use proper data

• fixes script
• uses ECMWF benchmark, observations in terciles from renku dataset
• works locally
• script runs quite fast. could precompute benchmark as benchmark is always the same, maybe future PR

ready to merge @tasko.olevski

README.md

@@ -6,11 +6,11 @@ Image used to calculate the scores for the S2S AI Competition
 
 ```bash
 pipenv install
-pipenv run python scoring/scoring_script.py submission_rb_prediction_2020.nc
+pipenv run python scoring/scoring_script.py ML_prediction_2020.nc
 ```
 
 Alternatively you can use the docker image itself
 ```bash
 docker build -t scoring-image .
-docker run -ti --rm -v $PWD:/work scoring-image pipenv run python scoring/scoring_script.py /work/submission_rb_prediction_2020.nc
-```
+docker run -ti --rm -v $PWD:/work scoring-image pipenv run python scoring/scoring_script.py /work/ML_prediction_2020.nc
+```
\ No newline at end of file