wednesday - rinse and repeat?, page-11

Good find. 2009 threw things out a bit (-8.6%, 24.5%) for 6th sig error.

Skipping most volatile Januarys (3 highest + 3 lowest, which includes last year), can get an ok linear fit: R improves from 0.55 → 0.67 with

Yr Change = Jan Change*3.1 + 5.3

Yet to do full stats (by sight std dev looks like around 10%) so from Jan being 3% so far, get year at +14% (σ: 10%)