Neuroshell

Does anyone here uses neuroshell software for designing and testing trading strategies? Any comments on how this software ranks vs Tradestation, Amibroker etc?

I can see an obvoius negative that you have to buy data from selected vendors instead of downloading it from free sources as amibroker does.

How good or bad are the backtesting abilities?How does the backtesting ability compare with the other software?

How easy is the use? I understand that you can write your strategies in NS without learning much programming.?

How good and open is the support and users interaction. I know they have a dedicated secret site for users. Are many people active there? How large is the community and how frequent is the interaction. I know there is an unofficial group on yahoo which is not very active specially compared to the amibroker group on yahoo which is very active.

I will appreciate your comments/experiences, I am trying to select a software platform for myself and neuroshell appears to be a good candidate.

many thanks for your help.

 

I have both NST and TS. I use NST mainly for realtime trading and TS for experimenting with new ideas. I’ve ported many of my TS strategies over to NST using DLLs.

In some respects NST is state-of-the art (Genetic algorithms (GA), neural networks (NN), fuzzy logic plug-in, automatic walk-forward optimization, etc.) when compared to TS. If one isn’t familiar with or has a use for the new technology, TS may be a more comfortable product. One drawback with TS is that its optimization engine is a brute force search compared to GA optimization in NST.


Re: “ you have to buy data”

You can import ASCII formatted data into NST from any source even free ones. But this is good only for EOD or for a training data set and not for realtime trading. You will have to buy realtime data if you are going use NST for realtime trading. TS can use tick data where NST can not.


Re: “How good or bad are the backtesting abilities? How does the backtesting ability compare with the other software?”

NST is better (1) because of the GA optimization, (2) far more objective functions to select for optimization, (3) automatic walk-forward optimization, and (4) portfolio optimization. But TS provides a far better set of test result data and graphs. Additionally, TS saves the top 200 optimization sets so you can pick the best one. NST saves only the best one. Many times you need to trade-off ROA, MaxDD, PF, etc. in order to select the “best” strategy.

NST will let you optimize a strategy for a portfolio of stocks (one set of parameters for all stocks). TS optimizes one stock at a time (N sets of parameters for N stocks). I believe that portfolio optimization is far more robust.


Re: “How easy is the use?”

If you are going to use canned indicators and strategies, TS is probably easier to use. I don’t like NST’s point and click “programming interface” especially when compared to TS’s EasyLanguage; it’s a real pain in the butt for complex strategies. Although non-programmers will probably prefer the point and click interface. One can just randomly add canned indicators to the prediction wizard and the GA algorithm will automatically find the best rules and parameters for you.

NST provides lots of options in its user interface, so many in fact that they have hidden much of the functionality because new users found it confusing and didn’t understand how to use them. For example automatic walk-forward optimization is not available until the user enables that functionality in the options menu.


Re: “How good and open is the support and users interaction.”

I never used their support, so I can’t comment. Their web site has a wealth of good information and ideas, including monthly strategies from S&C magazine that you can download. The discussion group has some great ideas as well but you have to read through the threads to find the nuggets, so its not handed to you on a silver platter. The site is not very active averaging a handful of posts each week with maybe one or two new threads a week. The web master answers technical issues typically within a day. But the activity comes in bursts, especially when an interesting topic is posted.

User interaction on the forum is minimal and is nothing like what you have here on ET. What you will not find is a lot of sharing of profitable strategies (or even indicators sets used in them); which may indicate that some users have developed strategies that they want to keep close to their chests, or alternatively that they have failed to develop anything worthwhile.

 

Many Thanks for your great post on Neuroshell, its comprehensive and answers the questions really well.

One clarification, You have written, "TS can use tick data where NST can not" I guess you mean NST cannot but NST day trader can probably use tick data?

Do you use only TS and NST?

Have you ever used Traders Studio(www.tradersstudio.com) Amibroker or Wealthlab.?

Any comments on Strengths and weakness of these platforms?

Many thanks for your help.

 

So like.. are you making money with NNs? I'm still waiting for someone to say that

 

sunnyskies,

many people at ET are posturing making money with all kind of things.

 

Friends money making with NN is possible though I have not seen many do it. But then I have not seen many make money trading anyway.

I know acrary on Elite Trader uses or has used some neural nets with good results.

 

Acrary is back and in a post today (9/26/05) he mentions using NN's for setting up strategies. Here is an extract:

If you hit at least 85% correctly then you can do one of two things. If you're a discretionary trader, setup the NN and preprocess the inputs every day and use the net to predict whether tomorrow has the target (in this example the low of the day is within 15 min. of the start of the session).

Jack

 

I wonder if more people here have used Neuroshell and can give their comments on it.

 

Yeah, threw it in the garbage bin 5 years ago.

 

I bought it a few years ago - along with a few of the add-on indicators (clustering, etc.)

I spent many many hours using it, optimizing, building NN's, etc., etc. Months and months of ideas and CPU usage. It is an amazing product in all it can do. You can make things so very complex. I thought of all sorts of methods and systems.

As a product, it is probably the best in the sector.

Did I make a lot of money with it? Not a dime.

Well, I do credit it with one thing - my realization that SIMPLE really does work and is the best in the long run. NN's, GA's, clusters, all that - or a line or two. The line or two isn't as cool for sure.