**Probabilistic Forecasting Planning**

**Attendees:**

**Agenda:**
* introductions
* @Svea gives overview of refactor state and next steps
* all to discuss work items, ways of working (e.g., joint stand-ups)
* agree next steps

**Refactor Idea**
* Instead of computing predicition intervals within the predict function we will separate stand-alone functions to predict intervals and to predict quantiles.
* Currently base class refactor is (hopefully almost done)

**TODOs**

sooner:
* Refactor Inheriting Classes
* List of Inheriting Classes that have probabilistic capabilities
* based on tag (maybe not always correct)
* tell by looking at _pred() fucntion
* testing predict() with pred_int = true
* use the test `test_all_forecasters/test_predict_pred_interval` (original or modify to "catch" the estimators)

* Refactoring guideline
* Must: implement `predict_quantiles()`
* Must remove pred_int argument from `_predict()` function
* Maybe: implement `predict_interval()` (if defaulting to predict_quantiles is inefficient)
* Update Extension Guideline
* Upadte Tutorial
* Impelement prediction intervals for forecasters that don't support probabilistic forecasting yet (e.g. Naive forecaster)
* Improve tests (so far only test stabs)

later:
* Add model agnostic default strategy in base class
* Think of how user can decide which stratgy to use

todo:
tests (test all forecasters, maybe also look at Svea's stab)
check which forecaster needs refactor