stockfetcher
980 posts
msg #119432 |
5/7/2014 1:50:44 PM
We're looking for your feedback on changes to how StockFetcher addresses analyzing past filter results.
We understand that an important part of developing a StockFetcher screen is understanding how it performed in the past and how to modify a screen to improve the outcome; however, StockFetcher's back-testing platform admittedly has multiple shortcomings. These shortcomings result in a back-testing offering that is nowhere near as robust as other tools available on the market today.
We've always stressed that StockFetcher back-testing is used as a gauge of how a filter performs in the past. However, given the naming of the feature and the options available today, more stress (than intended) is placed on the trading mechanics of a back-test.
That said, we are looking to eliminate any emphasis on "simulation" and focus more on "analysis" of historical filter results. We want your feedback on what functionality you would like to see. For example:
- specific statistics
- possible export options for 3rd party simulators
- report formats
- public filter ranking statistics
Also, please let us know if you have any preferred simulation tools, and how we might be able to work on bringing StockFetcher results into these other tools.
Finally, if you can't live without StockFetcher back-testing, let us know why that is. It is just that kind of feedback that would help direct us in the replacement functionality.
Thank you!
Tom
StockFetcher Support
|
karennma
8,057 posts
msg #123136
- Ignore karennma |
3/12/2015 10:48:55 AM
"EDIT" feature currently not working on filters.
Just an FYI ...
|
Kevin_in_GA
4,599 posts
msg #123151
- Ignore Kevin_in_GA |
3/12/2015 2:45:57 PM
Thanks for taking on what has been a chronic deficiency in SF. I would propose the following as ways to improve the backtesting functionality:
1. Allow for a longer backtest for all levels of subscribers - a four month window for most subscribers (the basic subscription) is simply too short to get a real sense of performance, and what folks have to do now is to stitch together results which is a pain in the butt. A minimum of 5 years to validate any decent system. Perhaps you can go to 10 or more years for advanced subscribers.
2. More useful metrics - The annualized ROI is basically useless. I like that you show actual equity growth, but that is the absolute minimum needed. I would include some measure of alpha, beta, and at least the Sharpe ratio and/or Calmar for any backtest results. Include a system MaxDD as well.
3. Limit Order entries and exits - not sure if this is done now since I don't use SF for backtesting, but most systems can be adapted to use limit orders based on the prior day price, and exits triggered intraday. An example would be "limit order entry at reversersi(2,5)" or "limit order exit at upper Bollinger Band(20,2) 1 day ago". If this is not already in place it certainly should be.
|
mahkoh
1,065 posts
msg #123206
- Ignore mahkoh |
3/15/2015 4:56:56 AM
Could you add ^SPX data to be used for single symbol backtest? When using SPY there's always the close/adjusted close issue. Also add volume data to make volume related indicators work.
|
abelduarte
9 posts
msg #123774
- Ignore abelduarte |
5/12/2015 2:04:04 AM
Support for more complex strategies such as selling part of your position after a certain goal is met would be good.
|
sohailmithani
192 posts
msg #124041
- Ignore sohailmithani |
6/16/2015 3:56:18 PM
Looking at subscriber comments to this one I do not see a call fro "eliminate backtest" instead need for more to the functionality.
How come SF has just decided to discontinue back test. Please help me understand how we will develop new strategies and share them with other users here without even testing them. What does it mean to focus more on analysis and less on simulation. Analysis of what????????
|