Taker outcomes

Straight contracts only, and this is not a rounding caveat. Novig's parlays are quoted by request and never become book markets, so not one of them resolves in the feed. They are , so everything here covers roughly the straight % of contract volume. It is never a venue-wide figure.

What made this possible. Novig is the only competitor that publishes an aggressor flag — every trade is tagged taker or maker — and the public GraphQL snapshot now supplies the settled WIN/LOSS the trade file never had. Both together are what make this true taker P&L, not the price-bias proxy other venues are limited to. Only the taker side of each trade is counted; each trade appears twice and summing both would double the venue.

Gross is the headline; net is a band. There is no settlement fee — a winning contract pays the full $1.00. Novig charges the taker 0.03 × P × (1−P) on live straights and nothing pre-game, and the feed carries no live/pre-game flag, so the true fee lies between zero and the ceiling drawn above. Because every window here is a loss and a fee only deepens a loss, gross is the conservative figure and net is more negative.

The status column is free text. It holds WIN, LOSS and TBD and numeric fair-value prices in one column; a naive equality filter would score every fair-value settlement as a loss. Decisive WIN/LOSS contracts drive both charts; the fair-value-settled contracts are P&L-computable (, folded into the daily gross) but have no binary outcome, so they are excluded from calibration.

Measurability. Thousands of prints on one game share a single outcome, so the standard error clusters on the fixture, not the print. — . Individual price bins are noisier and most do not clear on their own; that is expected across twenty bins and is why the venue-level test is the one that decides.

A short, forward-only history. Novig regenerated every identifier at its 2026-08-04 move to a CFTC-regulated exchange, so nothing joins before then and the series is about two weeks long — expected, not a gap. Settlement is read from the current snapshot; a daily re-pull can re-grade a market, so history can shift slightly. The GraphQL source is announced as sunsetting, and the model here is written so the transport can later move to Novig's keyed API without changing what these charts show.

The same-game comparison is a constructed join. Kalshi publishes no fixture id, so games are matched on date and unordered team pair through a hand-checked team crosswalk — MLB only, and it covers the games both venues list. Each side's price is the last straight moneyline trade strictly before the scheduled first pitch: one pre-game snapshot, not a close. Volume is one-sided contracts on each venue's own convention, so the ratio is indicative, not a like-for-like count.