In this issue:
I. NeuroShell Trader wins again!
II. Multicore optimization in release 6.0
III. Commentary by Steve Ward, CEO
IV. Reduced staffing days
V. One way to stop this newsletter
***********************************************************
I. NeuroShell Trader wins again!
For the ninth year in a row, those of you who are subscribers of Technical Analysis of Stocks and Commodities magazine have rated NeuroShell number 1 in artificial intelligence software. We are extremely pleased and grateful to all of you who voted. We strive to provide all of our customers the best software we can produce, as well as the most responsive technical support in the industry. This rating will assure that we continue to get new customers so that we do not have to charge the old customers for this support. Thank you all.
PS – If you are not a subscriber, please consider subscribing so you can vote next year. The magazine will also give you complete articles describing the tips which we implement every month. There are many other articles which are analytic in nature and lend themselves to NeuroShell implementation.
***********************************************************
II. Multicore optimization in release 6.0
Beginning in release 6.0 you have the choice of distributing optimization over some or all of the cores in your multi-core computer. If your cores are also hyper-threaded, each core will act like two cores to NeuroShell. That means if you have 6 hyper-threaded cores on your computer, NeuroShell will divide optimization into 12 parts so that the theoretical speedup is 12 times. If 12,000 individual backtests are required during optimization, approximately 1000 will be accomplished on each core.
In practice, of course, you will never reduce optimization times by the full 12 fold. That is because there is overhead required to fire up 12 cores and communicate between them during the process. Therefore, the longer each backtest requires, the more speedup you will see. If backtests are very fast to begin with, it is possible there will be no speedup at all.
If you are backtesting with the Paper Trading feature enabled (see example 10), less of the work can be distributed to the “client” cores because of some extra optimization algorithms at work. In fact, in the first release of 6.0, we didn’t get those extra algorithms working exactly the same during distributed optimization vs non-distributed. We have now corrected that problem in 6.1 beta 3, available on ward.net.
***********************************************************
III. Commentary by Steve Ward, CEO
1. Seminar – I am looking foward to seeing some of my old friends and meeting some new ones at this year’s seminar at the end of the month. Many of you I call my special friends have attended more than one seminar, and some of you have been to three or four I think. Marge and I have put a lot of work updating the content. See you there.
2. Investing – I do not risk all of my money in trading, as you might suspect. I also invest long term in stocks – dividend paying stocks almost exclusively. I use a service called Dividend Detective to help me do the stock picking. Dividend Detective doesn’t just give me top yielding stocks – anybody can find those, but often stocks have high dividends because of other problems and risks they carry. DD publishes portfolios of picks that change based on their analysis of fundamentals and news. To me it is well worth the $15/month they charge for the premium service. If you like dividend investing, check it out at www.dividenddetective.com.
3. PowerBasic. As most of you know, I like to do my DLL programming with the Power Basic compiler. It is easy to learn and creates fast, efficient code. There are many examples of indicators programmed with Power Basic on ward.net. Although the latest release is 9, costing $199, and release 10 is due out soon, at this writing they are still selling release 8 for $49. They call it Power Basic Classic for Windows, and it is perfectly fine for programming NeuroShell indicators. I used rel 8 extensively for a couple of years at least. If you ever even had a random thought about programming NeuroShell indicators, go pick up a copy. At $49 you can’t go wrong if you later decide to act on that thought.
4. Release 6.0 time restriction. One fairly obscure feature that we added to release 6.0 may be one of the biggest time savers for many of us, myself included. I like to do my intraday trading only in a specific time zone during the day, usually from 10 to about 2 or 3. Even if I stay in later than that, I usually want to exit my trade before close to avoid overnight surprises.
In release 5.7 I had to do the restrictions with complex time rules inserted into the trading strategy itself, like Time >= 10 AM. If there were multiple rules and I was doing rule selection optimization, it got much more complicated to make sure the time rules didn’t get optimized out. (Okay, in release 6.0 there is now the ability to lock a rule so that it cannot be optimized out even during rule selection of other rules, but that is not the feature I am discussing.) I couldn’t do anything with predictions at all unless I put them in a trading strategy.
In release 6.0 both the Trading Strategy Wizard and the Prediction Wizard now sport a little time box that allows you to set the beginning and ending times of the desired trading segment. No rules (or nets) fire before the start time, and all positions are exited automatically by NeuroShell after the end time, no matter what optimization does to the rules and parameters.
***********************************************************
IV. Reduced staffing days
Our offices will be open but operating with reduced staffing March 29-31 on account of some of us attending our Advanced Training Seminar. There may be some delay in processing your requests during the entire last week of March.
***********************************************************
V. One way to stop this newsletter
It is really easy. Just change your email address and don’t tell us.
***********************************************************