Sebastian Fainbraun, who sometimes goes by the name Alex, is a 26 year old graduate of George Washington University. He studied finance there, and then made his way to New York where he worked for a while in venture capital and corporate finance. He purchased the NeuroShell Trader about a year ago and, at least partially using the knowledge he thus gained, he landed a good job at a large international bank headquartered in New York.

Alex is a trader in the bond trading department, and the bank has allowed him to use the NeuroShell Trader to make interest rate predictions crucial to the high yield bond trading. It’s a good thing the bank has been open minded about artificial intelligence in general, and neural nets in particular, because Alex’s NeuroShell Trader system results in 85% winning picks and a 3 to 1 win to loss ratio! (The bank is only open minded to a point however, because they wouldn’t let us name them.)

Update November 19, 1998: Alex now reports that his profit for the year (which ends in September) on bond futures using the NeuroShell Trader was 393%!

Alex is exuberant: "The system is excellent – better than any package you could find on the market today. It’s amazing how well it works. My predictions are occasionally off, but never too far from the actual number."

We asked Alex about his methods. "I predict ranges of rates and the probability of being in that range. If the open is near the extreme of my predicted range, then I know there is a high probability of it going up or down."

Alex uses about 10 indicators as input to his nets, sometimes a few more or less. "You’d laugh if I told you what they are," he says a little sheepishly. "They are all just standard price related indicators. The rocket science is in the neural nets used by the NeuroShell Trader! I think the simpler you keep things the better." Obviously, Alex doesn’t need the over 800 indicators in the NeuroShell Trader!

Alex’s training sets are about 2 to 3 years of back adjusted futures data. His trading strategy is proprietary to the bank, so we didn’t find out anything about it, other than the fact that the nets are one of several indicators used in it, and that it is a "contrarian" strategy used for trading very short term.

Alex also uses the NeuroShell Trader for his personal portfolio, in which he trades stocks, futures, and options. He bases his options trading on the underlying stock predictions. Alex successfully impresses us: "My account is up 160% in 6 months."

Alex likes high volume stocks. He uses inter-market indicators for his personal stock trading: spreads between related stocks, or percent changes between them. These are easy to insert with the NeuroShell Trader. He uses these spreads both as inputs to the nets and in his trading strategies. Just as he does at work, Alex predicts stock price ranges, which then allows him to buy and sell options.

For his personal trading, Alex still uses about 10 indicators, but he adds stochastics and RSI. He also likes to look at spreads between the open, high, low, and close as well. He uses different training periods for different stocks, which he determines by looking at the stock history and volatility. He tests with 20 to 30 walk forwards of 3 to 6 months duration, depending on volatility. This doesn’t take very long with the NeuroShell Trader. "I can’t stress enough that you have to do a lot of testing," Alex advises. "Do some paper trading at first."

Alex used to override his neural nets, but he doesn’t any more. "In the long run, you lose doing that. You have to believe in your system." Alex has a great deal of wisdom for such a young man.

UPDATE NOVEMBER 2007

Guys,

Thanks again for creating such an amazing piece of software. As you know, I used your software quite successfully to trade bond futures at a large NY institution; I am now using your software for option trading (spread writing). I don’t want to get into mechanics or methods at this point because I don’t want to give anything away but I just figured out how to use your software to write options successfully 98% of the time. Now mind you that whenever I create a system, it goes through some serious statistical scrutiny before it is even traded (I have software that specifically test trading systems for statistical validity) and rather than prove that the system works, I try to prove my hardest that it doesn’t work and if that is null then I use the statistical software to make sure the positive returns are not due to luck or any sort of bias. Anyways, the method I am using is statistically valid and provides the same results over a 3 year training period and a 3 year forward test (close to 1500 sample observations). Just a little clue, people should stop trying to predict where prices are going to be in the future and concentrate on where they are NOT going to be. It’s a lot easier to prove and just as profitable!! Once you know the method, it is quite easy. It took me all of 1 hour to create the neural net for this system. I have tested in one market and envision running this in 5-10 non correlated markets.

Hope to catch you guys at next year’s seminar,

Sebastian

UPDATE MARCH 2008

I am so excited that you are coming out with a new product; I don’t know of any piece of software from anyone that gets me as excited as your products; it actually gets me excited enough that I want to write and just sing your praises.  As you know, I have had Neuroshell Trader since 1997 and that along with Excel are really the only two pieces of software for trading that I can’t live without. Over the years I have spent thousands of dollars on other pieces of well known software, yet I always come back to yours.

Recently, I started using the AI Trilogy, mostly Predictor and Classifier (the other piece is beyond me) not only for trading (just a clue, try to use classifier with volatility bands and VIX levels to identify if we are in a short term up or down trend and then use that as a filter for other long/short systems) but we are now trying to build models for our real estate investment business, as well as our window treatment company (one of the largest in the Washington Metropolitan area). We are using Classifier to try to identify possible default candidates in our multi-family properties and to identify growth trends in our window treatment business, as well as identify bid project with the largest probability of winning (we used a sample model to bid on the DC Ballpark, which we won!!).  Frankly, I’m surprised that you guys haven’t shut shop started a quant fund and kept the math to yourselves but I thank you for not doing so because I attribute some of my success to your products. Again, people considering buying your product need to wonder why you have so many testimonials and fanatical users. Again, thanks and can't wait to see what you geniuses came up with for your new software.  

Sebastian Fainbraun