An Alternative Vote system does not solve the major issues of First-Past-The-Post, and can still have wildly distorted outcomes relative to vote preferences. Unless combined with multi-member districts (aka Single Transferable Vote) it’s barely/not an improvement for Canada – it’s a convenient option for the largest parties because it rarely makes a difference to the outcome and the votes all trickle back to them anyway, so they can make a change without making a change.
Here’s an article worth reading. https://www.fairvote.ca/expert-dennis-pilon-sets-the-record-straight-about-the-alternative-vote/
The only desirable outcome of electoral reform is one which introduces at least a degree of proportionality – Single Transferable Vote if you’re really itching for a ranked ballot, or Mixed-Member Proportional Representation otherwise (my preference - but either would be a great improvement).
The other responder answered how many votes are considered extras, but I’m assuming you may be wondering about how those extra votes get allocated. If so, they would be divided up proportionally based on the 2nd choices of all the voters for that candidate or party. E.g. if Armchair won with extra votes, and all of Armchair’s voters 2nd choices were: 50% for Couch, 30% for Loveseat, and 20% for Stool, then the extra votes would be divided up to match those percentages.
This is a tangent, but when talking about STV I have seen at least one person comment that it does not solve for strategic voting, noting that if you liked a candidate but also knew that:
they were very popular and likely to win,
you had a rough idea of how the surplus votes would be split (possibly from prior polling),
you could make your vote count for more by voting for your 2nd choice (e.g. Stool) as your 1st choice, making your vote for Stool 5 times more potent than if you were just another extra voter for Armchair. I.e. it doesn’t eliminate strategic voting and so could cause distorted/degenerate outcomes. However, that relies on a lot of assumptions and is much less predictable than under First-Past-The-Post. There’s also at least one research article saying that strategic voting under STV is computationally difficult https://sci-hub.st/https://www.jstor.org/stable/41105995