what he's doing is taking national and regional polls and trying to cram the results into local ridings. he may look to past results in specific regions, but that's just going to confuse the outcome, if you try to separate it from context - sometimes, when you overcomplicate, you just confuse things. and, then you've got rounding - so he adds an extra 3.5 ndp seats as a consequence of rounding error. it's bizarre.
so, i look at this and i see 2.6 seats for the ndp in the east and the initial question i ask is "where". and, if your model throws out 2.6 seats (are you rounding up or flooring on that? or threatening to cut the seat in half and giving it to the one that objects?), and you can't find where they actually are, you need to pull back from the model, right?
well, i took a look through the map, and the only seat he's calling for the ndp is the one in st john's, newfoundland. so, his map contradicts his model. so much for mr. fournier.
he would appear to actually be calling for liberals 24, conservatives 7, ndp 1. if the ndp are lucky - that seat is listed as a toss-up.
likewise, i can only find one seat for the ndp in quebec, and it's in montreal, and isn't outremont. crucially, he has the bloc listed second, probably because they did well in that riding in the duceppe years. but, an urban riding that voted for duceppe isn't going to find blanchet nearly as appealing - this is a statistical relic, and all you're doing by bringing in duceppe numbers is confusing the reality of it.
that's probably true for the bloc across the province. blanchet may appeal to the nationalist core of the old bloc, but he's lost the social democratic bent that gave them their actual electoral victories. if they stick with the ndp - and vote qs provincially - that's going to leave the bloc uncompetitive in montreal. and, that is indeed likely the difference between the 30% they're at and the 40%+ they ran at under duceppe. but, how many seats do qs have, in a riding system that is more favourable to them?
if the ndp hang on, it will indeed likely be on the island.
but, i wouldn't count on it...
likewise, he has them winning three toss-up seats in toronto (despite local polling having the liberals running at 70% in the 416), holding three very close ridings in hamilton, holding a toss-up in london, and holding a toss-up in windsor, along with two more likely ridings here. so, that's ten seats, if you add them up - but at least seven of them are a coin toss. he also has them winning five seats in northern ontario based on the strength of ndp results ten or fifteen years ago, when all of the polling i'm aware of has singh doing exceedingly poorly with rural voters, for obvious reasons. and, three of them are toss ups.
so, that's fifteen seats - but ten of them are a coin toss, and that coin toss relies on the premise that people look at jagmeet singh and see jack layton, a pretty sketchy premise, given the actual data in front of us.
i penciled them in at five, and i was being conservative when i did.
i counted five liberal seats in the prairies (4 in wininipeg + ralph), and 2 ndp seats (niki ashton + one in saskatoon). so, why does he have the numbers he does? where? show me. his model is going to overshoot conservative numbers in winnipeg, on the strength of the conservative rural vote, and this actually happens every election, and the pollsters just don't fucking learn. so, when i said 5-7 in winnipeg, i was undoing the aggregates on purpose. but, still - he needs his numbers to accurately reflect his actual map, or he's just demonstrating that his model is flawed.
we agree that alberta will probably go entirely conservative, and by large margins. except he still has the ghost of linda duncan holding on...
i can count 13 ndp seats and 9 liberal seats on the map - numbers that are at least consistent, and that i have less disagreement about, even if it's by accident.
so, on his map, he has the ndp winning 32 seats, including ten very tentative wins in ontario that rely on singh getting the layton vote out. i had them at 20, give or take a reasonable error in bc. what i need to do to put the numbers in line is take that layton phantom effect out of the numbers - something i'm more than willing to do - and round them back down out east.
so, singh will need a divine intervention to get to 37 - and maybe he's into that, but i'm not. my numbers do not attempt to quantify the effect that jack layton's ghost will have on current ndp voters, many of whom were toddlers at the turn of the century, and are consequently more realistic...
again: the difference is those ten toss-up votes in ontario, and another handful in quebec and out east, which there's not any convincing evidence for - it's a relic of an over-reaching model. essentially all of the 15-17 seats should be liberal at the end of the day. the exaggerated bloc seat totals is also a relic of duceppe that is neither justified by the current political alliances nor by the actual numbers. that takes the ndp down to 20, the bloc down to 30 and the liberals all the way up to 161 - and is all in the east.
what's left is the question of the 905 and urban ontario - will the conservatives get 37 seats in ontario as the model suggests or 6 as i trollingly lowballed them at? the fact is that the 905 polls have them well ahead, meaning they're poised to not just hold but pick up. and, we'll see if the greens can split or not, but, i acknowledge the sketchiness of this in my write-up, and i'm less confident in really pushing it down.
as has become the norm, this election will be won or lost in the 905. and, if we split the difference between 37 and 7, you still have the liberals with 176 seats.
i tried to avoid this until the last minute, but there you go.