yepher
359 posts
msg #33684
- Ignore yepher |
10/26/2004 11:07:40 AM
xplorer,
I ran the filter that started this thread and here are the cumalitve results: (this are probably better viewed in Excel or some other CSV viewer)
Result For,samples, overall,Adv #,Adv %,Dec #,Dec %,Unch,IXIC,DJIA
1 Day, 226, 0.3, 14.39, 2.86, 11.2, -2.66, 0.5, -0.01, -0.02
2 Day, 225, 0.7, 14.57, 4.23, 11.27, -3.42, 0.32, 0.08, 0.02
3 Day, 224, 0.88, 14.56, 5.08, 11.39, -4.18, 0.27, 0.09, 0.04
4 Day, 223, 0.87, 14.55, 5.87, 11.51, -4.74, 0.22, 0.12, 0.04
1 Week, 222, 1.03, 15.01, 6.42, 11.14, -5.21, 0.24, 0.14, 0.06
1 Month, 208, 0.65, 13.16, 13.3, 12.75, -11.9, 0.21, 0.49, 0.49
3 Month, 176, -0.95, 13.02, 21.38, 14.99, -20.58, 0.14, 0.14, 1.52
Current, 226, -2.83, 11.06, 32.51, 15, -29.06, 0.03, -1.27, -3.37
FYI,
These results are calculated by stepping through the stockfetcher performance screen form date offset 1 to date offset 300. Summing of each fields and then divide by the total number of samples.
Here is an example how one field would be calulated.
adv % = (n=1∑300) / sample count
where
n is date offset
and sample count the actual results returned for that field
|
yepher
359 posts
msg #33689
- Ignore yepher |
10/26/2004 3:34:06 PM
As you say the results are biased toward the trigger day's open value. I don't believe there is a workaround for this within Stockfetcher.
> I assume you ran this outside of SF ? How can I incorporate this testing on any and every screen?
These results were generated directly from the StockFetcher's Performance tool. Although I collect the data automagically and do the calculations automatically.
For true backtesing I generally use Metastock. I have plans to autoconvert Stockfetcher to/from Metastock (and probably TradeStation) filters but have yet to achieve my goal. That goal had the nice side effect of producing my own StockFetcher manual.
> some of my screens are for trading, and are really written to hold from 0 to 10 days ... other screens for 0 - 30 days
I believe the data you seek is contained in the result set already.
This row is what the filter produced on average after 1 week from the trigger day:
1 Week, 222, 1.03, 15.01, 6.42, 11.14, -5.21, 0.24, 0.14, 0.06
And this row is what the filter produces on average after 1 month from the trigger day:
1 Month, 208, 0.65, 13.16, 13.3, 12.75, -11.9, 0.21, 0.49, 0.49
There ia a lot of "raw" data that accompanies these brief results but I did not want to post it because it is quite long. If you are interested in the "raw" form of the data let me know and I will provide a link to download the data.
I think you are probably already aware but Cegis has done some really good work with back-testing within Stockfetcher.
If you have a programming background and are interested in creating your own custom back-testing system you might have a look at:
http://eiffel-mas.sourceforge.net/
Regards,
|