TVL anomaly — % drop in <1h
Concrete's assessment for RD-F-098 — scored yellow on the v1.7.0 rubric. The evidence below is the curator's reasoning for this score.
Evidence summary #
TVL anomaly signal: T-09 v1 launch signal, fully applicable. Current TVL ~$991M (DefiLlama); 1d change +0.51%; 30d change -8.4% (gradual market-correlated). Threshold for firing: 30% drop in 1h vs 30d baseline — not breached today. However 90d CoV = 0.128 (mean $963.8M, std $123.8M per data cache) is elevated above what a stable mature protocol would show. Historical TVL series shows large step-changes (e.g., June 2025: ~$483M → ~$50M over a few days; September 2025: ~$118M → ~$949M) which appear to reflect vault-routing migrations or data artefacts rather than exploits, but would trigger false positives in a live F098 system without a pre-announcement allowlist. Score yellow: signal would not fire today but elevated CoV and structural step-change history indicate meaningful false-positive operational risk if this signal were live without protocol-specific change-management allowlists.
Sources #
- InternalT-09 RD-F-098 detection threshold specificationT-09 signals spec §4.1 RD-F-098 — threshold: TVL_now / TVL_baseline_30d < 0.70 over 60-min windowretrieved 2026-05-17
- Concrete data cache 90d TVL coefficient of variationdata cache §sources.defillama.tvl_cov_90d: cov=0.128429, mean=963825226, std=123782672, window_start=1771372800, window_end=1779024515, sample_count=90retrieved 2026-05-17
- DefiLlama API — Concrete TVL current and trendDefiLlama API tvl_usd 991139027, tvl_1d_change_pct +0.51, tvl_30d_change_pct -8.4retrieved 2026-05-17
Methodology #
Detect whether TVL drops >X% within 1 hour versus the trailing 30-day baseline (X configurable per protocol TVL tier).
See the full factor methodology and distribution across all protocols →