Refinement for XIRR #2487
Merged
Refinement for XIRR #2487
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
added 2 commits
January 6, 2022 08:33
Fix PHPOffice#2469. The algorithm used for XIRR is known not to converge in some cases, some of which are because the value is legitimately unsolvable; for others, using a different guess might help. The algorithm uses continual guesses at a rate to hopefully converge on the solution. The code in Python package xirr (https://github.com/tarioch/xirr/) suggests a refinement when this rate falls below -1. Adopting this refinement solves the problem for the data in issue 2469 without any adverse effect on the existing tests. My thanks to @tarioch for that refinement. The data from 2469 is, of course, added to the test cases. The user also mentions that an initial guess equal to the actual result doesn't converge either. A test is also added to confirm that that case now works. The test cases are changed to run in the context of a spreadsheet rather than by direct calls to XIRR calculation routine. This revealed some data validation errors which are also cleaned up with this PR. This suggests that other financial tests might benefit from the same change; I will look into that.
From https://github.com/RayDeCampo/java-xirr/blob/master/src/test/java/org/decampo/xirr/XirrTest.java https://github.com/tarioch/xirr/blob/master/tests/test_math.py Note that there are some cases where the PHP tests do not converge, but the non-PHP tests do. I have confirmed in each of those cases that Excel does not converge, so the PhpSpreadsheet results are good, at least for now. The discrepancies are noted in comments in the test member.
|
Great work. For what it's worth, I re-ran my workflow on my dataset with the fixes and it worked a charm. Previously, I was getting 285 failed convergences out of a total of 7,177 calls to the XIRR method. Now, there are zero. Furthermore, these were masking an additional 15 failures caused by bad data on my part which I've now been able to fix. |
|
Wonderful, thank you for the confirmation. I hope to be able to merge this by the weekend. |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Fix #2469. The algorithm used for XIRR is known not to converge in some cases, some of which are because the value is legitimately unsolvable; for others, using a different guess might help.
The algorithm uses continual guesses at a rate to hopefully converge on the solution. The code in Python package xirr (https://github.com/tarioch/xirr/) suggests a refinement when this rate falls below -1. Adopting this refinement solves the problem for the data in issue 2469 without any adverse effect on the existing tests. My thanks to @tarioch for that refinement. The Python project cites https://github.com/RayDeCampo/java-xirr as a reference. Thanks also to @RayDeCampo.
The data from 2469 is, of course, added to the test cases. The user also mentions that an initial guess equal to the actual result doesn't converge either. A test is also added to confirm that that case now works.
The test cases are changed to run in the context of a spreadsheet rather than by direct calls to XIRR calculation routine. This revealed some data validation errors which are also cleaned up with this PR. This suggests that other financial tests might benefit from the same change; I will look into that.
This is:
Checklist:
Why this change is needed?