Skip to content

STAR for skeptics — the honest 5-minute path

You think scoring candidates sounds gimmicky, you're not convinced anything's really broken, and "sounds too good" makes you suspicious. Fair — good instincts. This page meets you there: no cheerleading, just the fast honest path, and every claim below is backed by an election you can re-count yourself in this repo.

Level: 201 · for debaters If you came from the RCV camp specifically, jump to STAR vs RCV-IRV.

"This looks like a gimmick." It's two steps of grade-school arithmetic: add up the 0–5 scores, then the top two have a runoff. You can count it by hand — it's simpler to tally than most reforms, including RCV.

"Why change at all — what's actually broken?" Choose-one voting splits similar candidates and can elect the option most people didn't want — the spoiler effect. That's why people vote "lesser evil" instead of who they like. See it in 60 seconds: the team lunch.

"Isn't ranked-choice (RCV) already the fix everyone's pushing?" RCV-IRV is better than choose-one, but it has real, specific failures: it can eliminate the broadly-preferred centrist (center squeeze); ranking a candidate higher can make them lose (non-monotonic — watch it flip); and it isn't precinct-summable, so it's hard to hand-count and audit. Those aren't STAR talking points — they're in the neutral criteria table as runnable elections. STAR isn't RCV-IRV.

"Scores can be gamed — won't everyone just vote 5s and 0s?" The runoff removes the payoff. Your in-between scores still choose the finalists, but the winner is decided by simple preference between the top two — so exaggerating doesn't pay. Simulations rate STAR the most strategy-resistant of the common methods, and even the worst case just makes it behave like Approval — still a good method, not a broken one.

"Sounds too good. What's the catch?" — the real question, and the one that should decide it. Here's the honest catch: no voting method is perfect — that's a theorem (Gibbard–Satterthwaite). STAR fails later-no-harm and favorite-betrayal in rare constructed cases, is not a Condorcet method (it can miss a beats-everyone candidate), and asks a bit more of voters than a single mark. This repo states all of it out loud on the honest limits page. A reform that hides its weaknesses is the one to distrust; the honest read is that STAR's trade-offs are good ones, not that it has none.

"Is it even real, or a math toy?" Used by real organizations and political parties since 2019; not yet adopted by any government (recent Oregon ballot measures fell short). The honest, current picture: Is STAR actually used?

The skeptic's real advantage here

You don't have to take any of this on faith. This repo exists so you can check — every claim above is a small YAML election you can re-tabulate with an independent engine, cross-checked against a real platform. So don't believe the advocates (including us); verify:

You don't have to love STAR. But it's a serious, checkable proposal — and the fact that this library tests it against every rival, and concedes where it loses, is itself the reason to take it seriously.