Enough sample data?

[mvalley]

Quote from ntfs:

Mike,

Is there some place that records stocks that aren't listed anymore?

CSI is the only supplier that I know of for delisted stocks. Are you looking for stock lists for a certain indices? If so what indice?

nt

I just need to know the old symbols for the delisted stocks on NYSE, AMEX, NASDAQ & OTCBB.

Would CSI be able to provide information that is easily importable into AMI?

 

[mvalley]

Quote from OverKill:


Finally, in my first post I forgot to ask: are you checking for postdictive erros? This is the worst mistake one can make when coding a system. Look very carefully in your code for them.

Regards.

I'm not familar with that term. What kind of errors are those?

-Mike

 

[mvalley]

Quote from gbos:



Actual number is between:

backtesting result – 1.96 * standard deviation /sqrt(sample size)

backtesting result + 1.96 * standard deviation /sqrt(sample size)

with 95% confidence.

regards

This is very useful, Thanks!

-Mike

 

[OverKill]

Quote from mvalley:

I'm not familar with that term. What kind of errors are those?

-Mike

Commiting a "postdictive error" is the same as "peeking into the future". It occurs when you design your rules around information that would never be available when you actually implemented the system, for example, "enter long at tomorrow´s low and sell at tomorrow´s high". This is an absurd rule, since there is no way of knowing before hand which one will happen first or when they will happen. Postdictive errors don´t need to be so outrageous like this, but anything going along this lines will likely distort your results substantially.

 

[ntfs]

Mike,

I just need to know the old symbols for the delisted stocks on NYSE, AMEX, NASDAQ & OTCBB.

Sorry, can't help you with that. I did find where you can get some history of symbols of several indices.

Would CSI be able to provide information that is easily importable into AMI?

I don't use CSI but I have been thinking about getting it just for the delisted symbols. I heard it cost extra. Not sure how much.

I think on the main Yahoo/Amibroker board there is a folder with instructions on doing this.

 

[mvalley]

Quote from OverKill:

Commiting a "postdictive error" is the same as "peeking into the future". It occurs when you design your rules around information that would never be available when you actually implemented the system, for example, "enter long at tomorrow´s low and sell at tomorrow´s high". This is an absurd rule, since there is no way of knowing before hand which one will happen first or when they will happen. Postdictive errors don´t need to be so outrageous like this, but anything going along this lines will likely distort your results substantially.

That's funny, I commited one of those when I first got into backtesting. I accidently based the buy limit price at a support point calculated from that day's data.

Needless to say, the results were too good to be true

-Mike

 

[Hook N. Sinker]

I want to test on minimum 10 years of price data. Some securities have more than 20 year histories. I vaguely remember seeing over 40 years of price data for corn futures. GE stock might have about 44 years price data. I suspect the longer the better.

Sometimes I break the data into parts, optimize on one part and backtest with another part. Sometimes the results are as expected and sometimes very different.