Join GitHub today
GitHub is home to over 28 million developers working together to host and review code, manage projects, and build software together.Sign up
BUG/MAINT: Switch over to Google for benchmarking #1812
Due to recent changes in the Yahoo API, we're switching to Google Finance for benchmarking data to
We're using pandas-datareader to pull the data from Google Finance.
The one potential issue here is that the history that GF provides only goes back to around June 2001, and also data from
Jun 5, 2017
referenced this pull request
Jun 5, 2017
Another benchmark error that has to do again with missing index data for the user specified date range: Apparently, very recent benchmark data is also missing with the current google solution such as yesterday's prices. Trying to simulate a portfolio on June 6, 2017 with the Quantopian-Quandl bundle that i ingested and verified that it includes closing prices for June 5th, 2017. The simulation can NOT be run because the benchmark loader throws an error due to the benchmark data file not including June 5th, 2017. The benchmark SPY or specifically the ^GSPC csv file actually gets written or modified today, June 6 2017, but it only includes benchmark data up until June 2nd 2017 (<-3999 entry). I checked manually on the google finance website and the data for June 5, 2017 is available. If i do simulations not including yesterday's date for periods when i know the benchmark data is available from 2013 - 6-2-2017 all the simulation runs fine. As I mentioned in my earlier post if the benchmark modules can just fill in missing data this should not be an issue.
Here is the error:
During handling of the above exception, another exception occurred:
Traceback (most recent call last):
This was referenced
Jun 7, 2017
@freddiev4 Thank you! Changes made to fix the benchmark missing data and code error issues worked.
WARNING: Right now June 8, 2017 if one does a channel installation with: conda install -c Quantopian zipline=1.1.0 the following files crash the code:
Finally, i discovered if one makes the following modification it eliminates the benchmark date range error, if it is not the same as the user date range, and it no longer holds up the simulation.