Difference of Feature Sets

Hi, can someone help explain the feature_metadata json file? I was expecting the small feature set to be a subset of the medium, but it doesn’t appear so. If using the full set of data, will there be redundant overlap among different columns that I should try to take care of?

Also, I find these stats in the feature_metadata. Is there a definition somewhere that lists these out and what they are exactly?
{‘legacy_uniqueness’: 0.3646533080939621,
‘spearman_corr_w_target_nomi_20_mean’: -0.011053228109048541,
‘spearman_corr_w_target_nomi_20_sharpe’: -0.26929474693604694,
‘spearman_corr_w_target_nomi_20_reversals’: 0.0008786204121464577,
‘spearman_corr_w_target_nomi_20_autocorr’: -0.014497012511186493,
‘spearman_corr_w_target_nomi_20_arl’: 3.984732824427481}


Hey man,

Here’s the post explaining all your questions:
October 2021 Updates - Announcements - Numerai Forum

Thank you! I’ll check these forums more rigorously for answers in the future