fix: calculate meaningful price change properly weighting each pair

closes freqtrade#12588

Author: xmatthias

freqtrade/data/metrics.py

@@ -74,9 +74,10 @@ def combined_dataframes_with_rel_mean(
     df_comb = combine_dataframes_by_column(data, column)
     # Trim dataframes to the given timeframe
     df_comb = df_comb.iloc[(df_comb.index >= fromdt) & (df_comb.index < todt)]
+    rel_mean = df_comb.pct_change().mean(axis=1).fillna(0).cumsum()
     df_comb["count"] = df_comb.count(axis=1)
     df_comb["mean"] = df_comb.mean(axis=1)
-    df_comb["rel_mean"] = df_comb["mean"].pct_change().fillna(0).cumsum()
+    df_comb["rel_mean"] = rel_mean
     return df_comb[["mean", "rel_mean", "count"]]