Championship markets

Season-long markets for the soft-launch sports (F1, NASCAR Cup, MotoGP) on Polymarket, Kalshi and OG.com, what the syncs call them, and which of them a model prices. Taken from the code, the schemas and the captured listings in tests/fixtures/market (2026-09-29). Live listings were not checked from the cloud. Last reviewed 2026-09-30.

Model-priced means a model run writes a fair value that db.reads.model_prob or a command reads. Quote only means the market is synced and recorded, but no model prices it.

Sport	Exchange	Market	Listing	Kind (where)	Priced
F1	Kalshi	Drivers' champion	KXF1	champion (kalshi/sync.py)	Model: pricing.forecast standings, on the board
F1	Kalshi	Constructors' champion	KXF1CONSTRUCTORS	constructors_champion	Model: forecast constructors
F1	Polymarket	Drivers' / Constructors' champion	"F1 Drivers' Champion", "F1 Constructors' Champion"	champion / constructors_champion (polymarket/sync.py)	Model
F1	Polymarket	Season wins ("win N+ Grands Prix")	question wording	season_wins_ge	Model: forecast extra.wins_ge
F1	Polymarket	Standings head-to-head	question wording	standings_h2h	Model: forecast extra.ahead_of
F1	OG.com	Drivers' / Constructors' champion	F1-00001-2026, F1-00002-2026 (20 contracts)	champion / constructors_champion (exchanges/og.toml)	Model (venue shown with RACINGLINES_OG_VENUE=1)
NASCAR	Kalshi	Cup champion	KXNASCARCUPSERIES, KXNASCARCUPCHAMP	params.kind = champion, link unmodeled (sources/nascar/links.py)	Model, offline: racinglines model season nascar --quotes (off by default, read-only)
NASCAR	Polymarket	Cup champion	nascar-cup-series-2026-champion-…	champion from the wording	Model, offline: the same
NASCAR	OG.com	Cup champion	NSCAR-00002-2026 (17 contracts)	champion; og.toml modeled = false	Model, offline: the same
NASCAR	Kalshi	Regular-season champion	KXNASCARCUPSEASON	regular_season_champion	Quote only (settled for 2026: the regular season ended 29 Aug)
NASCAR	Kalshi	In-Season Challenge	KXNASCARCHALLENGE	in_season_challenge	Quote only (a bracket tournament, not the points table)
NASCAR	Kalshi	O'Reilly (Xfinity) / Truck champion	KXNASCARAUTOPARTSSERIES, KXNASCARTRUCKSERIES	champion, nascar_series xfinity / trucks	Quote only (only Cup results are ingested)
MotoGP	Kalshi	World champion	KXMOTOGP (22 markets)	none: unmodeled (sources/motogp/links.py knows only KXMOTOGPRACE)	Model, offline: motogp season-replay reads it as champion in memory
MotoGP	Kalshi	Teams' championship	KXMOTOGPTEAMS (11 markets)	none: unmodeled	Quote only
MotoGP	Polymarket	World champion (25 outcomes)	tag motogp	none: unmodeled	Model, offline: the same (from the wording)
MotoGP	OG.com	none listed

How far a model price goes. None of these champion prices has a backtest that held up. The F1 season forecast lost money on Polymarket's 2026 championship markets at every checkpoint and in every model variant (−$566 to −$797 over 33 markets, Market making): early in a new-regulations season the market knew more than the model. The NASCAR season sim (#93) is not calibrated: it has never been scored on a settled season, it ignores track types and form drift, and its first Mac run had Larson at 94.6% against 40–55% on the market (before its second revision). The replays that would test these prices on Kalshi and OG.com (F1: #95; NASCAR and MotoGP: #96) are merged on main (off by default, below) but not yet run on the VM as of 2026-09-30. Read every "Model" cell above as a fair value to compare with the quote, not as a validated edge. The full sport × exchange status: Sports and exchanges.

No exchange lists a NASCAR standings top 3, a standings head-to-head or season wins, although the season simulation could price them (markets.kinds.season_fair). No NASCAR manufacturers' or owners' championship was found either.

What pricing MotoGP's champion would take

The MotoGP race model (MotoGPRaceChallenger) prices one race at a time, so it could feed a season simulation the way the NASCAR model does. What is missing:

1. Sprint points. MotoGPRace.load reads the Sunday race only, and sprint classifications are not ingested yet (sources/motogp/ingest.py). The sprint pays 12-9-7-6-5-4-3-2-1 every weekend, so the season total is wrong without it.
2. The remaining schedule. The MotoGP ingest files completed events only. The events endpoint that fetch.py already reads has the dates.
3. Today's standings. Summed race and sprint points, or the world-standing classification in tests/fixtures/market/motogp_standings_2026_motogp.json, which nothing ingests yet.
4. Classifiers. KXMOTOGP should map to champion (and KXMOTOGPTEAMS to a team kind), and Polymarket's wording to champion. That changes stored link params, so it needs the owner's go and link --apply with a backup.
5. A points format. A ChaseFormat-like table with no playoff (25-20-16-13-11-10-9-8-7-6-5-4-3-2-1 for the race).

Steps 1-3 need a probe from the Mac first (standing rule), so none of this is built.

Update (2026-09-30): a replay-only version is built (racinglines strategy season-replay motogp, off by default with RACINGLINES_SEASON_REPLAY=1, read-only). It covers the steps above without changing any stored link. - models/motogp_season.py supplies the points format (5). - The remaining schedule is the stored events plus the fetched events.json (2). - The standings are summed from the stored Grand Prix and sprint points (3). With no sprint results stored, both sides are Grand Prix only and the report prints SPRINT POINTS MISSING (1). - The classifier runs in memory only: KXMOTOGP, or a listing naming the championship, is read as champion, and the rider comes from the name resolver (4). Nothing is written.

Championship replays (backtests)

racinglines strategy season-replay nascar and racinglines strategy season-replay motogp (pipelines/season_replay.py) trade the F1 championship sleeve's default strategy (markets/strategies/season.py) against each exchange's recorded champion prices. - Decisions: one at 12:00 UTC the day after each points race. - Fairs: an as-of season forecast from only the results before the decision. For NASCAR this is models/nascar_season.py summed from race results, because the points feed is today's only. - Costs: Kalshi's taker fee, OG.com's flat fee, or Polymarket's half-spread, each plus slippage.

Both commands are read-only and off by default (RACINGLINES_SEASON_REPLAY=1). They write only CSVs under data/runs/season-replay/<sport>-<venue>-<year>/ (decisions.csv, trades.csv, positions.csv, equity.csv, summary.json); --venue all (the default) runs every exchange the sport lists, one report each (CLI).

The F1 replay takes --venue kalshi|og too: racinglines strategy season-strategy f1 --venue kalshi|og replays the default championship strategy on that exchange's stored tape (no fetch). Kalshi is costed at its taker fee; OG.com at its flat $0.02 per contract, with no volume floor and prices read as its replay venue reads them (an empty book's 0.50 dropped). Polymarket stays the default and its output is unchanged. --save with another venue stores kind season_venue_replay, so no page reads it as Polymarket's; --paper stays Polymarket or Kalshi.

These are first backtest cells, not validated strategies. Both season forecasts simulate races independently from the form at the decision, and NASCAR's was more favourite-heavy than every exchange on 2026-09-30.