Great stuff but does this protect against look ahead bias? (Looking up data in a back test for a given back test data before they are publicly available.)

That is a good question. Quandl data is time series data, the fetcher will treat it as a DataFrame source and only make the bars <= the current bar available to tests so the issue would be the integrity of the original data coming from Quandl and not so much the backtest itself. I would use the pre_func and post_func with the fetcher to make sure the data has the correct index. My understanding is that Quandl only aggregates data from other sources and doesn't alter it so the time series data should be dependable.

I would check out their API for details.

David, thanks for sharing! This looks great.

For anyone who is curious: there is indeed potential to introduce look ahead bias in your algo. If you're fetching daily data while running an algo in minute mode, then the the daily data row will be fetched at the beginning of the day, instead of the end of day. To guard against the bias, you need to use the post_func function.

That is true, Quandl data is usually daily at best, an earnings report will have the same timestamp regardless of if it was released before or after trading hours. If anybody knows a data source with times and dates for things like earnings reports I would like to try integrating it to be used with Quandl data. It would have to be free and preferably a site with an export feature so the data doesn't need to be scraped from HTML.

Perhaps Quandl can add this data to their DB. This will be more convenient than getting it from some other place.

Does anyone have an example of a post_func function to adjust fetcher data by one day, to correct for the lookahead bias?

Hi Richard,

I think you can do something like this as a post_func:

def time_lag(df):
# using pandas' tshift to lag the index by 1 business day
df = df.tshift(1, freq='b')
return df

Let me know if that is too simple for your use case though. --Jess

Hi Jess,
That does the trick. Simple is good.

Hi, David

I want to call several Chinese stocks from Quandl using my strategy in the Backtest. Can you combine your post with my strategy to allow me to download multiple Chinese stocks and test the performance? Thanks!

The stocks I want to fetch into the algo are :
000023.SZ
000020.SZ
000019.SZ
000018.SZ
000014.SZ
000011.SZ
000010.SZ
000008.SZ
All these stocks have data set in Quandl. Thanks!

line 80 = QuandlRequest() <---- is this the line where we insert our api authentication key...?? in the enclose bracket?? if no where do we insert in your sample.

@QIFENG WU

ValueError: The truth value of a DataFrame is ambiguous. Use a.empty, a.bool(), a.item(), a.any() or a.all().
There was a runtime error on line 49.

Yes, line 80 is where you would insert your api token. I believe that quandl has changed their api since this was written so I'm not positive it still works. Hopefully it still does though.

David, this is exactly what I'm looking for but how do I access the version of Quandl's python library that you tweaked and get started using it?