When both sides are a little right: the case for split verdicts

Ask anyone in the middle of a dispute and they’ll tell you the other side is entirely wrong. Ask anyone outside the dispute and they’ll usually see something messier: a landlord with some legitimate deductions and some padding; a freelancer whose work was late and whose client kept moving the target; two roommates keeping two different, sincere ledgers. Most real disputes aren’t 100/0. They’re 70/30, 60/40 — both sides holding a piece of the truth and rounding themselves up to all of it.

The distortion of forced winners

A system that must name one winner does violence to every case in that middle band. Forced to choose, it rounds 60/40 up to 100/0 — and both parties learn the wrong lesson. The “winner” hears that their padding was fine; the “loser” hears that their legitimate 40% counted for nothing, and correctly concludes the process was unfair. Worse, the anticipation of winner-take-all changes behavior upstream: when the verdict will be all-or-nothing, both sides inflate, because admitting any weakness feels like handing over the whole case.

What an honest split looks like

This is why Settle was built to be able to say “split” — and to mean it mechanically, not rhetorically. Referees from four labs — Anthropic, DeepSeek, Google, and Groq — each rule independently, and each ruling is tagged unanimous, majority, or split based on the referees’ real positions, computed rather than asserted. The synthesis judge rules issue by issue, not case by case: a dispute isn’t one question, it’s five — was the deadline missed, was the scope changed, was notice given — and ruling on each separately is how a verdict ends up honestly proportioned instead of artificially total.

Dissent is preserved, not averaged away

The detail we’re most deliberate about: when one referee disagrees with the majority, the verdict names the dissent and its reasoning rather than blending it into a mushy consensus. Averaging dissent away would make reports look cleaner and more authoritative. It would also be quietly dishonest — a 3–1 ruling and a 4–0 ruling are different facts about a case, and you deserve to know which one you got. A visible dissent tells you exactly where your dispute is genuinely contestable, which is precisely the part worth another look — or a professional’s eyes — before you rely on it.

Why a split verdict still ends fights

It seems paradoxical: how does “you’re both partly right” resolve anything? In practice it resolves the middle-band cases better, because it matches what both parties privately suspect. A 65/35 finding, issue by issue, with a settlement proposal proportioned to it — a specific amount, a deadline, mutual release — gives each side a story they can accept without self-betrayal: I was mostly right, I wasn’t entirely right, and the number reflects both. (Settle’s verdicts are non-binding neutral opinions — not arbitration, not legal advice — which is exactly why their power has to come from being believable rather than enforceable. A verdict that can honestly say “split” is more believable than one that can’t.)

If you’re in one of those 60/40 fights right now, file it — $14.99, the other side responds free — and see what four referees who don’t need a winner actually make of it.