I've cobbled together two models for the upcoming election, and hope to add to that over the next two years. The first one is a province-wide poll tracker, where I'll be taking the various opinion polls and estimating the trend in voting preference, as well as estimating the actual voting preference and confidence in that preference. This takes into account the date and length of time between each poll taken, as well as the sample sizes of each poll.
Previous update December 7 2016: As you can see, there was quite a bit of volatility in November, with the PCs and Wildrose effectively switching positions in two polls. Two subsequent polls at the end of November and early December seem to have reverted back to previous positions, so we'll see if this volatility quiets down a bit.
The second model I've put together is a seat projection. This takes the poll numbers, and estimates support in each riding based on patterns from previous elections. When possible, polls that breakdown support by regions are used as they've proven to be more accurate. The darker coloured bars indicate the 50% likely range of seats, and the lighter coloured bars indicate the 95% range of seats.