Strategic Voting Across the Equal Vote Methods¶

Strategic voting (also tactical or insincere voting) is casting a ballot that doesn't reflect your honest preferences, in the hope of a better outcome. The Equal Vote Coalition treats honesty as a core test of a voting method: can a voter safely express her sincere opinion, and how much does the method reward insincerity?

Two facts frame everything below:

No method is immune. The Gibbard–Satterthwaite theorem proves that every ranked or scored method has some scenario where a voter can gain by voting insincerely. So "you can construct a strategy" isn't a mark against a method — the real questions are how often it pays, whether it's actionable, and whether it backfires.
Honest ≠ the same as strategy-proof. A method can fail a criterion on paper (STAR fails Later-No-Harm, for instance) yet still make honesty the practical best play. This page separates the paper property from the practical incentive, and tries to be even-handed to all three Equal Vote methods — STAR, Approval, and Ranked Robin.

External references worth reading directly: the Equal Vote Coalition's Strategic STAR and Gaming the Vote, and STAR Voting's Is STAR vulnerable to strategic voting?. (These are advocacy sources; this page reports their claims and evidence while flagging the honest counterpoints.)

The four kinds of insincere vote¶

The Equal Vote taxonomy names four distinct ways to vote insincerely on a rated or ranked ballot. The names matter — people argue past each other by confusing them:

Type	Also called	What the voter does	Worked in full
Strong insincerity	Favorite betrayal, compromising, decapitation	Gives someone a higher score/rank than their true favorite	Favorite betrayal — Voting 301 · the pair
Weak insincerity	Burial, skipping	Keeps the favorite on top, but ranks/scores a less-preferred candidate below an even-less-preferred one	Burial (topic hub)
Restrictive sincerity	Tactical minimization, bullet voting, truncation (down-voting)	Lowers support for non-favorites (e.g. scores everyone but the favorite a 0)	Restrictive sincerity
Expansive sincerity	Tactical maximization (up-voting)	Raises support for non-favorites (e.g. gives a strong front-runner a 5 to hedge)	Expansive sincerity

Only the first two distort preference order; the last two only distort how much. That distinction is why fully-expressive methods (STAR, Score) face the bottom two while choose-one and ranked methods face the top two.

→ The four side by side, with the runnable elections for each and a four-question test for telling them apart: the four kinds of insincere vote.

How each method handles each strategy¶

This is a comparison, not a verdict — each method resists some strategies well and is exposed to others.

Strategy	Choose-One / IRV (the contrast)	Approval	Ranked Robin (Condorcet)	STAR
Favorite betrayal	Strongly incentivized (choose-one) or possible (IRV center squeeze)	Not needed — approve favorite freely	Rare; sincere ranking usually best	Rare; runoff makes it backfire
Burial	n/a (choose-one) / possible in IRV	n/a (no order)	The main worry — you can bury a rival	Possible but rarely pays (see below)
Bullet voting (min)	is the ballot	The central tension — approve only your favorite?	n/a (must rank)	Runoff discourages it
Tactical maximization (max)	n/a	The other side of the tension — approve a front-runner you dislike?	n/a	Runoff discourages it
Main honesty question	"Is my favorite electable?"	"Where do I draw my approval line?"	"Do I dare bury a rival?"	"Do I dare mis-order the top two?"

A few honest observations from that table:

Approval's strategy is the most actionable. Its one genuine dilemma — the approval threshold: do you approve only your favorite (bullet voting) or also the tolerable front-runner (tactical maximization)? — is easy to reason about, which cuts both ways: simple to use, but also simple to game. Approval never asks you to betray your favorite, which is its real strength. See Approval's honest limits.
Ranked Robin's exposure is burial. Because it's purely ordinal (order, not strength), a faction can in principle bury a strong rival below weaker candidates. In practice sincere ranking is usually optimal and Condorcet methods are considered strongly strategy-resistant — but burial is the named risk. See Ranked Robin's honest limits, and — because burial is a cross-method strategy, not Ranked Robin's alone — the Burial topic hub, which runs it against Copeland, Borda, STAR and IRV side by side.
STAR's exposure is the bottom two, softened by the runoff. Because scores are summed in round one but treated only as order in the runoff, exaggeration has limited upside and real downside (below). STAR does fail Later-No-Harm — honestly scoring a second choice can, in constructions, help them beat your favorite — which is a genuine, admitted property. See STAR's honest limits.

Why exaggeration tends to backfire in STAR¶

The mechanism is worth understanding because it's STAR's clearest honesty argument, and CPython's Tim Peters put it crisply: STAR "uses each [system] to counter the worst of the other." Scores are cardinal (summed) when choosing the two finalists, then ordinal (just which is higher) in the runoff:

Betray your favorite (score a "lesser evil" above them) → you might knock your favorite out of the runoff, and your full runoff vote then goes to the lesser evil instead. Backfires.
Bury a strong consensus candidate (the FairVote hypothesis) → to do this you must raise someone you like less, which raises the chance your own favorite gets squeezed out of the top two. And if several factions all try it, all but one end up worse off than if they'd been honest.
Bullet-vote or exaggerate the scale → the runoff only reads order, so a 5-vs-4 counts the same as 5-vs-0; you gain nothing in the runoff by inflating, and you lose your say among candidates you flattened together.

The honest limit on this argument: it says exaggeration rarely pays and often backfires — not that STAR is favorite-betrayal-proof. It isn't. The measured version is the brute-force simulation in Favorite Betrayal — Voting 301, which finds STAR and IRV fail the favorite-betrayal criterion at about the same low rate — STAR's real edge is that a betrayal backfires far more often than it works, not that it never works.

What the simulations say¶

Strategic incentive is often summarized as a ratio — how often a strategic vote helps the voter vs. how often it backfires. From Dr. Jameson Quinn's Voter Satisfaction Efficiency work:

Choose-One Plurality ≈ 17:1 — strategy almost always pays (the "lesser evil" problem).
IRV ≈ 3:1 — strategy pays about three times as often as it backfires.
STAR ≈ 1:1 — a strategic vote is about as likely to backfire as to help, so it isn't worth attempting.

Two nuances kept for honesty: (1) these are model outputs — the exact numbers move with the electorate model, field size, and assumptions (this repo's own simulations make the same point about not quoting a bare rate); (2) a separate finding is that STAR, IRV, Score, and Condorcet methods all produce more representative outcomes when voters are honest — the difference is whether the method also incentivizes honesty at the individual level, where STAR and Condorcet score well and IRV less so.

Actionability is the practical clincher: even where a STAR or Condorcet strategy exists, it only bites in a close three-way tie, and executing it needs near-perfect polling on everyone's scores — and if one faction tries it, rivals can respond, making the outcome less predictable. That's why the honest advice under STAR, Condorcet, and 3-2-1 is simply show your sincere preference order.

The even-handed bottom line¶

Approval never makes you betray a favorite — its only tactic is where to set your approval line, which is easy to reason about (a strength and an exposure).
Ranked Robin is strongly strategy-resistant in practice; its named risk is burial, and it can't express no preference as gracefully as a scored ballot.
STAR fails Later-No-Harm on paper, but its cardinal-then-ordinal runoff makes exaggeration and burial usually backfire, and its two-round count makes why honesty pays unusually easy to explain.

No method removes strategy entirely (Gibbard–Satterthwaite guarantees that). Among the Equal Vote methods the differences are about how actionable the strategy is and whether it pays — and on those measures all three are dramatically safer than the choose-one status quo.