First of all, thanks to everybody that contributed money to this project. The money contributed allowed me to upgrade the hardware increasing the backtesting capability significantly, which should benefit all users here.

As an experiment I have decided to stop accepting contributions for now and test Google AdSense instead. If money trickles in I plan to use this to subscribe to additional data services, and more hardware upgrades when that becomes useful. 

Various updates today:

Daily data

Free

Trendvest. Stock Name & Symbol Changes, Stock Splits .

Commercial

Marketsource Online. Free trial .

EODData, name says it all .

Added a new function named MAXEQUAL to the backtester today which people have asked for. Typical usage is:

I recently backtested all the standard screens from 1997-08-31 using a 28-day/monthly trading cycle both with and without the new MAXEQUALfunction recently introduced in the backtester (holding top 10,holding cash whenever there were not enough picks).

 Most screens (with a couple of exceptions) seems to benefit - as measured by the CAGR/GSD ratio and/or UPR ratio (using a MAR of 1%/month).

I ran the screen first as normal, and then by adding MAXEQUAL(1,[SI Sector]) after the final sorting step in each screen. 

DELTADAYS(date1, date2) returns the number of days between the two dates. If date2 is later than date1 the number will be positive, or else negative. The typical usage will typically be:

Mechanical investing is about designing investing strategies that anybody can do - possibly even automate - by following typically backtested strategies. Using one or more screens which contains selection criterias with a sorting filter, a list of stocks is generated and one picks from the top of this list (typically at least a few stocks deep).

You can find examples of automatically generated lists and strategies in the stock picks area.

'The backtester has a method for working with short screens designed to also work in blends (mixing long and short screens). By making sure a screen ends with "_s" (underscore + s), the backtester will treat the screen as a short screen and reverse gains and losses for each period. So if the backtester reports a positive CAGR for a short screen it means you would have made money by following the screen.

I have now updated the study to avoid the backtester putting all the
money in a few stocks when enough picks do not satisfy the ndmax price
criteria, and will comment on the differences from my original study
below.

In general, my initial conclusions looks wrong whereever fewer stocks
were available, so I will restrict my comments where less the number
of stocks actually held was 3 or less (of the wanted 5) for several
settings of the ndmax variable.

The SIPRO backtester at http://keelix.com/ traditionally uses closing
prices on Fridays for selling AND buying. Since the rankings are
reported on Sundays for closing prices on the preceeding Friday, it is
impossible to do exactly what the backtester does. Research done
previously on this board have indicated that one or more trading days
delay should not significantly affect returns.