Skip to content

The STAR second round — FAQ, and how to answer twisted claims

Everything people push back on about STAR's automatic runoff (the "second round"), sorted into what's a fair point, what's misleading, and what's simply false — plus the one honest philosophical trade-off worth conceding.

→ Mechanics: The Automatic Runoff · Why two rounds? · the caveats: STAR's honest limits · official: starvoting.org/why_the_runoff.

First, what each round actually measures

Most bad-faith arguments start by mislabeling the two rounds. Get this straight and half the "gotchas" evaporate:

flowchart LR
    B["Your 0–5 ballot<br/>(scores every candidate)"] --> R1
    R1["<b>Round 1 — Scoring</b><br/>sum every 0–5 score<br/><i>total expressed support:<br/>breadth AND intensity</i>"] -->|top two advance| R2
    R2["<b>Round 2 — Automatic Runoff</b><br/>your full vote → the finalist<br/>you scored higher<br/><i>majority preference,<br/>equal weight per voter</i>"] --> W["Winner"]
  • Round 1 (scoring) already captures breadth. A candidate can top the score round with a pile of moderate 3s and 4s (broad, mild support) or with fewer intense 5s. It rewards total expressed support — breadth and intensity together. This is where the expressive, "how much" signal lives.
  • Round 2 (the runoff) is the majoritarian check. Among the top two, every ballot counts as one full vote for whichever finalist it scored higher (equal weight, regardless of the gap). It asks: does the score leader actually beat the runner-up when every voter counts equally? That guards against a candidate riding lopsided intensity from a minority — Score voting's known weakness — and guarantees the winner is majority-preferred over the runner-up.

So the honest one-liner: Round 1 measures how much support each candidate has; Round 2 checks that the leader can actually win a majority head-to-head. (If someone says "round one is intensity, round two is breadth," they've got it backwards — the breadth is already in the scores.)

The claims, sorted

Claim you'll hear Verdict The honest answer
"The runoff throws your scores in the garbage." ❌ False Your scores did the most important job — they chose the two finalists. In the runoff your ballot is still counted, as one vote for the finalist you scored higher. Nothing is discarded.
"If my favorite isn't a finalist, my ballot doesn't count in the runoff." ❌ False It counts for whichever finalist you scored higher. Score your favorite 5 and the finalists 3 and 2? Your full vote goes to the 3. You're only "no preference" if you scored the two finalists equally — and that's your own stated intent, not disenfranchisement.
"Equal-score voters are silenced in the runoff." ⚠️ Misleading Their scores were fully counted in Round 1 (helping pick the finalists). Scoring the two finalists equally means "I'm happy with either" — counted exactly as intended. It doesn't tip a race you said you're neutral on.
"STAR is just IRV with extra steps." ❌ False IRV eliminates and transfers over many rounds and only ever looks at each ballot's current top choice. STAR has one runoff, between the two highest-scored candidates, using your whole ballot at once. Different ballot, different count, different failure modes.
"A 50.1% runoff isn't 'broad support'." ⚠️ Half-fair Correct that "wins the runoff" means "majority-preferred over the runner-up," not "loved by all." But breadth was already measured in Round 1 (the score totals). The runoff adds the majority check on top; it doesn't claim to be the breadth measure by itself. Cite both numbers, not just one.
"The highest scorer should always win — a reversal is a bug." ⚠️ Philosophical This is the real debate (see below). It's a deliberate design choice, not a malfunction: STAR trades a little cardinal-utility maximization for majoritarian legitimacy. Reasonable people weight those differently.
"STAR can elect someone who loses head-to-head to a third candidate (fails Condorcet)." ✅ True Genuine limit. STAR only runs the runoff between the top two scorers, so it can miss a Condorcet winner who scored third. It's rare in practice and STAR still usually elects the Condorcet winner — but yes, not always. See honest limits and three notions of winner.
"Two rounds is too complicated for voters." ⚠️ Weak Voters do one thing: score 0–5, like rating movies. The two rounds are how it's counted, not extra work at the ballot box — and it's fully summable / hand-countable. Choose-one is simpler to count and far worse at representing people.
"The runoff punishes enthusiasm / builds a beige government." ⚠️ Philosophical Same trade-off as the reversal debate. STAR does deliberately stop a passionately-loved minority candidate from beating a broadly-preferred one — most people call that a feature, but it's a values question, not a factual error.

