Live coverage of PMU via the ParlayAPI moneyline / spread / total / player-prop aggregator. Classification: French ANJ-licensed sportsbook. Source key: pmu.
h2h 5,600 rowsspreads 4,312 rowsmoneyline 2,119 rowsmoneyline 2,800 rowsh2h 1,400 rowsspreads 992 rowsmoneyline 5,707 rowsh2h 1,912 rowsmoneyline 7,696 rowsh2h 2,567 rowsmoneyline 7,046 rowsh2h 2,350 rowsmoneyline 6,523 rowsh2h 2,178 rowsmoneyline 8,971 rowsh2h 3,000 rowsmoneyline 1,206 rowsh2h 402 rowsmoneyline 6,981 rowsh2h 2,349 rowsmoneyline 4,071 rowsh2h 1,357 rowsmoneyline 5,196 rowsh2h 1,734 rowsmoneyline 3,015 rowsh2h 1,005 rowsmoneyline 8,789 rowsh2h 2,934 rowsmoneyline 4,482 rowsh2h 1,503 rowsmoneyline 8,535 rowsh2h 3,600 rowsmoneyline 8,578 rowsh2h 3,600 rowsmoneyline 144 rowsh2h 48 rowsmoneyline 7,839 rowsh2h 2,613 rowscurl -s 'https://parlay-api.com/v1/sports/baseball_mlb/odds?regions=us&markets=h2h&bookmakers=pmu' \
-H "X-API-Key: $PARLAY_API_KEY"
We poll PMU on the International / non-US freshness tier, which means we expect a new observation within 90s under healthy operation. Observations between 90s and 300s are flagged "degraded". Between 300s and 1800s we mark "breach" and page an operator. Beyond 1800s the source is "stale" and we typically pause comparisons against it until it recovers.
These thresholds are measured on the book as a whole: they say how long we can go without any new price from PMU, and not how old the price on one particular game is. A source that walks its slate one event at a time refreshes any single market less often than the tier above. Measured on production 2026-07-31, Fliff wrote a price every 5.3s at the median book-wide while its median WNBA game went 169s between writes, both healthy. For the age of a specific price, read the last_update field on that row rather than this tier.
Full per-book threshold table at /v1/meta/per-book-sla. Live operator dashboard at /uptime. Historical hourly observation rate at /v1/historical/source-quality.csv?source=pmu.
← All sportsbooks · Cross-book coverage matrix · Operator uptime · Machine-readable parser coverage