Predictions are judged as predictions, not statements of fact. In general, possible outcomes are modeled on a bell curve. Mathematically, if 97-50-3 has some probability of occurring you simply cannot label his prediction "wrong
Hate to interrupt the conversation, but lay off Polnut and recall that sampling bias is a thing. The rest of that post is skeptic claptrap I had thought Fivethirtyeight had quashed well...
First of all, I am not particularly impressed with your opinion of the merits or demerits of some unstated argument by some sports prognosticator. If you think his arguments actually have some merit by all means share them.
It is not "skeptic claptrap" that another mathematical model of the British Columbia election showed a BC Liberal popular vote and seat majorities being something on the order of three- and two-sigma events respectively with the actual comfortable margins being achieved even more highly improbable. Is it really "skeptic claptrap" to point out this objective fact because a four-sigma event just happened by chance, or was it the underlying model itself that was the "claptrap?"
Blaming "sampling bias" makes no sense. It is akin to arguing, "My model wouldn't have garbage out if it didn't have garbage in." It would seem immediately obvious that if you want to model the variability of an election result that projection would model the variability of the sampling. Obviously, some hand-waving occurred. Turns out that the simplifying assumptions resulted in the model being simplistic. That is just the objective fact, unless you want to argue sometimes lightening stricks the same person twice in one year.
Another assumption in such models is that what happens during the balance of the election doesn't matter. Apparently, prognosticators can't model last-minute doubts about the Labor defeating John Majors, or the BC NDP returning to power after their previous stint, so they assume such events don't occur. Again, their models are counter-factual.
The person who has to "lay off" someone is Polnut. His formulation has every indication of him speaking ex cathedra when he is simply in no position to do so whatsoever.