Mentioned this last night but the weightings they are using seem way off to me. This tweet shows it well.
A 65% youth turnout last time dropping to 52% this time while the pensioner vote increases 10-15% seems off to me. Could be significant as those two age groups are the ones who largely support one party, the youth vote going to Labour and the pensioner vote to Conservative.
Given how tight some of the marginals are, a 1% swing can equate to 10 seats or even in some circumstances as many as 20 seats, either or both of those turnout predictions being wrong and the actual turnout being nearer the level of the last election could make a huge difference.