If the model were calibrated only on presidential elections since 2000, Clinton's chances would rise to 95 percent and Trump's fall to 5 percent.

This is a counterfactual about Silver's own model, so the record is FiveThirtyEight's published 2016 analysis or methodology. True: a FiveThirtyEight publication from 2016 states that calibrating the model only on presidential elections since 2000 would give Clinton about 95% and Trump about 5%. Mostly true: the published figures differ slightly, e.g. 94–96%. Mixed: the published figures differ substantially, or the counterfactual is described differently, e.g. using a different period or calibration. Mostly false: the published analysis shows a materially different counterfactual figure. False: FiveThirtyEight's published analysis contradicts the claim.

Set before the research began.

1. E1 Nate Silver’s Forecast (Probably) Overstates Trump’s Chances New York Magazine (Intelligencer) · 7 November 2016

   “Silver’s model looks at the race’s volatility over the past three months — and the relatively large numbers of third-party and undecided voters — and discounts the security of Clinton’s longtime polling advantage.”

The only evidence, E2, is a New York Magazine piece saying Silver's model gives weight to the race's volatility and to the large number of third-party and undecided voters. It does not mention calibrating the model only on presidential elections since 2000. It also gives no counterfactual probabilities for Clinton or Trump. No FiveThirtyEight publication was provided, so the claimed 95%/5% figures cannot be checked against the record the criteria require.