Instructions:

1. The scripts expect a data/ folder in the same directory that contains the JSON files provided by Metaculus
2. The plots comparing aggregation algorithms can be reproduced by running parametric_study.py
3. The table showing a correlation between frequency and magnitude of updates and accuracy can be reproduced by running atanasov.py
4. The plot correlating accuracy with description length can be reproduced by running nlp-exploration.py on the nlp-exploration branch