Legend is not accurate in plot_predictions()

[Bisaloo]

example1 <- scoringutils::continuous_example_data

scoringutils::plot_predictions(example1, x = "value_date",
                               filter_truth = list('value_date <= "2020-06-22"',
                                                   'value_date > "2020-05-01"'),
                               filter_forecasts = list("model == 'SIRCOVID'",
                                                       'creation_date == "2020-06-22"'),
                               facet_formula = geography ~ value_desc)

^{Created on 2021-11-23 by the reprex package (v2.0.1.9000)}

The colours in the legend do not match precisely what is visible on the plot because the colours on the plot result from the overlap of various ribbons (with alpha = 0.4). The colours on the plot are effectively darker than what is specified on the legend.

A solution might be to remove alpha = 0.4? If the quantiles are ordered correctly and since we plot median & truth data afterwards, I can't think of anything that might be under the ribbons.

[nikosbosse]

👍 sounds good. Do you want to do that or shall I?