Home Page
Yahoo Charts

    Feb 24th, 2001 : GPF Pro V1.44 reposted.  There has been a few occurences of registration problems. It has now all been corrected. Note that it is so minor that there is no change in the version number.

    Feb 17th, 2001 : GPF Pro V1.44 A few additions mainly to respond to the growing interest from day traders.  Intraday specific settings have been added, and more generally signals can now be generated using Stop or Limit orders.  A few GPF V1.43 have been out for testing, so this release is V1.44.

    Nov 30th, 2000 : GPF Pro V1.42 is announced. It adds important statistical calculations to better validate GPF patterns (T-Stat significance and serial correlation of signals).  A new web based user registration is being implemented.

    Nov 28th, 2000 : NeuroShell Trader users can purchase the GPF Pro with 33% discount!! , i.e. only $100!!!

     Oct 31st, 2000 : GPF Pro V1.411 adds access buttons to the ForeTrade web site, and a few more comments. It also solves insignificant errors, and added the comments and cosmetics to make it more pleasant to run.  More seriously, a true stop loss feature has now been implemented.  This is consistent with our strategy to bring more and more trading parameters into the signal definition.  Unlike most if not all other products, we believe that the quality of a signal must be related to the subsequent trading strategy.  Too many practitioners on the contrary calculate fancy indicators and then work out a way to use them in their trading.  This is too restricting.  Yet, by the same token, we have too agree that the GPF does a LOT of calculations, and is considerably slower than other optimizers.

The GPF Pro is considered as final. The plain vanilla GPF is still available and supported, but will not include the latest features. The GPF Standard now costs $75, and entitles to the GPF Pro for just the difference in price. It is still a good way to start experimenting features at your own pace.

    Oct 20th, 2000 

1.    A new exciting on-line service will soon be launched.  It is called WebGPF, and will be a Excel 2000 spreadsheet which automatically downloads data from Yahoo and patterns from our site, calculate GPF composite signals to be used in your trading strategy.  The latest version of the GPF Pro does generate WebGPF sheets for your own use, or for FREE distribution over your own site.

ForeTrade will post WebGPF sheets on this web site.  Some will be free for all users.  Some will be available at a small nominal fee.  All WebGPF patterns will be free for GPF registered users. More details coming soon...

2.    The GPF Pro will in this upcoming release better allow for Out of Sample (OoS) analysis.  It was previously already possible to load data, run an optimization, analyze stats, then load new data to check how the newly found patterns fare on 'unseen' data.  While this is still possible, it is now easier to just hold some data out at the end of the sample, and calculate stats to analyze pattern statistical stability.

The Analysis form now shows OoS statistics which can be calculated either cumulatively (In-S + OoS), or separately (OoS).

3.    Internal genetic settings have been improved including default parameters, usually only available to GeneHunter registered users.  While this will increase optimization times, there also brings much better patterns.  It was previously not uncommon to see duplicate optimizations yield different patterns.  Among other things, pattern populations are bigger, and the genetic selection process is even more elitist.

Default parameters should be updated to allow Fitness peaks to be at least 30 to 40.  As a guideline, a regular optimization should take a minimum of 100 generations for 2 to 3 years of data to reach an optimum.  Some advanced users, particularly academics, have argued that fitness must peak for 75 to 100 generations to ensure an optimum solution.  This may be true, but our experience showed that the best patterns are often the second or third one, which often display better statistical stability.

Today: - Page last modified: December 08, 2007
Copyright © ForeTrade Technologies 21st century and thereafter