In this issue:
I. Live from the Forum
II. Commentary by Steve Ward, CEO
III. Voting for the Readers’ Choice Awards
IV. ChaosHunter discount for NeuroShell Trader Pro users
V. Office closure
VI. One way to stop this newsletter
***********************************************************
I. Live from the Forum
Topics lately on the NeuroShell User Forum include:
ETF? Or Stock?
Portfolio Systems – Building from Predictions
Limit Price
Overfitting – when to trade it
What should a beginner do???
Twitter for stock prediction
Using ChaosHunter to optimize NST Trading strategy!
Real DLLs created with VB6?
***********************************************************
II. Commentary by Steve Ward, CEO
1. Selecting the in-sample periods for trading models. This topic is important whether you are optimizing rules or training and optimizing neural nets. It is relevant for NeuroShell Trader Pro, DayTrader Pro, or ChaosHunter. Next to choosing correct rules and/or net inputs, it could be the most important part of the art of building good trading models.
They say variety is the spice of life, but it is also the spice of modeling. Optimizers work by finding a formula (like a prediction formula or set of trading rules) that work the best over a set of data usually called the “in-sample” or “training” data. What often happens is that your in-sample period is a long trend, usually upwards, and so the model only winds up learning about up-trends, which is great as long as the up-trend continues as long as you trade the model. Even if there are slight pullbacks in the in-sample up-trend, the optimizer will probably learn to ignore all but the largest of the pullbacks, because it realizes more money can be made just going long most of the time. That is, it takes the safe path of “when in doubt, go long”.
So suppose you are using daily data and in addition to the recent slowly rising bull market we have had, you decide to include the big crash in 2008. The flip side can come into play, because the crash was so sharp, the optimizer need only go short during the period to make a boatload of money. It makes so much money short in the bear market, the gains in the slow bull period don’t add much in relation to the bottom line. You might need to add the slow bull market before 2007 to help balance things out, because balance is important.
This brings us to what I will call focus. The traditional wisdom is that the longer the in-sample period the better. This use of a “sledge-hammer to kill the fly” wisdom grew out of the need to include variety and balance in your in-sample data, and the faulty (I believe) assumption that you can find a model that works well forever. But just starting your in-sample data as far back as possible might be counter-productive. For one thing you may get so much variety that the optimizer learns to only detect the largest moves correctly. For another thing, the variety that you capture and learn could be so old as to be useless tomorrow. It might be better to focus on a smaller part of that most recent slow bull market.
But won’t that cause overfitting? It could, but if you are careful and focus on a period of cyclic activity with good inputs/rules, the result might be that the model is better able to detect the onset of smaller activity and reversals. There is less hay to hide the needle. Yes, you might have to retrain or reoptimize as the markets shift.
Clearly the issue that you are trying to trade matters, in view of the above.
You could even isolate segments in the time series on which to train. The easiest way to do that is to use the Toggle Indicators in Advanced Indicator Set 3 in NeuroShell, and the new Time Flag series feature in ChaosHunter. Contact tech support for more details if you are interested in pursuing that.
2. Please vote for NeuroShell. If you are a subscriber of Technical Analysis of Stocks and Commodities Magazine, we need your vote. Please do not sit this election out, because continuing to bring you the free tech support we have for 22 years really does depends on a good showing in the Readers’ Choice awards. We have never charged anyone for help, and we do not want to start. Some of you do not require much help, but someday even you might need it. The next article tells how to vote.
3. A fun experiment. Neither NeuroShell Trader or ChaosHunter was built for any kind of visual pattern recognition, but that doesn’t mean it wasn’t fun to play with something simple. Since you can see the neural net formulas that ChaosHunter evolves, I wanted to see what kind of net formula evolution would create for pattern recognition of printed digits.
So I set up a ten grids of 5 rows and 3 columns each, with dots in the grids to draw the ten digits. There were several ways each digit could have been drawn, but I only made 10 patterns. (Thus I ignored the spice of life advice above, which would have been necessary for a more robust digit recognizer. But again, ChaosHunter was not built for this type of classification anyway.)
I labeled the rows R1 through R5 and the columns C1 through C3. Therefore, there were 15 inputs, e.g. R3C2. In the training data each input was either 0 for missing or 1 for dot. So this was a seven:
1 1 1
0 0 1
0 1 0
1 0 0
1 0 0
The outputs for the ten examples in the training set were 0 through 9. I turned on only the neural functions and scaled the output. It took a while but eventually every digit was recognized in an interesting evolved net, which consisted of a string of four neurons, each feeding the next, and another such string of two neurons. Each of the neurons in the string had 3 or 4 inputs. Both strings fed into a final neuron which also had an input. I have posted the training data and model on ward.net in the ChaosHunter section.
NeuroShell Trader Pro solves the problem too, with input selection turned on and set to minimize R squared.
4. Something I learned about ChaosHunter. The documentation says ChaosHunter will save runner-up models as well as the best one, but it doesn’t say it will save as many runner-up models as you want to save. I tried saving multiple ones the other day and it works just fine. You just have to first apply (backtest) each model you want to save, and give each model a different file name. This means you could fire multiple models in NeuroShell or other trading platform, perhaps as a panel of experts.
***********************************************************
III. Voting for the Readers’ Choice Awards
Here is how to vote as Steve Ward requested in his commentary:
1. Get your copy of the Stocks & Commodities magazine and visit www.Traders.com (any issue will do; it doesn’t have to be the October one). Look on the left side of the web site for the 2011 Readers’ Choice Awards and click on the Vote Now! link. On the page that is displayed enter your subscriber ID from your magazine address label (look at their example to see which numbers on the label are the subscriber ID). Next enter your last name.
2. After you are logged in, look for the Standalone Analytical Software, $1000 and more category on page 2. Be sure to mark you ballot for NeuroShell Trader Professional (Ward Systems Group Inc.)
3. Do the same for the Artificial Intelligence Software (Expert, Neural) category on page 3. Look for NeuroShell Trader Professional (Ward Systems Group Inc.). THIS IS THE MOST IMPORTANT VOTE!
4. Also in the Artificial Intelligence Software category, we’d appreciate a second place vote for GeneHunter, the genetic algorithm optimizer that is included in NeuroShell Trader Professional.
5. Also on page 3, right before the Artificial Intelligence Software, there is a new category called Software Plug-Ins. Be sure to vote for MESA 8 for NeuroShell Trader.
The staff at Ward Systems Group Inc. extends a very heartfelt thanks to all of you who take the time to vote for us. Please do it soon before you forget.
The winners will be announced in the 2011 Stocks & Commodities Bonus Issue that will be mailed in late February.
***********************************************************
IV. ChaosHunter discount for NeuroShell Trader Pro users
If you are already an owner of NeuoShell Trader Professional or Daytrader Professional, we will give you a 20% discount on the purchase of a new ChaosHunter until Thanksgiving Day, November 26, 2010.
If you like neural nets, read this carefully:
www.chaoshunter.com/chnn.asp
***********************************************************
V. Office closure.
Ward Systems Group offices will be closed November 25 and 26 in order that our employees may celebrate the Thanksgiving Holidays with their families. We will reopen November 29.
***********************************************************
VI. One way to stop this newsletter
It is really easy. Just change your email address and don’t tell us.
***********************************************************