Methods and payout calculations

Supporting material by Splitrule. October 14, 2026.

Read the article | Contract sources

Selection and matching

The article examines selected reported disagreements and two dated contract comparisons. It does not estimate how often venues disagree. Five earnings candidates were checked from a prior list of seven. The other two were not checked.

Pairs were matched by event, keyword, exact contract identity, and chronology. Ticker suffixes alone do not establish event dates. Walmart remains a qualified pairing because its subtitle conflicts with its trading and settlement chronology.

Kalshi finalized records and settlement values support its recorded outcomes. Polymarket resolved event records were cross-checked against winning-token flags from its CLOB API. A quoted market price or a closed flag alone was not treated as final settlement.

Rule versions and source clocks

Rules retrieved on October 14 do not establish that the same wording applied to every earlier trade. Current templates can differ from historical versions. The article does not allege an adjudication error or a proved retrospective rule change.

The Microsoft finding comes from the complete official written transcript. Two separate parsers checked keyword occurrences and speaker attribution. The original audio was not audited. The transcript supports the proposed speaker explanation without proving the adjudicators’ reasoning.

Event deadlines, trading-close times, resolution times, and customer funds availability are different measures. Article dates use Eastern time where specified. Source-observation timestamps on the companion page use UTC. No customer-payment delay was calculated from unlike fields.

Payout units

Each example uses one unit of each leg, with a maximum face payout of $1 per contract. Dollar-denominated units are treated at parity for the arithmetic. Stablecoin, custody, transfer, and execution risks are outside these calculations.

Let sA and sB be the Yes settlement fractions. Where Yes and No payouts sum to one on each venue:

• Yes A plus No B returns 1 + sA - sB.
• No A plus Yes B returns 1 - sA + sB.

Matching binary results return $1 per pair. Opposite results return $0 in one direction and $2 in the other. With sA = 0.26 and sB = 1, the returns are $0.26 and $1.74.

The invented $90 purchase covers 100 units of each leg. Matching binary results return $100, for a $10 gain before fees. The stated opposing results return zero, for a $90 loss. The reverse portfolio is not assumed to have the same purchase cost.

Refunds and half payouts

In the invented example, the first leg costs $0.30 and the second costs $0.60. A refund of the first leg leaves a total return of $0.30 or $1.30. Relative to the $0.90 entry cost, that is a $0.60 loss or a $0.40 gain.

A fixed $0.50 settlement of the first leg leaves a total return of $0.50 or $1.50. Relative to the same entry cost, that is a $0.40 loss or a $0.60 gain. Both operations scale with contract quantity. Only the refund in this example depends on the price paid.

Fees and rebates are excluded because no actual fill is reconstructed. These figures describe contract payouts and hypothetical positions, not audited customer receipts or executable opportunities.

Scope and limitations

Exact venue records support the reported final outcomes. The Bitcoin comparison settled in the same direction on both venues despite different reference-price definitions.

The checks do not establish historical rule continuity, customer payments, blockchain redemptions, executable liquidity, or a market-wide disagreement rate. The speaker-containment example holds other rules constant. The Hantavirus example follows the retrieved descriptions without assuming a medical relationship between the classifications.