well, i was by far the closest of any of the predictions i've seen. what i got wrong was that the split in quebec helped the liberals in places nobody would have expected them to win. but, the bloc vote is still coming in, and it could still get a bit closer.

but i figured it would be a seat or two short of a majority, and it does seem like they managed one, even if it does get a little closer to where i was predicting.

i'll need to look at specific data more specifically in the morning, but it's actually aligned almost exactly to my predictions, except that swing in quebec.

nobody got quebec right. everybody else suggested the ndp holding 35-40 seats. give me credit for realizing that that didn't make sense, but i seem to have jumped on a last-minute bloc bump that was really just statistical noise [although it was present in the data, it's always a coin flip whether you're picking up a bump or not over the last 36 hours - it's equally likely to be noise, and appears to have been this time].

the liberals seem to have picked up around 35% of the vote in quebec. nobody had them that high.

so, i can make a tentative guess as to what went wrong. i suggested that an ndp--->liberal swing would not be enough for the liberals to win rural quebec, and that the bloc would come up the middle. this would have been a decent guess if the numbers would have stabilized at libs 30, ndp 25, bloc 25.

what happened, instead, was that the swing to the liberals was larger than any pollster picked up. we ended up with libs 35, ndp 25, bloc 20. that meant the swing was actually enough for the liberals to win, after all.

i got everything else dead on, almost.

i suggested the conservatives would be around 100.

i suggested the ndp would be around 45.

but, i suggested the bloc would be around 35 (they got 10) and that the liberals would be around 160 (they got 185).

i did far better than the official pollsters, and by citing precedent and using logic rather than using modelling (which i was certainly right in suggesting was inapplicable to this election). but, you have to understand that quebec was impossible to predict. it wasn't even sure what order the parties were in. and, with margins of error over 5%, you couldn't really peg them better than an unordered 10 point spread - which meant you were literally left making a guess.

that said: if there were some numbers putting the liberals higher than 30, consistently, i think i would have seen this coming. and, i suspect we'll find out that turnout in quebec was low, which propped the numbers up a bit. i did see that coming at some point, but i didn't factor it in. and, even so, i would have had a hard time suggesting liberals could win in a lot of the seats they did.

so, i got most of it dead on. quebec was completely unpredictable; i broke with the conventional wisdom, but what actually happened was even more unlikely than my break with the conventional thinking. although, i think we'll find out in the end that turnout was very low, and it's really what caused these unexpectedly high liberal numbers there.