-
Notifications
You must be signed in to change notification settings - Fork 3.9k
Add SMMA, replace RSI with original study #925
Conversation
Thanks a lot! Great stuff, let me know when you have both parts ready and I'll go over them. |
Part 2/2 of #759 RSI has been replaced with the original study. Once again, tests are included and verified. Keep in mind that this I'd suggest doing a minor increment when implementing this, for this reason. I assumed you'd want both parts together, so I put them together in this PR. |
This PR looks great, thanks a lot for the time you put in :)
What exactly will this break besides calculating different results? |
This is exactly it. I used the term break just to clarify that this change isn't a simple fix. I should've worded it differently, you're right. Users of Gekko are probably used to working with the current results and I can imagine they wouldn't be too happy if we lower their revenue by implementing this change. Assuming you're using Semver, if this version were to be 1.0.0 or higher, this change would classify as a semver-major - because we will change users results. |
Great, I will make it part of the release described in #911 and merge it into
Semver describes versioning according to API changes, but since Gekko is not a library that exposes a single API I found semver hard to follow:
For that reason I found it impossible to follow strict semver since all the above can be argued to be "APIs", and a lot of them have had breaking changes in the past. Also I am not comfortable labeling this v1+. |
Part 1/2 of #759
I'm going to split the pull request as this one is backwards compatible. The next part is not - it will replace the current RSI function with the one from the original study.
This means the patch version can be incremented for this specific PR, instead of a semver-minor.
Tests are included, however, it looked like the test suite is broken for this repository.
Anyway, I just added them in the same style as the tests for the other indicators.
The pre-calculated results are taken from my own repository, which is documented and tested thoroughly. Sources used to determine the correct algorithm are documented, as well.