i was tired last night, i'd been up since before sunrise and had done a lot of walking, so let's try to get this across again.

i got the west almost perfect - i claimed ndp-15, liberals-15, greens-4, conservatives-rest. if jwr joins the greens, i almost nailed it.

i argued the models were overselling the bloc, who would get closer to 30 than 40. they got 32. i argued that they'd be shut out of montreal; they were. i claimed the liberals were being lowballed and they would win the province with about 40 seats; they ended up with 35. i argued that the conservatives would get 13 seats in the east (including quebec); they got 14.

i argued that the ndp were experiencing a bradley effect in the polling and would win roughly 5 seats east of winnipeg, all in ontario; they got 8, 6 in ontario. i pointed out the possibility of the greens causing chaos in the maritimes, and they did in fact score an unexpected upset win. i had the liberals at 27 seats in the maritimes; they got 26.

i argued that the liberals would sweep toronto and ottawa and pick up seats in the 905 as well as in smaller urban centres like windsor, and they did.

but, i argued that the conservatives would experience a massive vote split in rural ontario due to everybody moving to the greens, and that did not happen - and that is where i was substantively wrong, and what cost me my prediction. but, i should make the point clearer - it's less that i was wrong and more that the polling was wrong. i can only rely on the data, which had the greens in double digits in rural ontario; that simply didn't happen. if the polling had the greens where they ended up, i would have fallen in line with predictions of a liberal minority.

the thing is that it made sense to me to have everybody move to the greens - racist new democrats (and i don't like singh, but for different reasons than them), scared conservative property owners & pissed off liberals, alike. it was a perfect storm. and the numbers looked like 2000 - i will defend the argument.

but, it didn't happen - the greens got 6% when the polling had them at 16%. so, blame them...

that said, i acknowledged that it seemed like a stretch; i saw the data, but realized it was kind of unusual, too. so, i corrected myself - i claimed the following numbers in ontario were more likely, roughly, and gave myself some room for error on top of it:

liberals - 95
conservatives - 21
ndp - 5

but, i'm still off. why?

i still overshot the effect.

i was looking very carefully at the polling, which had the liberals roughly flat (they finished down about 3 points, which is roughly flat), the conservatives down by 8-10 points (they finished down about 3, which is actually roughly flat), the ndp roughly flat (they were. literally.) and the greens up by like 10-12 points (they finished up by around 3). so, in working this through my bullshit detector, i reasoned that, even if the greens only took 5% instead of 10% from the conservatives, it would still be a good bite, so long as it was localized in rural ridings  - and still enough to let the liberals squeeze through, so long as they polled roughly flat. so, i was essentially arguing that the greens and conservatives would split the vote, and let the liberals up the middle; if it was really bad, it could wipe the conservatives out, but if it was a more moderate split, it would just knock them down on their ass - and leave them with 20+ seats.

but, it just evaporated entirely; it didn't happen at all. the greens were up by a few points, but it didn't pool anywhere, and they appear to have had little effect, whatsoever.

what did happen in the rural areas was a movement from the ndp to the conservatives, which allowed them to make up for their losses in the 905 by stealing a few seats from the liberals and the ndp that they may have a hard time keeping in the long run.

if the greens ever get their break in ontario, my projection for this election may be prophetic - it's what would happen if the greens were to actually get that kind of bump. the results in the east are instructive. and, the polling presented the signal, too. but, it didn't happen this time...

what i actually should have done, the error i actually made, was that i should have given myself a larger error bar - i should have talked about 30 seats instead of 15.

i'm going to make a series of posts before i move on.