finalize

intermediate_leaderboard.py

@@ -55,8 +55,9 @@ def make_intermediate_lb():
     df_latest_raw = df_latest_raw.query("Endpoint == 'Average'")
     df_latest_raw['latest_time_per_user'] = df_latest_raw.groupby('user')['submission_time'].transform('max')
     latest_submissions_df = df_latest_raw[df_latest_raw['submission_time'] == df_latest_raw['latest_time_per_user']].copy()
+
     latest_submissions_df = latest_submissions_df.sort_values(
-        ['RAE','user', 'Sample'], ascending=True
+        ['RAE'], ascending=True
     ).reset_index(drop=True)
 
     # Get the unique users in the order of their first appearance
@@ -81,7 +82,7 @@ def make_intermediate_lb():
     tukey = pairwise_tukeyhsd(endog=latest_submissions_df['RAE'], groups=latest_submissions_df['user'], alpha=0.05)
     tukey_df = pd.DataFrame(data=tukey._results_table.data[1:], 
                             columns=tukey._results_table.data[0])
-    
+
     # add CLDs
     cld_dict = cld(tukey_df)
 
@@ -114,6 +115,9 @@ def make_intermediate_lb():
         )
         cld_df = cld_df.merge(metric_stats[['user', f'{metric}_mean', f'{metric}_std', f'{metric}_display']], on='user', how='left')
 
+    # re-sort by RAE mean, lowest is best
+    cld_df = cld_df.sort_values(by='RAE_mean', ascending=True).reset_index(drop=True)
+
 
     cld_subset = cld_df[['user_fixed', 'fixed_letter'] + [f'{metric}_display' for metric in METRICS]]
     cld_subset = cld_subset.rename(columns={'user_fixed': 'user', 'fixed_letter': 'CLD'})