MRPs and mega-polls: every seat projection against ours

The projections, newest first

Seats as each firm published them (Great Britain, or the UK where the firm included Northern Ireland). Ours: our projection on the last day of the poll's fieldwork, recorded that day or rebuilt from the polls public by then (how the forecast moved). Sources: each firm's release where linked, and Wikipedia's table of seat projections.

How the final MRPs did

In 2024 every final MRP gave Labour more seats than it won, from 425 to 516 against 411. The British Polling Council put the MRP average at 441 Labour seats, 29 too many, and 101 Conservative, 20 too few: the same direction as the ordinary polls' miss, made bigger by the seat count.

2019: YouGov's final MRP gave the Conservatives 339 seats and Labour 231; they won 365 and 202. In 2017 its MRP called a hung parliament when most forecasts had a Conservative majority, with the Conservatives about 15 seats short of the 317 they won. MRPs can read the shape of an election well; their seat counts carry the error of the poll underneath.

2024 figures: each firm's own final release where linked, otherwise the seat tallies compiled by Peter Inglesby (the seats where each party leads, which can differ by a few from a firm's headline). Result: Labour 411 seats; the British Polling Council's review counts 412, with the Speaker.