Live Score Computation Process
Computing the live score of each day is a very computationally intensive process.
Active datasets lookup
This board allows you to visualize which dataset is currently active and scored against live data.
Live Score computation
For each primary submission, the scorer will compute the spearman correlation score between your prediction for that target (30, 60, 90 days) and what happened live.
Mean by dataset
For each dataset, your live score for this dataset is the mean of your last two submissions for this dataset.
Mean of mean
The global mean score is the mean of all your live scores over the universe of strategy. If you miss a round you get a score of -5%.
Normalization
Since December '21, all round's scores are normalized to avoid some rounds having to much impact on the overall score: https://discord.com/channels/770586873260933120/770586873260933124/905097810502910033
Computing global leaderboard
Crunchers are ranked by their mean live spearman correlation score normalized overall strategies.
Internal Schema
Here is a schema on how the process is working internally:
