# January 2013 Newsletter – Consensus models / Mechanics of Multiple Data Frequencies in Power User Versions

In this issue:

I. Commentary by Marge Sherald, CEO

II. Denham Ward, Lead Developer of NeuroShell Trader

III. New ChaosHunter Model Available for Download

IV. ChaosHunter Trader One Year Anniversary Sale

VI. Office Closed January 21, 2013

VII. How to Stop This Newsletter

***********************************************************

I. Commentary by Marge Sherald, CEO

Building successful models is a challenging process, even when you have some of best modeling software such as NeuroShell Trader and ChaosHunter. One technique that produces results in both NeuroShell Trader and ChaosHunter is to build ensemble models. We have posted the following before, but I believe it is worth repeating.

It is easy to prove that consensus models (independent ones) work better than any of their parts if you think of your models at any time as being either right or wrong, as say in a reversal strategy where you are either long or short. Let’s say you have three models (A, B, and C) and each one has a probability of 0.6 of being correct at any given time. In other words, the models are 60% right. To figure the probability that the average (which is the consensus) is right, think of it this way. The average is right if 2 or 3 of the 3 are right. Now,

probability of A right, B right, C wrong = 0.6 x 0.6 x 0.4 = 0.144 probability of A right, B wrong, C right = 0.6 x 0.4 x 0.6 = 0.144 probability of A wrong, B right, C right = 0.4 x 0.6 x 0.6 = 0.144 probability of A right, B right, C right = 0.6 x 0.6 x 0.6 = 0.216

The sum of the probabilities is 0.648 so the average will be right about 64.8% of the time.

If you assume the models are 55% right, the average of the 3 is 57.475% right.

To build an ensemble model in NeuroShell Trader, you can create several neural networks based on different indicator categories such as price momentum, volume, and regression slopes. The next step is to incorporate the nets in a trading strategy that says to buy if two of the three are true. You can also combine nets and trading rules. For some detailed instructions on how to create an ensemble system, visit www.ward.net and look in the Tips and Techniques section for an article called “More Tips from the President”.

If you’re building models in ChaosHunter, you can simply save three of the models with different names and use the ChaosHunter Trader to send trades to Interactive Brokers only if a majority of your models generate trading signals.

It’s simple to save ChaosHunter models with different names. After optimization is complete, go to the ChaosHunter apply screen and you’ll be able to view a list of the models that were created. Select a model that is profitable in the out-of-sample set, and click on the Apply Model button. Next, go the File menu and choose “Save model”. Repeat for other models, but be sure to save them with different names. If you have the ChaosHunter Trader, you can connect to Interactive Brokers and then load up a chart for the instrument you modeled in ChaosHunter Trader. When you specify the trading setup, add all three models and choose a majority rule of two, for example. If you don’t know about ChaosHunter Trader, check out the article below.

**********************************************************************

II. by Denham Ward, Lead Developer of NeuroShell Trader

The Mechanics of Multiple Data Frequencies in Power User Versions

The Power User versions of NeuroShell Trader include the powerful ability to combine multiple frequency data on a single chart. Not only does this allow displaying of different data frequencies like 1 day, 2-hour, 3 minute, 2.5 range, 5,000 volume, 7500 tick and 30 second bars all on the same chart, but it also allows indicators, predictions and trading strategies to incorporate different frequencies in a single calculation, forecast or trading decision. To understand the multiple frequency capabilities of the Power User versions, it’s important to realize that there are three different data frequencies that can be varied: the chart frequency; the output frequency of an indicator, prediction or trading strategy; and the frequency of inputs to an indicator, prediction or trading strategy.

The chart frequency is simply the data frequency that was chosen when the chart was created and can be changed using the dropdown box at the top of every chart. When the chart frequency is changed using the dropdown box at the top of the chart, any of the data, indicators, predictions or trading strategies that share the same data frequency as the chart will also change to this new frequency. However, data or calculations that do not share the same frequency as the chart (i.e. chart independent frequencies) will remain at their original frequency.

When inserting data, other instrument data or creating a new indicator, prediction or trading strategy, you will be given the option of choosing the output frequency of this new addition. If you choose same frequency as the chart, then the new addition will have the following characteristics:
1) Same data frequency as the chart data
2) Data points line up with the original chart data
3) Changes frequencies in conjunction with the chart when the chart frequency is changed using the dropdown box at the top of the chart.

However, if you choose a chart independent output frequency (for instance 2.5 range bars on a 5-minute chart), then the new addition will have the following characteristics:
1) Different data frequency than the chart data
2) Displays on the chart with more or less data points than the original chart data depending upon whether the frequency is higher or lower than the chart frequency.
3) Data points generally do not line up with the original chart data, especially when combining non-time based frequencies like range, volume or tick bars. For instance, when adding 5 minute data to a 10 minute chart, every other 5-minute bar will match a 10 minute bar, but when adding 2.5 range bars to a 10 minute chart, the bars may never line up at the same timestamp.
4) Do not change frequencies with the chart when the chart frequency is changed. To change the frequency, you must modify the original indicator, prediction or trading strategy.

The last frequency that can be varied is the frequency of the inputs to an indicator, prediction or trading strategy. The input frequency is simply the output frequency of each individual input and is set when you add each individual input. When the inputs are all the same frequency as the indicator, prediction or trading strategy output, then the data timestamps all line up and calculation proceeds as expected. However, when the input frequencies are different than the output frequencies and don’t line up in timestamps, the calculation uses the last available data value for each particular input frequency. So for instance a 10-minute Add2(1-minute Volume, 1-Hour Volume) will produce a value every 10 minutes using the current 1-minute volume value and the last available 1-hour volume. Although there will always be a current 1-minute volume that matches each 10 minute time stamp, the 1-hour volume value used will not be current except at the 10-minute bar that occurs at
the
top of each hour. So even though you can easily combine multiple input frequencies, you must keep in mind that lower frequency data and non-time matching frequencies like range, volume and tick bars may be using the last available value and not always a value calculated at the exact same time as the output frequency time stamp.

