Backtesting enhancements and fixes are live 🥳
almost 3 years ago by Ellery Berk
Myst Platform release (2022-02-09)
Hello Myst Platform users!
We have released an updated version of the myst-alpha Python library. Please upgrade your myst-alpha package to the latest version at your earliest convenience. Upgrading instructions can be found here.
Updates in this release:
- ⚡️ ENHANCEMENT Some backtests were slow to complete because result aggregation and metric calculation lingered in an incomplete state. Now, you should observe that backtests complete in approximately one hour, consistently.
- 🐞 BUG FIX If you have not deployed your model before running your backtest, you now will receive the following error message: “Model with identifier
{model UUID}is not deployed.” This new message is intended to clarify that, for now, you must deploy your model before running a backtest.
Backtesting tips
- We recognize that deploying your model before running a backtest is not intuitive – and that switching from the client library to the web UI to deploy is clunky. We are working on improvements to this experience.
- If you would like to avoid proliferating model fits and forecasts when you deploy before running a backtest, you can uncheck the “Active” box in your policies.
- When you are ready to generate recurring forecasts, please check the “Active” box and re-deploy!
As always, please do not hesitate to reach out via the chat embedded in the Myst Platform UI or email us at [email protected] with questions or feedback.
We appreciate your partnership!
Ellery and the Myst team