THE BLOG

# The Statistics Prove It: We've Got The Groundhog Day Rule Backwards

02/02/2016 08:24 EST | Updated 02/02/2017 11:06 EST

Every now and then, some statistical research comes along that shifts paradigms, rewrites perceptions and sets the stage for a new era in human thought.

This is not one of those times.

No, this is about apparently long-ignored research on the value of a rodent in determining medium-term weather forecasts.

As it turns out, groundhogs are shockingly poor predictors of the weather -- so shockingly poor, in fact, that if we bet against them, we'd come out on top.

The groundhog rule states that if a groundhog sees its shadow, winter will linger ("six more weeks of winter"), but if the groundhog sees no shadow, it'll be an early spring.

But statistical data for 13 Canadian cities shows that, if you used cloudiness on Feb. 2 as an indicator of an early spring, you'd only be right 37 per cent of the time.

If you just flipped a coin to forecast an early or late spring, you'd expect to be right 50 per cent of the time in the long run. Groundhogs are worse than flipping a coin at determining the weather.

But the groundhogs would be right much more often if we just reversed the rule. If a sunny day meant an early spring, then that 37 per cent accuracy score would instantly become a 63 per cent accuracy score -- much better.

Those accuracy rates -- 37 per cent and 63 per cent -- aren't very meaningful when the sample size is this small. Either way, the groundhogs aren't reliable indicators of anything.

But in some parts of Canada, the reversed groundhog rule would actually take on some statistical significance.

Take Edmonton, where the groundhog rule is right only 26 per cent of the time, according to data cited in this Canadian Encyclopedia article.

Reverse that rule, and you have a predictor of Edmonton spring that's accurate 74 per cent of the time. That might beat the city's local forecasts.

In Toronto, the groundhog rule is right 29 per cent of the time. So the reverse groundhog rule would be right 71 per cent of the time. Those sound like decent odds.

So let it henceforth be known throughout the land, that groundhogs can mostly be taken at their word on the weather -- so long as we always assume they're lying.

Those pesky rodents.

Odds of a groundhog being right, under traditional rules. The first column indicates likelihood of a groundhog seeing its shadow, in per cent. The second column indicates how often a groundhog would be correct, in per cent. Source: The Canadian Encyclopedia.