I Now Have Delisted Stock Data!
Despite the fact that this will absolutely confirm me as a geek, I am very excited to say that I now have delisted stock data, going back to 1985.
Without delisted data, one runs the risk of Survivorship Bias creeping into his backtests.
I got my data from Norgate Investor Services, (the same folks that provide my end-of-day feed). They only charge a one-time fee for the delisted data, while some of their competitors charge as much as 3x Norgate’s one time fee with the charge recurring annually!
Since adding the delisted database, I have not noted any great differences in the historical results of the systems I work with. I have stated a few times that it is my belief that short-term systems that hold stocks for a few days to a week are not likely to suffer greatly from survivorship bias. So far, this belief is proving to be true.
The new data will prove invaluable as I start working on some long-term trend following systems.
you definitely are a nerdtron
Clearly and geek and a nerd, but so are most of the rest of us or we wouldn’t be up late at night trying to beat mother market!
Keep it up!!
5-0, did you design the TF systems you were speaking about on Jake’s blog?
Does that delisted data come in as CSV files? Or do you connect to it directly?
I know this method isn’t exact but I used some app to pull historical data from Yahoo to write up my own program to play around with backtesting data … hoping to get CRSP data through school eventually.
_johny, I connect to it directly, but then it is saved (and is probably transmitted) in metastock format.
AmiBroker has an app that will pull yahoo historical data. If you get serious about it, you will need to a real data provider. You do not want to put real money on the line if you aren’t sure the data is bomb proof.
Completely agree … but until I have enough time to put towards something serious I’ll just be playing around with some theories from here and there.
Good show. Nice to see you representing good modeling practices.
Not that I ever thought you would do otherwise…
Thanks Cuervos. How’s the new gig?
Hey Wood,
As you said it would be interesting to see how longer term strategies faired with this added information, but it would also be nice to see how a short only strategy faired with this information. Primarily one that target stocks that would become canidates to be delisted.
During the tech crash I heard someone say that stocks that fell under $10 for the first time in their history where more likely to see $5 before they saw $11. Of course the problem backtesting this strategy without the data you now have is that so many of the winners would not show up in the data.
Just a thought. Thanks for passing this info along.
Red, I’ve been trying to figure out a test to put this data to use. I like your idea of <10 going to 5 before 11.00 Should be pretty easy to test. I’ll try it out soon and if it shows anything interesting, I”ll put a post up.
One thought I had about that, is how to handle split adjustments.
For example, here is a lifetime chart of MSFT: http://finance.yahoo.com/q/bc?s=MSFT&t=my&l=off&z=m&q=l&c=
The price of MSFT according to this data was less than $10, but I am assuming, and I have not done the calculations, that if you reversed the split adjustments MSFT was never below $10. I do not know how the data provider handles such corporate actions, but it is just one more layer that might make your test more complex, but hopefully not impossible.
Just thought I would throw this out there.