We know customers who retrain every day, every week, or no earlier than six months and they are all successful. But there are some Ward Systems Group add-ons to NeuroShell Trader that update neural nets automatically.
1. Adaptive TurboProp2
The neural network in the Adaptive TurboProp2 add-on is the same neural network algorithm used in the Prediction wizard, with the added advantage of a parameter you can set to tell the net when to retrain itself to incorporate the latest changes in the market. The network has another parameter called trainbars where you specify the number of bars to be used in the training set, so every time the net retrains it drops off the oldest data and trains on the newest data. You can use the nets inside of a Trading Strategy by creating a Long Entry trading rule such as Adaptive TurboProp 2 prediction > 0 (or other threshold value that either you choose or let the optimizer determine). You add similar rules for a Long Exit and the Short side.
2. Adaptive Net Indicators
These nets do a one bar walkforward and evolve with the pattern in each new bar, again incorporating recent market changes in the model. There is one set of indicators for predictions and another for classifications which produce probabilities that the current pattern matches a previous pattern in the training set. Unlike the neural net in the Prediction wizard, you can tell the nets which inputs are more important than others by setting a contribution factor, or you can let the optimizer figure that out as well. You can vary the settings from a very strict match to recent patterns or an imprecise match to allow for some latitude in the decision making process. The nets in Adaptive Net Indicators are the PNN and GRNN networks from our general purpose software, NeuroShell 2 and NeuroShell Predictor and Classifier, but they have been modified for trading. Note: because these nets are designed for pattern matching, they will not predict exact prices very well if the price rises from $25 to $50, for example. These nets can never predict a price higher than the highest price in the training data. They work better if you predict a percent change in the next open.
***********************************************************
II. Commentary by Denham Ward, Lead Developer of NeuroShell Trader
Often when exploring new trading systems, being too loyal to your original ideas and trading plan can cause missed trading system opportunities. In this article I’m going to explore the relationship between contrarian trading, trading costs and time frame by taking a very simple trading model and showing a methodology one might use to turn a losing idea into a winning idea. Note that you can download the charts described in this article from www.ward.net in the New and Updated Examples section.
Let’s assume I’m of the mindset that I want to create a trading system with lots of trades during the day so I don’t get bored and I also want to utilize a well-established trading methodology. To meet the first criteria of lots of trades, I load up a 1-minute bar chart of the E-Mini S&P 500. To meet the second criteria, I create a breakout trend following system that goes long on a 5 Bar High Channel Breakout, goes short on a 5 Bar Low Channel Breakout and only trades during the primary market hours of 9:30 am to 4:00 pm. However, when backtested over the last 6 months the system loses $4,300 dollars trading a single contract.
Since my original trading system loses money a logical next step is to try the exact opposite. Although generally referred to as contrarian trading, my father liked to call this Bizarro Trading after the Seinfeld TV show episode in which George does everything in life by doing the exact opposite of his initial gut instinct. So instead of following the trend on breakouts, I’ll try going against my gut instinct and create a system that trades against the trend. When I reverse the system so it goes long on a 5 Bar Low Channel Breakout and short on a 5 Bar High Channel Breakout, the system now makes $4,300 dollars trading a single contract over the last 6 months.
Although it looks like I have discovered that trading against the trend works on my 1-minute chart, I want to first make sure it checks with reality by adding in my brokerage’s trading costs and adding in a little slippage to account for real world fill prices. However, once I add in a $1 per contract commission and a 1/2 tick (.125) slippage, I quickly find that my bizarro trading system loses $40,000 dollars and my original trading system loses $49,000 trading 1 contract over 6 months.
I now realize that whether I trade my original trading rules or the exact opposite rules, my trading costs are going to overwhelm any desire I have to trade frequently throughout the day. So faced with reality, I slowly increase the chart frequency from 1-minute bars to larger bar sizes. Although I no longer have as many trades during the day, at least now my trading costs won’t be the most dominant factor in the trading losses. As it turns out, the original trend following breakout strategy of 5 bar high and low channel breakouts makes a $6,600 profit when trading 1 contract on 20 minute bars and taking into consideration trading costs.
Although the methodology above was very simplistic in nature, performed on in-sample data and didn’t utilize optimization, it does go to show that by not being so loyal to your original trading plan and ideas, what looks like a losing idea can be turned into a winning strategy.
