June 2014 Article

June 16, 2014
 

Position Sizing Improves Profits

by Marge Sherald, CEO

Using NeuroShell DayTrader Power User, I built a prediction for EUR/USD June 14 futures contract using 10 minute bars.  The model optimized on only a few days worth of data but held up on several days of out-of-sample data.  The signals looked pretty good, hitting peaks and valleys in both the in and out-of-sample data.
 
Chart displays the trading signals from a neural network prediction for EUR/USD
 
Trading rules from the prediction may be used in a trading strategy
Next I wanted to see if the Position Sizing feature in the Power User would improve the profits.  I inserted the trading rules from the prediction into the Trading Strategy wizard by selecting them from the Prediction outputs folder.
Long Entry:  Prediction > Long Entry Threshold 4 bar %Change in Open
Long Exit:  Prediction < Long Exit Threshold 4 bar %Change in Open
Short Entry:  Prediction < Short Entry Threshold 4 bar %Change in Open
Short Exit:  Prediction > Short Exit Threshold 4 bar %Change in Open
By inserting the prediction rules from the list, the predictions will not be optimized even if the Trading Strategy is optimized. In order to match the results from the prediction, you need to insert all four of the prediction rules.
On the dates tab in the Trading Strategy wizard, check the box to match date ranges for out-of-sample data from the prediction to the dates used in the Trading Strategy. Also be sure to match the time frame from the prediction, which in this example only allows trades between the hours of 5:30 am to 11:30 am.
Position Sizing
To learn if position sizing increased profits, I chose three different methods on the position sizing tab in the Trading Strategy wizard:  fixed units, Fixed Dollar Amount per Unit, and Fixed Dollar Risk per Trade. The optimizer chose a position sizing method of Fixed Dollar Risk per Trade of $10,000.
Results

The Trading Strategy using a Fixed Dollar Risk per Trade of $10,000 showed returns as follows:

Trading Strategy in-sample:  $58,274.30
Trading Strategy out-of-sample:  $10,708.45
This trading system is meant to be used as a guideline for improving profits from predictions by using position sizing in a Trading Strategy.  Of course you have to work within your account balance and the amount you’re willing to risk per trade to match your trading style.

New Version of NeuroShell Trader with FXCM Available

If you trade FOREX and have an FXCM account, now is your chance to use NeuroShell Trader with FXCM data and to send trades to your FXCM demo account.  A new beta version of NeuroShell Trader with the FXCM interface is available from WARD.NET in the Release news, upgrade information and beta tests section.
This new release makes it easier to download daily data plus intraday data.  You can still build charts with FXCM data without installing any other software.  All you have to do is select FXCM from the Tools menu, data sources, server tab in NeuroShell Trader.  To send trades to your FXCM demo account, select FXCM as the brokerage from the Tools menu, options, trading orders tab.
NeuroShell Trader Now Works with TradeStation Data and Brokerage

If you have a TradeStation account, now is your chance to use NeuroShell Trader with TradeStation data and to send trades to your TradeStation simulated account.  A beta version of the TradeStation interface to NeuroShell Trader is available from WARD.NET in the Release News, Upgrade Information and Beta Tests section. A single install program includes both the TradeStation Data Server and Broker programs.  Note that TradeStation does not have to be running in order to obtain data or send trades to your TradeStation account.
Setup in NeuroShell Trader
Once you’ve installed the TradeStation Data Server and Broker programs, open NeuroShell Trader and select the Tools menu, data sources, server tab and then look for the TradeStation Data Server on the list of available options.  When you have charts running in NeuroShell Trader, the TradeStation Data Server will display a list of symbols that are currently loaded in charts that are running in NeuroShell Trader, along with the bar size and update information.
To send trades from NeuroShell Trader to your TradeStation simulated account, open the NeuroShell Trader Tools menu, options, and trading orders tab and choose the TradeStation Broker.  To automatically send trades from a Trading Strategy in a NeuroShell Trader chart, activate trading in the Alerts and Orders window available from the View Menu in NeuroShell Trader.  The TradeStation Broker program will run in the background and display information about your accounts and current balances, positions, and order history.  You may also manually send trades to TradeStation from the TradeStation Broker program.
Because this is a beta release, users are cautioned to only use their TradeStation simulated accounts at this time.  
Beta testers must currently own a 6.x license of NeuroShell Trader and be running NeuroShell Trader 6.4 beta or higher in order to take part in this beta test.  Comments about the beta test should be sent to 

su*****@wa*********.com











.

The TradeStation Data Server and Broker programs are a no cost update to NeuroShell Trader.

NeuroShell Trader Selects ChaosHunter Model

by Marge Sherald, CEO

An ideal trading model would not only generate trading signals before a large change in price, but also be able to generate profits during sideways markets.  However, it’s a rare model that can excel in both types of markets.  That’s where an ensemble system can bridge the gap.

Since ChaosHunter generates a variety of models during optimization, you can select several that work under different conditions and build an ensemble system in NeuroShell Trader.

First create a model in ChaosHunter.  Once optimization is stopped, go to the apply screen and click on different models and apply them to the training data.  If you see a model you want to keep, give it a unique name.  To create this example, I chose three different ChaosHunter models and named them Signal A, B, and C.

I displayed these models on the Trader chart by using  an indicator in the External Program and Library Calls category called ChaosHunter Trading Signal.  The chart below shows how the trading signals are different for each model.

Three different ChaosHunter models were chosen for inclusion in the Trader model based on their ability to find different trading opportunities
 

How to Implement a ChaosHunter Panel of Experts in NeuroShell Trader

I chose to use the ChaosHunter Trading SIGNAL indicator rather than the OUTPUT indicator because the signal tells you whether you should enter or exit a long or short position. If you use the output, you have to compare that value to a threshold. The Signal implicitly knows about the thresholds.

Below are the signal values and meanings. The signal starts out at 0, which is a neutral (not in any position). The signal stays at a given value until a new entry or exit occurs.

signal = 1:

Enter into a long position on the next bar.

signal = 0:

Exit the current position on the next bar and enter neutral (no position).

signal = -1:

Enter into a short position on the next bar.

To implement the ChaosHunter signal in a NeuroShell Trader Trading Strategy rule, use the  A = B indicator from the Relational Category as follows:

Long Entry:  signal = 1

Short Entry:  signal = -1

Since I had specified a true reversal when I created the ChaosHunter model, I didn’t use Long and Short exit rules in the Trader chart.  Also since I was using true reversal in ChaosHunter, in NeuroShell Trader I had to turn on the option for Long/Short entries exit existing short/long positions on the Sizing tab in the Trading Strategy Parameters to insure the results will be the same.

   

The NeuroShell Trader optimizer decided which ChaosHunter signals to use in the trading rules.

Note that this ensemble system differs from others we have featured in our newsletters where two or more trading signals had to be true in order to enter a trade.  For this example, the optimizer in the Trader was set to rule selection and only one rule had to be true in order to enter a trade.  The idea behind this set up was that the Trader’s optimizer would select from several diverse models the one trading system that would work in current market conditions.  Optimization should be repeated on a regular basis to select the best model for existing conditions.

Was this article helpful?

Yes No

Related Articles