what is my record on recent election predictions?

- my prediction in 2015 was a large liberal minority - that they would almost get a majority, but not quite. this was considerably better than any of the aggregate sites, which i criticized very heavily. the reason i was off was that i traced the ndp collapse to an increase in bloc support, and it ultimately ended up helping the liberals more than the bloc. with four way splits and small sample sizes, the polling in quebec was vague and messy, and nobody saw what happened coming; i at least got the right idea. the reason i was able to make a very good prediction here is that it was a long election and that there was a lot of data. my writing exists on this page, but the basic takeaway is to take aggregate sites with a grain of salt because they utilize too much flawed data; it's largely garbage in, garbage out.

- after the field stabilized a little, it became clear to me that trump would win because he was the most moderate candidate in the republican field, and i do believe that this is the truth of the matter, regardless of the narrative around it. had the party picked a moderate, they could have beat him; instead, people ended up voting for trump because he just wasn't nearly as bad as rubio or cruz.

- i paid more attention to the democratic side, and found myself having to learn the demographics of a foreign country in order to keep up. i figured out very early on that the polling companies were skewing the data on purpose, and i made some good predictions - i predicted sanders winning michigan, for example. i also predicted clinton winning kentucky, which surprised a lot of people. when i made errors, and i did, it was mostly due to a misunderstanding of those demographics, or a lack of clarity as to the process. for example, i assumed that the large black population of dc would transfer over to northern virginia, and it did not - stuff like that happened over and over. i also come face to face with the reality of voter suppression in the democratic primary system, and the difficulty in trying to analyse polling in a system that isn't actually fair. i stand by the focus of my analysis, and by my critique of the media narrative around race as such a defining issue in a race between two older white folks. i largely corrected myself as i went through, and would defer to my own writing. in the end, my record was mixed but at least as good as anybody else's, even if the larger takeaway was a learning experience about the process.

- my prediction in the 2016 election - and this is still sitting in those files - was that the shadow government would rig the election for trump. that doesn't sound very scientific, granted, but it's what allowed me to avoid the error that everybody else made. i guess that silver came closest with his gigo-model, but he was simply less wrong. the fact is that the data clearly projected a clinton victory, and i realized that, but i also realized the election was rigged. so, when the narrative switched to the russians rigging the election, i just rolled my eyes - it was an inside job, and i saw it coming. there's two ways to talk about this, after the fact. the first is that you can reference vague, shadowy bodies without any real evidence, which is both obvious and sketchy at the same time; call it a vast, right-wing conspiracy if you must, but realize that it is the obvious truth. the second is to point to more concrete concerns about voter suppression, the civil rights act, etc.

- my analysis of the 2018 ontario conservative leadership convention was nearly spotless, if cynical and distant.

-  the 2018 ontario election might seem like a spot on my record at first glance, but i disagree with that. there are two things i tried to make clear, here. the first is that the polling was really not detailed enough to be predictive - we had tons of online research, but almost no actual polling, and the little bit of polling we had was not focused enough on regional variation. the gigo models actually drastically underestimated conservative numbers, so to suggest they were accurate is to hit a very large target - the conservative numbers were mostly well outside the margin of error in the polling. so, i tried to make that point clear - i didn't have enough data to work with. the second point i tried to make is that ford was repeating all of the same warning signals we heard from trump around the fairness of the election, specifically these projections that the vote isn't fair. i also became suspicious of these online panels, and what they were really up to. i stopped short of explicitly predicting a stolen election and instead suggested an ndp minority is most likely (with very low confidence due to a deficit of data), but i am extremely skeptical of the numbers i saw come in and have called for an investigation. so, i don't have a lot of confidence in my analysis of this election, but i don't trust the results, either; and, if some process of stuffing ballots is shown to be true in the end, i think my analysis gets upheld.

- i did not pay attention to the 2018 midterms in the united states, but pointed out the importance of the democrats finding a way to win white voters.

i don't expect to be as interested in either of the upcoming cycles, but how well i do in any of these elections will depend on whether i can get enough data to identify flaws in the gigo models and whether or not the elections are fair in the first place.