Hi all,
We are currently in the process of rolling out an update to our fundamentals database. As with any data source, there are always issues of data cleanliness and accuracy with our fundamental data. With the help of the community, we identified and fixed many problems with our fundamental data. We certainly haven’t fixed all the problems, but we have tackled a bunch of the big ones.
Some highlights of these fixes:
- More accurate point-in-time data for daily metrics. When we generate point-in-time data for the time period prior to our subscription to the fundamental data, sometimes we need to estimate when that data would actually have been available to traders. For example, Q1 earnings aren’t available to the market immediately after Q1. There is a delay between the end of the quarter and the filing. Sometimes we have the filing date for a metric. Sometimes we don’t. When estimating the availability date for metrics that update daily (like market cap, enterprise value and others), we were being overly conservative and estimating a delay that was too long. We’ve fixed this and these updates are available to trading algorithms the following day. This affects the timeframe of 2002 - May, 2014.
- Friday updates. Similarly, for the time period of 2002 - 2014, we were not processing updates for Fridays. The previous value was forward filling for use in backtests. We’ve remedied this.
- Further, there were other time miscellaneous time periods for which we failed to process updates over the last year and half, since our subscription to the fundamental data started. These windows were remedied as well.
The result is that our database now has improved data quality.
We want to make sure you are aware of this update because it could change your backtest and paper trading results if your algorithm uses the get_fundamentals() method. We have proactively reached out and notified those community members with live algorithms affected by this change.
But we encourage you to run fresh backtests on your algorithms with get_fundamentals() to assess the impact on your algorithms.
TL;DR: We fixed data bugs with fundamentals data. You might want to re-run your backtests or re-examine live trading results tomorrow if they use get_fundamentals()
All the best,
Josh