***********************************************************
III. Save 20% on purchase of ChaosHunter and Adaptive TurboProp2 and Adaptive Net Indicators add-ons
If you haven’t checked out the power of ChaosHunter, we’re giving you the opportunity to do so and save $299 in the process. Buy ChaosHunter for $1196 instead of the regular price of $1495 by visiting www.chaoshunter.com. The price on the order form will reflect the discount.
We’re also offering a 20% discount on the following WSG add-ons:
-Adaptive Net Indicators, usually $299 – now $239.20
-Adaptive TurboProp2, usually $299 – now $239.20
These offers expire on Tuesday, September 4, 2012. Use the following order link to receive the discount: www.ward.net/60.asp. The prices on the order form will reflect the discount.
***********************************************************
IV. Noxa Analytics announces expert advisor for NeuroShell Trader
We are stoked to announce the private release of Noxa AME v1.0, our Adaptive Mixture of Experts, effective August 20th, 2012. It is risk adjusted, self-learned, adaptive and it has only one parameter to set.
You can probably tell by now; markets are broken and fragmented, leaving us in a quandary. High frequency trading is responsible for two-thirds of the trading volume. You may have noticed odd price movements on your charts, ones that are too fast and too consistent to be human. Computers caused them. We have to acknowledge this new reality that computers are increasingly driving the market.
We wanted an indicator able to adapt to not just changing market conditions, but changes in the fundamental way the market works. You can read more on how AME does it from this link: http://goo.gl/Mgi5Z
Also you’ll read about a new index-tracking strategy that trades volatility ETFs quite successfully. Essentially, AME has cracked the volatility VIX index!
Remember, this release is not advertised on our website yet because it is only for existing NeuroShell users and people who read this newsletter. This is the first indicator we’ve made publicly available since the release of CSSA five years ago. We are very happy to be able to put something out there again for you.
AME is $890. It is offered at $666 or a discount of 25% until the end of September. That’s a $224 value. In addition you can get $100 refund on your purchase of NEI and CSSA.
These products must be ordered from Noxa Analytics directly. Call 1 613 599-7844 or send an email to Philippe Lonjoux at p.lonjoux@noxapredict.com.
Happy trading.
Philippe Lonjoux
Noxa Analytics, Inc
75 Spruce Meadows
Kanata, ON K2M2K4, Canada
www.noxapredict.com
1 613 599-7844
***********************************************************
V. Ward Systems Group closes for Labor Day and returns to fall schedule
Ward Systems Group offices will be closed on Monday, September 3, for Labor Day. The staff will return on Tuesday, September 4. Also effective September 4, office hours will be as follows: 9 am to 5 pm Monday through Thursday, and Friday from 9 am to 4 pm eastern time.
***********************************************************
VI. Tech support questions
We are proud that we have provided tech support free of charge to registered users for more than 20 years. However, it is very expensive for us. To continue free tech support in the future, we are trying to be as efficient as possible in order to minimize cost. It will help to achieve this goal if you adhere to the following guidelines.
When you send in a technical support requests via E-mail, it will greatly expedite the process of reviewing and responding to your request if you initially include as many details as possible such as:
— A detailed description of the issue(s) at hand, including steps to reproduce.
— Screenshots of what you are seeing.
— For NeuroShell Trader users, the chart (.cht) file(s) in question, preferably with data saved in them, ChaosHunter (.md) files if used in your chart, along with any external DLL files for indicators you have created. If needed, external data files in .csv, .txt or .prn format are also needed.
— For ChaosHunter users, the .csv file, along with the .md and .work files.
— Your product serial number and numeric version of the software used.
— Specifics on your computer setup such as processor, RAM, any virtualization used, and type of internet connection.
— Details of datafeed, brokerages and third party add-ons used.
We do understand that there are times where you may not want to share your charts and other files with us due to secrecy or for items of a proprietary nature. Rest assured, your files are NEVER provided to ANYONE outside of Ward Systems Group. We only open your files for the sole purpose of responding to your request. We have NEVER nor will we EVER use your files for any other reason.
There is only one exception to this policy, and that is problems that involve third party add-on vendors where we may need their assistance. We only take this step after we have received your permission to send your files to the vendor, and then the files are only accessed internally by those who have a need to use them to respond to your request.
Understandably, you may still decide to not provide your files. While we will do our best to investigate based upon information you provide to us, our response time will be slower and the priority lower, than for requests where files have been provided.
***********************************************************
VII. One way to stop this newsletter
It is really easy. Just change your email address and don’t tell us.
***********************************************************