Rich Fordiani of California has competed at tennis to the Nationals level and set a record for sailing from San Francisco to Kauai, Hawaii. He’s now applying that competitive spirit to trading with the NeuroShell Trader. He has been trading for 8 to 10 years and has been working with the Trader for 2 years, mostly trading end of day SPY and related index models.

“For the first three or four years I traded I lost money. I used systems I purchased from other people or services. I tried a number of things that were not successful so I built my own systems. I started to get profitable,“ said Fordiani.

“Then I felt like I outgrew traditional trading systems. I looked at other neural nets and found the NeuroShell Trader easier to use. You can include other data such as the dollar in models for indexes. That is not as easy to do with other systems,” according to Fordiani.

He prefers systems rather than just looking at indicators on the screen and making a trade. “I like to be able to back test and look at results. I currently have five models built with the NeuroShell Trader that I actively trade. I trade multiple indexes, mostly related to the S&P. One of the systems uses neural indicators, another Adaptive TurboProp 2, one uses the Trading Strategy wizard, while another uses the Prediction Wizard. Another uses a system of five neural nets that I put into a Trading Strategy. All of the nets have to agree in order to trade. I started out with three nets like one of the examples that come with the Trader. I gradually revised the system so it had five nets. I found that individual nets had too many trades. Combining them made them more profitable,” he concluded. He’d next like to have the optimizer tell him which net is working best.

“My usual procedure is to download data at night and open up my models in the Trader to see if there is a signal. I use a reversal system without money management or trailing stops,” said Fordiani.

He currently retrains the nets once a month, but how often to retrain is the problem he’s currently working on. “I’m testing by training the nets on a month by month basis. I start training from scratch and look for different results. Once a month is working out ok, but I will try quarterly. The model that uses Adaptive Turboprop retrains every 10 days. I go back about 2 to 3 years to test. I found out there’s a lot of ways to do it, that’s why I trade multiple systems. I evaluate my results based on the percentage of winners compared to losers, the number of winners compared to losers, the Ratio Avg Win/Avg Loss and maximum drawdown.”

When asked if he traded every time a model produces a signal, he quickly responded that “You have to take every signal or throw the system away.” Following his models has produced solid results. “Last year my returns were greater than 45 percent and this year I’m up 20 percent so far. I’ve been successful but you can’t just put your feet up on the desk and relax. You always have to be working on something.”