Tweet by @NateSilver538

Nate Silver @NateSilver538

They had Tories +33 seats when the final result was Tories +55. That's decent as these things go because this stuff is **really hard** to model. But certainly not fantastic. Also they got a bit lucky because they *underestimated* Labour's vote share.... — PolitiTweet.org

The 2017 YouGov model did better at forecasting seat level vote shares than a regional swing model *with known vote… https://t.co/ezzTubGLmS

Posted Nov. 27, 2019 Hibernated

Preceded By

Nate Silver @NateSilver538

So the news in YouGov's poll tonight is that it has Tories up 11 UK (fairly typical of recent polls). It's no surprise that Tories would expect to win lots of seats if that's the final margin. But MRP is not magic. It's an OK/reasonable technique but its powers are overstated. — PolitiTweet.org

Posted Nov. 27, 2019 Hibernated

Followed By

Nate Silver @NateSilver538

...meaning, if they had gotten the national margin exactly right (Conservatives +2.4 instead of +4), they would have shown an even narrower Tory seat margin and so would have been further off from the actual seat count. Luckily for them, their errors canceled out. — PolitiTweet.org

Posted Nov. 27, 2019 Hibernated