Ok, it's important first of all to explain that I am fully aware that very specific examples can be bandied around to 'prove' anything.
That said the mechanics involved in this example are universal.
Candidates A B C D
Candidate A has 10 votes
candidate B has 7 votes
Candidate C has 4 votes
Candidate D has 2 votes
Fptp would say A is the winner
AV would eliminate D...
D's votes transfers to C.
We now have
Candidate A with 10 votes
Candidate B with 7 votes
Candidate C has 4 first pref votes and 2 second pref votes.
C is eliminated. Say 1 1st pref vote goes to A. the rest goes b
And we now have our winner
B wins with seven 1st pref, three 2nd pref and two 3rd preference.
A with ten 1st pref and one 2nd pref.
Now this result sounds pretty dodgy to me as it is.
We have asked 23 people for first preferences, 6 people for a 2nd preference and 1 person for a 3rd preference.
What about the 17 people who didn't get their 2nd preference registered. Being that you only need 12 votes to win, it is very silly to ignore these 17 2nd preference votes and to go on to other people's less relevant 3rd choices.
What if all 17 had C as their second preference?
Or any other type of combination...this is still more relevant and representative than going onto 3rd preference votes.
This can mean (and will mean) that under AV, people will get elected who wouldn't under Fptp, and who wouldn't if the two most popular candidates supporters weren't ignored after their 1st preference, but were treated the same as everyone else ( or as it is known in the wider world outside the yes 2 AV club, fairly or equally).
Candidates will be voted in not on support but by the quirks of a counting system.
Under Fptp because we do not ask for preferences, we have to go with what we have.
Under AV they ask for preferences...and then ignore in most cases most of the data..so even in a four horse race someone can be elected even if they have less first, second and third preference votes than other candidates.