ok I did the 2008-18 data. -14.5% from new method vs -8.8% from the old.
so overall old method wins. new method basically gave more importance to the recent past. that's proving bad. long-term past wins.
well this was a fun project, learned a bit about programming. no harm done