The current situation with multiple tactical voting sites giving differing recommendations is not new. This site was the most accurate in the 2017 general election, getting 95.4% of our tactical voting recommendations correct.
|Site||Total rec.||Accurate rec.||Errors||Accuracy|
(known in 2017 as stopthetori.es)
|Best for Britain||508||472||36||92.9%|
After the election it is easily possible to see whether a tactical voting recommendation was 'right' or 'wrong'. Just ask: did the party this site recommended either win the seat, or at least come in second place? If so, it was a correct recommendation. If not (ie. third place or lower), then it was no good as a tactical voting recommendation.
The different 'total recommendation' figures reflect that sites varied on how many seats there were where they did not publish a recommendation, or stated 'no recommendation'. For example, in 2019, tactical.vote is currently (25 Nov) publishing a recommendation for 586 seats out of the total 650.
Generally we were correct that the 2015 general election was the best guide to 2017 tactical voting decisions, despite many assertions that the 2016 EU referendum had changed the whole electoral map.
Despite an often-repeated view online that our method 'overestimates Labour' or 'underestimates Lib Dems', the reality is that in 20 out of our 26 incorrect recommendations in 2017, we recommended a Lib Dem vote when it turns out that Labour was the correct tactical vote.
We did underestimate the East Devon independent candidate, who we are recommending in 2019. Sadly our recommendation to vote Green on the Isle of Wight was also incorrect.
You can download a spreadsheet of these results to look through for yourself.
Thanks to Tactical2017 for initially compiling some of this data. You can read their assessment of 2017 here.