No more tactical voting under AV

A tactical vote is, by definition, always sensible. The issue here is under what circumstances is a tactical vote a sincere vote with Alternative Voting.

As a rule, the best tactic with Alternative Vote is to:
1) Determine which two candidates are most likely to win.
2) Rank your favorite OF THOSE TWO in first place.

I explain this here in great detail:
http://www.electology.org/irv-plurality

The short explanation is that, of all the "what might happen" scenarios you discuss, only TWO are relevant to a rational voter's expected utility calculations:

1) The tactical vote switches the winner from my less preferred assumed-frontrunner to my most preferred assumed-frontrunner. (The tactic HELPS.)

2) The tactical vote switches the winner from my sincere favorite to my most preferred assumed-frontrunner. (The tactic HURTS.)

#1 is clearly obviously more likely, since #2 requires that the assumed-weak candidate you favored surprises and defeats BOTH assumed-frontrunners, whereas #1 only requires him to defeat ONE frontrunner.

Maybe it would help to see a Princeton math Ph.D. explain it.
ScoreVoting.net/TarrIrv.html

Oh, and those four circumstances you list are inaccurate. NONE of those is required for the expected value calculation I explained to be in favor of the insincere tactic.

As I explained in your previous article, the particular example is irrelevant. It absolutely does not matter that the LibDems are in the middle, which is why your third condition above is inaccurate.

The ONLY thing that matters is that a weak party is more likely to defeat ONE major parry than to defeat BOTH. This is simple statistics. One possibility is a subset of the other.

Anthony,

"If your favourite candidate is the centre candidate, then voting for the third-place candidate is unlikely to help"

But it's even LESS likely to hurt, which means it's still an advisable strategy.

This is an issue of RELATIVE probability between two outcomes, and that's a point you seem to keep missing.

"you clearly have some ulterior motives in all of this."

My motive is educating. My organization serves as a "FactCheck.org for voting systems". We came about largely as a result of some extensive insights by a Princeton math Ph.D. named Warren D. Smith, which challenged a lot of conventional wisdom on voting theory. Smith went on to co-create the site RangeVoting.org (aka ScoreVoting.net), which I discovered in 2006. We are featured in a book called "Gaming the Vote", by William Poundstone. You can see more about our new organization (currently incorporating as a California non-profit corporation) here:
http://www.electology.org/about-us

"Go and read your Princeton PhD friend’s example again, and see that, whatever the numbers are, tactical voting does not make sense if your favourite candidate is the middle one."

That is a single example, not a statistical aggregate, so it says no such thing. It is purely meant to serve as a tool for expressing the core concept.

"My point about the Lib Dems was that if tactical voting under AV helps the middle party (as your Princeton PhD friend claims)"

No he does not claim that. He actually claims that IRV tends to elect extremists.
ScoreVoting.net/IrvExtreme.html

What me meant there by "middling" was more like this (left-right axis for simplicity):
Green .. Labour (middling left) .. LibDem (centrist) .. Tory (middling right) .. Fascist (or some UK analog)

He just left a lot of that out for simplicity.

An example would be the French election where Royal was the left, Sarkozy the right, and Bayrou the centrist. Polls actually showed that Bayrou was the Condorcet winner, but because of the "center squeeze" effect of runoffs, he didn't make it to the final round. IRV essentially has the same effect, but much more pronounced, apparently due to the fact that with IRV, voter's cannot change their vote between rounds. Here we speculate on possible explanations:
scorevoting.net/TTRvIRVreasons.html

It is quite conceivable that, under IRV, the LibDems would eventually disappear. Or Labour could disappear, and then LibDems would replace them but then gradually move further left. Which, in a long-term sense, would be roughly the same as changing the name of the Labour Party to the Liberal Democrat Party. That would fit perfectly with the trend we see, that Top-Two Runoff tends toward multipartism, while IRV massively maintains bipartism:
scorevoting.net/TTRvIRVstats.html

"If a party has a good chance of defeating one of the major parties, then it is no longer a weak party in that election, by definition."

But if that minor party has historically been weak, then the only way you would KNOW that it had a good chance of winning would be if you had some other indicator (like good polling), which you said in your last post, doesn't exist. (That was actually about the Burlington, Vermont, USA elections, which you apparently are some kind of expert on, since you can assert they didn't have pre-election polls.)

The problem is, once you get two-party domination with IRV, any minor parties WON'T grow to the point of having "a good chance of defeating one of the major parties". Tactical exaggeration keeps them weak. It is extremely difficult for a party to grow from "weak minor party" to "contender" in the span of a single term of office.

"If my favourite party is likely to defeat BOTH of the other parties, then why should I vote against them?"

A minor party is, by definition, NOT likely to defeat both major parties. And therefore your best tactical bet is to betray that party, by ranking your favorite MAJOR party in first place. And then that minor party will STAY weak, because lots of its supporters will vote strategically.

I have placed phone calls to Australian Green Party staffers, for instance. They have told me that one of the most common phone calls they receive is from supporters who ask them, "why should I throw away my vote by ranking Green in first place?" That should serve as a reality check to you. These people have lived their whole lives with IRV in use, and yet they STILL don't how it works. I'm a software engineer in San Francisco, and my fellow engineers who VOTE here don't even know how IRV works. Check out this instant message conversation I recently had with one of them, in which he thinks the tabulation rules are essentially those of the Borda system.
http://www.electology.org/san-leandro-inamdar

You have got to deal with the real world, not your theoretical assumptions about what voters will do.

"if the two major parties are neck-and-neck, then tactical voting against the “weak” party is frankly and obviously absurd under AV (not so under FPTP)."

On the contrary, that is precisely when it makes the MOST sense to vote tactically, since your vote has the biggest chance of helping to flip the election outcome.

"But anyway, this is all irrelevant, since the expected order of the top three candidates is (I strongly suspect) something that cannot be accurately predicted"

WRONG! Since..

A) You don't HAVE to know the finish order, ONLY which two will MOST LIKELY be strongest (regardless of finish order), and
B) It ABSOLUTELY can be accurately predicted based on historical data. E.g. if you live almost anywhere in Australia, you know that there's a 99% chance the winner won't be from a party other than Labor or NatLib.

"And this is all irrelevant again, since you’ve said nothing to defend FPTP over AV."

I'm not defending FPTP. I'm just pointing out that IRV won't help break up two-party domination. It obviously has lower Bayesian regret than FPTP:
ScoreVoting.net/BayRegsFig.html

Also, do you use FPTP, or Top-Two Runoff? There is evidence that TTR is BETTER than IRV, even if FPTP is worse than both.

Also, Score Voting and Approval Voting are worlds better than either.
ScoreVoting.net/CFERlet.html

I want to point out that on one of the most obvious aspects of tactical voting under IRV, you got it exactly backwards.

“if the two major parties are neck-and-neck, then tactical voting against the “weak” party is frankly and obviously absurd under AV (not so under FPTP).”

On the contrary, that is precisely when it makes the MOST sense to vote tactically, since your vote has the biggest chance of helping to flip the election outcome.