>> Mathematically impossible?
what 3×3 meant to say – “statistically impossible”
the reason for this misconception – you can come up with 1000s of different patterns which counties are important – since 1980, or maybe since 1890, or 1089 – and then focus on the pattern where current election is “statistically impossible”
a better way is to fix prediction rules before the election and then check them after.
another way is to construct multiple predictions and then use machine learning and cluster analysis to see if this one is different from others