Add Backtesting to the Screener
What would be really useful is the ability to run the filter on a specific dataset e.g. Finals December 2005. All that is required is the set of stocks a filter might have produced for the dataset, together with a price on the final day of the month. The user could then manually work out what to do with all the changes in ownership and get an idea of performance. I accept that running on specific dataset limits the shares that would be thrown out but I can't think of an 'easy' way to do it. What do you think?
I like the idea of forward testing but would prefer to keep my filters and their performance 'close' until they are proven to work. I guess that means a 2/3 year lag from when you make the facility available but it would surely be worth it in the end.
We discuss backtesting a bit here – http://help.stockopedia.com/knowledgebase/articles/80621-can-you-add-longer-timeframe-performance-histories
Doing backtesting badly (as some sites do) is easy but the results are more harmful than anything else. Doing it well and dealing with survivorship bias etc. is much harder but we do plan to do this. Timing is uncertain though given the complexity.
Backtesting would be really helpful. If the cost of computing power is an issue - you could charge and extra add-on for those who want to use backtesting functionality.
Not backtesting per se, but the ability to run a screen against values in the past. E.g. I have a large cap value screen, run it at 01/01/2013. I can then build a model portfolio from the results, at 01/01/13 prices and track its performance over time.
Backtest facilities, so users can set up a screen, select (say) 10 stocks that comply, and test these over (say) a year.
Backtesting is a full time jobs, and I would really would like Stockopedia to allocate ressource on over features. We already have plenty of academic material. Let's use that.
The only problem with these academic materiel is that they target institutional investors and not individual investors that invest only in 10-30 stocks, and have other liquidity isssues.
Willy GABRIEL commented
Possibility to backtest my Screen
I quite understand that finding clean (and cheap!) fundamental historical and price data with no biases is hard (especially for the UK market).
Stockopedia has now been running since January 2012 during which time I assume you've built up your own "historical" fundamental and price database going back to the date (please tell me you've kept a point in time db of all the changes ;) ).
If that's the case it should be a fairly simple matter to support backtesting going back to Jan 2012 with no survivorship or look-ahead bias.
Of course backtesting only to 2012 is not ideal (e.g. no big market crash like 2008 to backtest over) however it would still be a a hugely useful and compelling feature imho, and over time the database will grow anyway :)
I would love to see a backtesting feature, along with a feature where the user could specify rules in their portfolio for how to balance it and when to trade.
This could be used both for backtesting with historical data, but also for live/demo trading with a real broker. Long/short hedge portfolios could be supported if used e.g. with spread bets/CFDs
If the trading / rebalancing rules could be scripted even better. I'm basically thinking a bit like Quantopian but for the UK.
It would be nice to see the past performance of a screen. You have this hidden gem "Analysis"-Tap in the Portfolio Section. Please make this available for the screener interface.
To save computing resources you could limit it to the first 25 stocks of a screen.
This could be a nice pre stage for backtesting (though it only would show the performance of the current stocks).
I often find myself creating temporary portfolios from screen results just so I can view simulated performance (i.e. clicking on the "analysis" tab) to get a rough idea of how the screened stocks might have performed if bought a year ago.
It would be nice if we could get this functionality directly from the screen results page without having to create a portfolio first.
I feel another essential feature is a log of what has dropped into and out of your screened selection.
Can't be difficult to implement as we do it on the Guru screens? Or is that done manually?
Creating an investment strategy without at least some scientific basis / test data is just guess work and does not give you the confidence / conviction to stick with it. Worst yet, its feedback loop could be 2-3 years without a back-testing. I agree, back-testing is a must
This is a must.
Dave Brickell commented
See the discussion here: http://www.stockopedia.co.uk/content/seeking-safe-value-69448/?comment=4#4