There is a dividing line between the 4 major (seat winning) parties, the 3 medium sized but smaller parties, and the many other parties that ran in the election. Only one provincial party that was registered and eligible to run did not run, that was the NDP. The party was only registered earlier this year and, simply, was not ready for an election. It was a wise tactical decision to sit the election out rather than do poorly.
The table shows how ratios can be calculated, by dividing the popular vote share between this and last election. The ratios for this election (compared to last time) were as follows
PLQ - 1.33
PQ - 0.79
CAQ - 0.85
QS - 1.27
Apply this, for example, to SMSJ.
If you do you get 8,099 PQ votes, 7,353 PLQ votes, and 9,188 QS votes.
In the end, the PQ took 7,612 votes, the PLQ took 8,346 votes, and the QS took 8,437 votes.
This is part of the reason you need to do two things.
1 - use ratios and not raw swing. A raw swing would have projected 8292, 8247, and 7524.
2 - use your gut, your head, and history. The raw math will never be 100% correct, and it is up to you, the predictor and projectionist, to adjust where needed.
One thing I've learned is
2a - Adjusting a few ridings to match your gut is no big deal. Adjusting a number of ridings, all to keep the same party out; that's just the math trying to tell you something unexpected is going to happen.
Now that we have cold hard numbers, I decided to re-examine a number of possible alternative scenarios.
And in this final scenario, I third the vote of the PQ, but only relative to the CAQ and QS; thus presuming that the PLQ would not take the gains from this. For this map I did a few adjustments. First, the Liberals came very close to winning two ridings, so, I gave these to them. Secondly, in ridings where the CAQ would come within 0.05% of victory, I gave it to them (just one riding) and lastly, in the remaining ridings (3 of them) where the ON took more than 1% of the vote (all of these ridings) I gave the ridings to them.