The Runoff Reversal — the honest core of the debate

A Runoff Reversal is when the candidate with the highest score total loses the runoff to the second-highest scorer. The engine even names it and refuses to call it a bug:

[Runoff Reversal]
 - Score Round Winner(s) = (Uma)
 - Runoff Round Winner   = (Rye)
  Candidate Uma earned the highest total score, but
  Candidate Rye won the automatic runoff — not a malfunction,
  STAR working as designed: the runoff elects the finalist preferred
  by the majority (of voters with a preference).

But be honest: not all reversals are equally convincing. Some clearly serve the majority; some are genuinely debatable. Show both.

Scenario A — a convincing reversal (the runoff earns its keep)

Three candidates, 100 voters: Max is a passionate-minority favorite, Nora the broad compromise, and Cal a no-hoper who never reaches the runoff.

Ballots:            Max  Nora  Cal
  45 voters      ×   5    2    0     (Max's intense base)
  55 voters      ×   2    3    1     (mildly prefer Nora)

Round 1 (scores):  Max 335  ·  Nora 255  ·  Cal 55   → Max & Nora are finalists (Cal drops)
Round 2 (runoff):  Nora 55  vs  Max 45               → Nora WINS

Max's high total came from 45 people who love him. But 55 of 100 prefer Nora. Pure Score voting would crown Max on the strength of an intense minority; STAR's runoff catches that a majority actually wants Nora. Here almost everyone nods — the reversal is obviously right.

Reproduce it: reversal_convincing_c3_b100.yaml.

Scenario B — a jarring reversal (the real philosophical drawback)

Same setup, different electorate. Uma is a near-consensus, high-utility candidate, Rye is polarizing, and Tao is a no-hoper.

Ballots:            Uma  Rye  Tao
  51 voters      ×   4    5    0     (love Rye, but Uma is a strong 4)
  49 voters      ×   5    0    1     (love Uma, reject Rye)

Round 1 (scores):  Uma 449  ·  Rye 255  ·  Tao 49   → Uma & Rye are finalists
Round 2 (runoff):  Rye 51  vs  Uma 49               → Rye WINS   (Uma avg 4.5 vs Rye 2.6)

Uma is nearly everyone's near-favorite — a 4.5/5 average, almost twice Rye's total. Yet Rye wins the runoff by a single vote, 51–49. A pure utilitarian says STAR just elected the candidate who makes the electorate less happy overall, on a razor-thin ordinal majority, and threw away a huge cardinal signal. A majoritarian says: 51 people preferred Rye, and "more than half preferred them" is the most legitimate thing an election can say.

Reproduce it: reversal_jarring_c3_b100.yaml.

Concede it plainly: this is a genuine philosophical drawback of STAR, not a bug to explain away. STAR is a hybrid — a cardinal (score) ballot with an ordinal (majority) finish — and Scenario B is exactly where those two values pull apart. STAR chooses majority preference over utility maximization at the final step. That's defensible, and it's also a real cost; whether it's the right call depends on whether you think an election should maximize total satisfaction or honor majority rule. (Score-voting advocates would elect Uma; STAR and every majoritarian method elect Rye.)

How to argue it well

  1. Lead with the honest concession, then reframe. "Yes, STAR fails the Condorcet criterion and yes, the score leader can lose — here's why that's a deliberate majoritarian choice." Owning the limit disarms the gotcha and builds trust; overselling invites backlash.
  2. Always cite both numbers. The score total and the runoff margin. Most twisted claims work by quoting one and hiding the other.
  3. Separate facts from values. "Scores are thrown away" is a factual error you correct. "The highest scorer should always win" is a values position you engage — don't treat a philosophical disagreement as if it were a factual mistake, or you'll sound evasive.