Coverage & gaps
What’s indexed, right now
A visible gap is credible; a guess presented as coverage is not. This is the exact pipeline every launchpad below moves through from onboarding, to the one-time backfill, to the ongoing cronjob that keeps every score current.
From onboarding to a live score
Name, slug, deployer address recorded. Docs checked for a Bitquery/Mobula entry.
Backfilled once: everything under 100 launches, else 20% of the upstream total, capped at 1,000 most-recent-first.
5 weighted dimensions, clamped 0–100, gated by sample size before a star rating is shown.
Hourly ingestion rotation + a daily scoring sweep no manual step re-runs this.
Cronjob status
Collectors + Normaliser: refreshes metrics for tracked launches, polls the <24h new-launch fast path.
Scorer recomputes launchpad_scores for the day; drift checks run alongside it.
Pipeline architecture
One adapter per source Dexscreener, Blockscout, Bitquery/Mobula for documented launchpads, RPC watcher as fallback. Fetch, tag with data_source, push to queue.
Maps each source's payload into the canonical snapshot row, resolves disagreements between sources, writes it, then triggers a recompute.
Pure function over snapshot history graduation status, wash-trading flag, holder concentration. Does not decide a score.
Deterministic composite formula across 5 dimensions. Reproducible: same inputs and algorithm_version, same output, every time.
Stateless, reads only from Postgres. Stays up even if every collector is down serves stale-but-labeled data instead.
Active collectors
Price, 24h volume, liquidity
metricsContract verification, holder distribution, transfers
verificationDocumented factory addresses, decoded launch events, live trades
discoveryPrice history (OHLCV) → peak multiple, plus liquidity, volume and bonding status for tokens still on a curve
metricsSelf-indexed deploy/graduation events the fallback with no third party in the loop
fallbackIndexed launchpads & sources
| Launchpad | Status | Discovery source | Sample / upstream total | Confidence floor | Last update | Next update |
|---|---|---|---|---|---|---|
| Pons | Indexed | RPC self-indexed | 936 / 287,738 | high | Sep 26, 2026 | — |
| Bags.fm | Indexed | RPC self-indexed | 876 / 4,378 | high | Sep 26, 2026 | — |
| Klik Finance | Indexed | RPC self-indexed | 814 / 10,131 | high | Sep 26, 2026 | — |
| LaunchHood | Indexed | RPC self-indexed | 600 / 4,167 | high | Sep 26, 2026 | — |
| Virtuals | Indexed | RPC self-indexed | 465 / 121,906 | high | Sep 26, 2026 | — |
| hood.fun | Indexed | RPC self-indexed | 265 / 10,783 | high | Sep 26, 2026 | — |
| Clanker | Indexed | RPC self-indexed | 146 / 14,721 | high | Sep 26, 2026 | — |
| Pools.trade | Indexed | RPC self-indexed | 120 / 599 | high | Sep 26, 2026 | — |
Confidence floor here is a coarse, display-only read on sample size not the scoring engine's binary confidence gate (see Methodology). A launchpad can be “high” coverage and still carry a low score.