By combining different chart, output and input data frequencies, the NeuroShell Trader Power User versions have the ability to create an almost unlimited combination of data frequency displays and calculations. Just imagine a 10-minute chart displaying 2.5 range bars, 30-second other instrument data and 1-week data with a 4 minute prediction of the 1-minute change in price using 3-minute stochastic, 1-day relative strength and 10,000 tick bid & ask trade volume ratio as prediction inputs, which is then fed into a 25,000 volume trading strategy which confirms the 4-minute prediction with the latest 45 second MACD before issuing a buy signal.

***********************************************************************

IV. New ChaosHunter Model Available for Download

We’re continuing the process of updating the models on our web pages. This month we’ve created a ChaosHunter Forex example that uses the time flag tip from the November 2012 newsletter to restrict trading hours. We first created a 15 minute chart for the Swiss Franc/US dollar pair in the NeuroShell Trader and added the Stochastic%K indicator plus the time flag indicator (X <= Time <=Y(Date, 5:30:00 AM, 11:30:00 AM). We also included data for gold, the Euro, the British Pound, and Hong Kong dollar. Next we exported the data from Trader as a text file.

On the ChaosHunter inputs tab we selected the Stoch%K, Euro high and low, and the British Pound high and low as inputs and the Swiss Franc open as the output. On the optimization tab we chose Evolution Strategy, a population size of 300 and a buy/sell threshold range of +/- 10. We traded an account size of \$100,000; costs were assumed to be \$10 each way, and we checked the box for smooth equity curve. We selected long and short trading positions and true reversal. We checked the Time Flag series box, and selected the time flag indicator column from the NeuroShell Trader data file. Since the last time in the indicator is 11:30 AM, ChaosHunter exits any position after 11:30 AM. (Time flags are useful if you want your ChaosHunter model to trade more often because of this forced exit.) On the formula tab we chose all of the arithmetic operations, -x and 1/x from the algebra group, sin and cos from the trigonometry and transcendental category, plus neuron2 and neuron3 from t he neural category.

The out-of-sample results showed 62.5% profitable trades based on a ChaosHunter formula that used a 2 neuron neural net, the low of the Euro and British Pound plus some constants and the Stoch%K.

The screen shot you see on the web site is from the “out-of-sample” time period, which is the period in time just after the optimization took place. In other words, the system did not “see” this data when the model was being built; the backtest simulates trading that would have occurred had you built the model earlier and then traded it during the dates shown on the chart. (Note that many intraday models work better if they are re-evolved as often as weekly with more recent data.)

You can download the data, template, and model files from www.ward.net if you own ChaosHunter.

Disclaimers: This model might not provide good results very much longer than shown, so there is no guarantee that a model that performed well “out-of-sample” will not lose money later. All trading involves risk, including loss of principal. Furthermore, the same inputs will not necessarily work as well with other ticker symbols and in different time periods. Not all of the models we build work this well, and therefore your results may vary as well. This model is shown simply as an example of the types of models you can build with ChaosHunter.

***********************************************************************

IV. ChaosHunter Trader One Year Anniversary Sale

Our ChaosHunter software (www.chaoshunter.com) was introduced five years ago after we discovered an efficient way to use evolution to construct formulas for many purposes. On the one hand you can evolve formulas for making business and science models, such as for predicting sales, events in nature, or future prices. On the other hand, it can construct formulas for producing buy/sell signals in the financial markets. It seems to meet the needs of those traders who don’t have the time to experiment with constructing rules for making buy and sell decisions. The fact that ChaosHunter makes simple trading formulas that can be transported to other platforms means that Ward Systems could now capture the business of those traders who weren’t willing to switch to NeuroShell Trader Professional from some platform to which they were already inextricably attached.

There are conversion libraries to assist in transporting formulas to many of the popular trading platforms, and it is a simple indicator call to load ChaosHunter models into NeuroShell Trader. Then we realized that many of our ChaosHunter users really would have no need to move ChaosHunter formulas to some full featured platform if they could just trade the ChaosHunter formulas fast and directly. ChaosHunter Trader (www.chaoshunter/chaoshuntertrader.asp), which is separate and distinct from ChaosHunter itself, was introduced in January 2012.

ChaosHunter Trader trades ChaosHunter models with highly popular Interactive Brokers (IB) accounts. It even exports data from IB for ChaosHunter consumption. You don’t need anything else to make a whole portfolio of models.

To get your trading year off to the right start, we’ll honor the ChaosHunter Trader introductory price of \$625 compared to the \$895 list price for one week only, until January 25, 2013.

To buy ChaosHunter Trader at this special price, order only from the following web link or give us a call at 301 662 7950 or Skype us at Ward Systems.

www.ward.net/60.asp

***********************************************************************

V. WSG Add-on Sale

-Fuzzy Pattern Recognizer
-Advanced Indicator Set 1
-Neural Indicators
-Cluster Indicators
-Advanced Indicator Set 2
-Turning Points
-Fuzzy Sets
-Pattern Matcher
-Advanced Indicator Set 3

To buy the WSG add-ons at this special price, order only from the following web link or give us a call at 301 662 7950 or Skype us at Ward Systems.

www.ward.net/60.asp

***********************************************************************

VI. Office closed January 21, 2013

The Ward Systems Group Inc. office will be closed on Monday, January 21, 2013, in honor of the birthday of Dr. Martin Luther King, Jr.

***********************************************************************

VII. One way to stop this newsletter

It is really easy. Just change your E-mail address and don’t tell us.

***********************************************************************