If you used Industry or Sector nodes in ranking system with technical functions they have not been working. The technical functions were returning N/A, which made all the stocks get the same score for that node.
For example the following Ranking System which you can see here, which scores the sector performance of the stocks only, scored all the stocks with the same rank since all the values were N/A. This is what it looks like
It is now producing the intended results. Stocks in the same sector get the same ranks and all the ranks are combined to produce a multifactor sector rank.
This problem is ONLY with Sector and Industry nodes of the Ranking System. We are still investigating how it broke, but it has been broken for a while. Most likely when we switched to FactSet few years ago which created many changes, and this was simply not tested.
These nodes are not commonly used, so overall most users will not notice. We will alert the users that use these nodes with a email.
Truly sorry about this.
What classifies as a “technical function”? Does Close()?
Yes, anything in the TECHNICAL section
How long had the issue been in effect? Trying to understand if I need to redo some prior analyses. Thanks!
We’re still investigating. We think it was when we switched to FactSet, so around three years
@marco sounds like it might be unrelated then, but I’ve seen another issue in the past 48 hours. On ranker adding slippage now gives nonsensical results. Wasn’t true before and screener with slippage is still working it seems. (I’ve also checked with different minimum prices and had the same type of results)
Looks like this fix Future prices now return NA if they do not exists broke something with slippage
Thank you for reporting it. Should be an easy fix
Hi Marco, appreciate the transparency!
Question: I was / am using this ranking system, where this formulas affected?
The “Ind” factors are a different animal and were not affected by this.
See the documentation for any of the “Ind” factors. They are calculated using using Aggregate with trimming and averaging. The time series for the industries are calculated using cap-weighting and daily returns. A completely different method.
Thanks Marco for the update!