538 has an interactive model now. Some interesting PA & FL stats

[Statistical]

8. One of the dangers in 2016 was people treating states like a coin flip. They aren't.

They are not independent events (coin flips or dice rolls) they are semi-related. If Trump wins PA it means the polls were off in PA and if they are off in PA they likely are off by similar amounts in states with similar geography and demographics like MI and WI.

The 538 model takes that into account. It predicts if Trump wins PA that Biden's chance of winning MI and WI aren't 93% & 88% anymore they drop to 56% and 59%. If Biden loses PA the odds of him picking up both WI & MI are only about 1 in 3. That isn't impossible but far from a certain thing.

Likewise Trump winning both FL and PA likely means he is performing better than the polls predict across the country that means the odds of Biden winning any close states say less than +4 in polling drop significantly. States like NC, GA, and AZ go from even odds to long shots and Biden would need multiple longshots which is an even longer shot.

In 2016 other sites showing a >99.9% chance of Clinton winning were based on flawed analysis of treating each state as an independent coin flip. Trump was the underdog in a dozen battleground states so the the odds Trump would win six or seven of them were incredibly low or they would be if states were independent coin flips. The reality is once Trump outperformed a single underdog state the odds of him doing well in other states rose dramatically and those sites (not 538) didn't take that into account.

In the 538 model they show the revised chance for each state when you mouse over them after picking a state.