The strategic pathologies — five "Molochs," and where STAR stands¶
Every single-winner voting method has a strategic failure mode: a situation where a voter's honest ballot works against them, and the individually-rational move is to lie. Jameson Quinn (Center for Election Science) catalogued these as "Molochs" — coordination traps where each voter's self-interest drags the whole electorate to a worse place. This page lays out the five, names the game each one is, and gives the honest scorecard: which ones STAR and Ranked Robin escape, which they only soften, and which nobody can escape.**
Attribution & lean. The "Moloch" framing is Quinn's, and he's advocacy-adjacent — his own preferred methods are Approval and 3-2-1, not STAR. That makes this a useful outside lens: where the scorecard credits STAR, it's crediting it against a yardstick built by someone who doesn't champion it. Each row links to a runnable case in this library so you can check the result yourself.
The five Molochs at a glance¶
| Moloch | The game it is | Method it plagues | Runnable here |
|---|---|---|---|
| Lesser Evil | coordination trap (wasted-vote) | First-Past-the-Post | vote-splitting / spoiler · spoiler effect |
| Center Squeeze | premature elimination | RCV-IRV (Hare) | center squeeze · favorite betrayal |
| Dark Horse | prisoner's dilemma | Borda / ranked-point | the Dark Horse |
| Chicken / Burr | chicken (snowdrift) | Approval / Score | the chicken dilemma |
| Condorcet Cycle | Gibbard-universal | every method | cycle resolution · Gibbard–Satterthwaite |
The scorecard — STAR, Ranked Robin, and 3-2-1 on each¶
Legend: ✅ escapes · ⚠️ softens (non-slippery, not immune) · ➖ shares (nobody escapes). 3-2-1 is Jameson Quinn's own rated method, included because it's the one he built to fight these very Molochs.
| Moloch | STAR | Ranked Robin | 3-2-1 | Notes |
|---|---|---|---|---|
| Lesser Evil | ✅ | ✅ | ✅ | rated/ranked ballots remove the wasted-vote fear; a third candidate never spoils |
| Center Squeeze | ✅ | ✅ | ✅ | STAR's runoff, RR's pairwise wins, and 3-2-1's fewest-Bad finalist step all keep the broadly-liked center IRV eliminates |
| Dark Horse | ✅ | ✅ | ✅ | STAR reads support; RR rewards no burial (but has its own sincere seam — fine print); 3-2-1 bolts on an explicit ½-Good guard |
| Chicken / Burr | ⚠️ | ✅ | ⚠️ | STAR & 3-2-1 make the slippery Approval slope non-slippery via a final head-to-head step; RR just lets the majority beat the third candidate |
| Condorcet Cycle | ➖ | ➖ | ➖ | Gibbard: no method escapes; cycles are rare (~1–5%) and must be resolved somehow |
The honest tally: STAR and 3-2-1 score identically — escape three, soften one (non-slippery chicken), share the cycle. That's no accident: both are rated ballots with a final pairwise step, close cousins compared head-on on the 3-2-1 page. Ranked Robin escapes four and shares only the cycle — at the cost of its own sincere dark-horse seam (fine print). None is strategy-proof — Gibbard forbids that for any real method — but all three dodge the traps that actually bite in practice, which is exactly what separates them from Plurality, IRV, and Borda.
Fine print (three caveats the scorecard glosses, raised in the discussion of Quinn's originals): - PR doesn't fully kill Lesser Evil. Proportional representation softens it, but any threshold recreates a milder wasted-vote fear — a small new party below the cutoff still feels unsafe to support. Escape is partial, not total. - Ranked Robin has its own, sincere dark horse. RR escapes the strategic Borda Dark Horse (above), but a Condorcet method can crown a thinly-supported unknown everyone merely tolerates as a second choice — a real RR limit. STAR resists both, because it reads support, not just order (preference vs support). So on Dark Horse, STAR is a touch ahead of RR. - Arrow only binds ordinal methods. The impossibility that dooms ranked methods (Arrow) doesn't apply to a cardinal ballot — which is why STAR escapes Arrow but not Gibbard (STAR escapes Arrow, not Gibbard). The Condorcet-cycle row is a Gibbard limit, not an Arrow one.
The five, one paragraph each¶
Lesser Evil (FPTP). The one you already live under. When you can only pick one, voting your true favorite "wastes" your vote, so you're pushed to the least-bad viable option — and the field collapses toward two poles. Score and pairwise ballots remove the wasted-vote fear entirely. → spoiler effect.
Center Squeeze (IRV). Fix wasted votes with instant-runoff and you get a new trap: a broadly-liked centrist with few first choices is eliminated early, even though they'd beat either wing head-to-head — so their would-be supporters, by ranking them honestly, help elect the candidate they like least. This is the library's most-documented pathology. → center squeeze, favorite betrayal, and the real Alaska 2022 and monotonicity cases.
Dark Horse (Borda). Under a strictly-ranked point method, the only way to oppose a rival is to rank someone above them — so factions bury their rivals behind a harmless nobody, and if everyone does it, the nobody wins with zero honest support. A prisoner's dilemma. A score ballot lets you oppose without being forced to support. → the Dark Horse.
Chicken / Burr (Approval). Two allies must cooperate to beat a third, but honest cooperation leaves them tied, so each is tempted to bullet-vote — a slippery slope that can hand victory to the majority-opposed candidate. STAR's runoff converts the slippery slope to a non-slippery one (the same fix as Quinn's 3-2-1): support both allies honestly, and a few defectors can't start an avalanche. → the chicken dilemma.
Condorcet Cycle (universal). Sometimes a majority prefers A to B, B to C, and C to A — a rock-paper-scissors with no honest winner. This is at the heart of the proof that no method escapes strategy (Gibbard–Satterthwaite). Cycles are genuinely rare (~1–5% in realistic models and empirical data), and every method — STAR and Ranked Robin included — must simply resolve them somehow. → cycle resolution.
The takeaway¶
The Molochs aren't a reason to despair of voting reform — they're a method-selection tool. Four of the five can be designed away; only the Condorcet cycle is permanent, and it's both rare and mild. The lesson Quinn draws, and the one this library's cases keep landing on:
Don't force voters to dishonestly support one candidate merely to oppose another. A ballot that lets you say how much you like each option — independently — sidesteps Dark Horse and softens Chicken; a runoff or pairwise check sidesteps Center Squeeze and Lesser Evil. That's STAR and Ranked Robin, and it's why they hold up against a scorecard built by someone rooting for neither.
Sources¶
- Jameson Quinn, The Six Voting Molochs — the framing and the four constructed scenarios (Dark Horse, Lesser Evil, Center Squeeze, Chicken). Advocacy-adjacent (Quinn favors Approval/3-2-1); used here as an outside lens, with every result reproduced in-repo.
- Gibbard–Satterthwaite theorem — why no method is strategy-proof (the cycle is the seed of the proof).
- Scott Alexander, Meditations on Moloch — the "Moloch" metaphor for multipolar coordination traps that Quinn borrows.