Game-line rules - ties, overtime and pushes

Two venues can list the same moneyline and still pay differently. The clearest case is an NFL game that ends level: Kalshi pays each side half, ProphetX voids the contract and returns the stakes. The matched row looks identical; the payout is not.

Every moneyline, spread and total on the sports matching endpoint carries a one-word verdict on that, settlement_equivalence, and a small rule matrix behind it that you ask for with include_rules=true, so the difference is visible before you trade. Matching itself is unchanged: venues are still paired exactly as before, and a venue that settles a tie differently stays in the row. The matrix is the evidence you use to decide whether a pairing is a hedge.

The verdict is always there; the rows are opt-in

The rows are the evidence: each one names every source market, describes the rule in words and links the clause. That makes them about 60% of the bytes of a busy football event, and a page of events with every submarket was about 11 MB with them, large enough to run into the request deadline. So a response leaves them out unless you ask. The usual pattern is to read the verdicts on a page and, for a game whose verdict is different or unverified, look it up with the rows:

GET /v1/matching-markets/sports/lookup?event_id=nfl-pit-no-2026-10-25&include_rules=true

include_rules reads the same on the list and the lookup. It changes what each event lists and never which events come back, so a cursor from one request continues with or without it. It takes one value, an empty value means false, and anything that is not a boolean is a 400. A player prop is not affected: it carries its nine rows whatever include_rules says. The SDK parameter is includeRules in TypeScript, include_rules in Python and IncludeRules in Go.

The three rules

With include_rules=true, each game line has three rows in rule_comparisons, one value per source market (source_markets[]). Its summary, settlement_equivalence, is there either way.

not_possible means the contract cannot end level: a league that plays on until one side wins (the NBA, WNBA, MLB and NHL) cannot end a full game tied, and a spread or total has no tie result of its own because its line decides it. Only the NFL can end a game level, so that is where tie differs in practice. For push, not_applicable means there is nothing to land on: a moneyline has no line, and a half-point line, which is every line matched today, cannot be landed on exactly. For overtime, it means overtime cannot change the result: an NHL game tied after regulation is decided by exactly one goal (sudden-death overtime, or a shootout that adds one goal to the winner's score), so a puck line of more than one goal, such as -1.5, settles the same whether a venue counts overtime or not.

Each value comes with a description in plain words and, for every known value, an evidence_url: the venue's published rules (a Kalshi contract-terms PDF, SX Bet's help center, a ProphetX CFTC filing) or the record that holds the market's own rules text. Where a venue writes no rule but has settled games it decided, the value comes from those settlements and the evidence is the venue's own record of one of them: Predict paid the extra-innings winner of every MLB game that went past nine innings, AlphaArcade paid the full-game winner of every game that went past regulation, and SX Bet voided the moneyline of an NFL game that ended tied. An unknown value has a description that says why and no evidence, because there is no clause to point at.

Reading settlement_equivalence

unknown never means "no restriction", and an unknown value never makes a row equivalent. A venue whose rules say nothing about overtime, for example, leaves the row unverified however many other venues agree. A known difference always wins over an unknown, so different is never hidden by a venue that is silent.

An excerpt

An NFL moneyline where one venue voids a tie, as include_rules=true returns it (abridged to the tie row):

{
  "settlement_equivalence": "different",
  "rule_comparisons": [
    {
      "rule": "tie",
      "label": "Tied game or period",
      "comparison": "different",
      "source_values": [
        {
          "provider": "kalshi",
          "market_id": "...",
          "value": "half_payout",
          "description": "Kalshi: with no \"Tie\" strike listed, \"Yes\" holders are paid $0.50 and \"No\" holders $0.50; ...",
          "evidence_url": "https://assets.kalshi.com/contract_terms/FOOTBALLGAMEWIN.pdf"
        },
        {
          "provider": "prophetx",
          "market_id": "...",
          "value": "void",
          "description": "ProphetX: \"If, after the completion of overtime, the result remains tied, all orders on the outright event winner contract shall be void ...\"",
          "evidence_url": "https://framerusercontent.com/assets/3xBUPoodNIUMWAauxaaccjNRSo.pdf"
        },
        {
          "provider": "sxbet",
          "market_id": "...",
          "value": "unknown",
          "description": "SX Bet's help center does not say what happens to a moneyline when an NFL game ends tied."
        }
      ]
    }
  ]
}

Filter rows with comparison == "different" for known differences and comparison == "unverified" for gaps. A consumer that hedges across venues usually wants equivalent, or a different whose rows it has read and accepted.

What it does not do

• It never changes which venues are paired, the submarket's key, or any identifier. It is published beside the match. (The one overtime rule that does decide a pairing is a second half's or fourth quarter's own, which is in its key; the matrix reports the same rule beside it.)
• It carries no prices and no payout estimate. A half_payout is a rule, not a number.
• Agreement is not a complete contract review. Evidence quotes the clauses that were read; a venue can have further terms.
• The player-prop matrix is separate: props always carry nine rows, with or without include_rules, and the policies on player-prop matching. A total's rules.settlement still says whether the line was keyed with overtime (so does a second half's or fourth quarter's spread or total that counts it), and the overtime row of its matrix never contradicts it.

Where the rules come from

Kalshi, SX Bet and ProphetX publish rules by product family, so their values are read from those documents. Hyperliquid's NFL game winners all use one published template, which pays 0.5 to each side on a tie, and each market states whether overtime counts in its own countedPlay, which is read like a market's text. Polymarket, Predict, Pred, Limitless and AlphaArcade publish their rules in each market's own text, so their values are read from the market's text when it states them (for example "Overtime is included if played" or "If the game ends in a tie, this market will resolve 50-50"). A market whose text is silent is unknown for that rule. A full-game line also reads the text of its game beside the market's own: Polymarket's event, and for an NBA spread or total Predict's game description, which says once for every market of the game that the final score includes overtime (Predict's spreads and totals say nothing about it themselves). A period market states its own rule and reads nothing else. The description on each value says which it is.