Liquidity Score Changelog
Full version history of Liquidity Score methodology decisions, from v1.0 to v6.98.
What this controls
Liquidity Score Changelog version history
Use this changelog to trace how Pharos changed the related scoring, data-source, threshold, or interpretation rules over time. Entries record the contract active at each release; archived rules are retained so historical charts and citations can be read against the rules active at the time.
Primary references
Latest version
Aero venue coverage after the Aerodrome/Velodrome merger
Aero, the 2026-10-22 merger of Aerodrome and Velodrome, is now read directly from its factories on Base, Ethereum, OP Mainnet, Arbitrum, Robinhood Chain and Arc, and its provider listings receive the same venue-quality tiers as Aerodrome instead of the generic 0.3x fallback. Component weights and legacy Aerodrome/Velodrome handling are unchanged; no budget constant changed, but the two new sources add serial RPC reads inside the existing direct-API phase.
Impact snapshot
- Aero Slipstream (concentrated liquidity) pools use the Aerodrome Slipstream tiers: 1.1x at a 1 bp base fee, 0.85x at 5 bp and 0.4x above. The tier is the tick spacing's base fee from the factory, not the hook-set dynamic per-swap fee. Aero basic pools score 0.85x when created by the stable-pool factory and 0.4x when created by the volatile-pool factory; DefiLlama `aero-basic` rows, which carry no stable label, stay at the volatile tier as Aerodrome classic rows do. GeckoTerminal `aero-slipstream-<chain>` rows use the 0.7x Slipstream weight.
- DefiLlama `aero-slipstream`/`aero-basic`, GeckoTerminal `aero-slipstream-<chain>` and DexScreener `aero` identifiers now group as one `aero` protocol, so the same physical pool deduplicates across direct, staged and provider rows instead of counting under unrelated generic protocol names.
- Two direct sources enumerate every Aero factory pool holding a tracked stablecoin. Concentrated-liquidity pools must prove factory binding and exact getPool membership at one pinned block; basic pools must prove factory, metadata decimals and getPool membership. Volatile basic pools derive an untracked side from the constant-product reserve ratio; stable-curve pools are admitted only when both sides are independently priced. Both sources are bounded samples and cannot veto staged pools. Ink is not read directly because the reviewed Worker RPC profile has no Ink endpoint.
- Aero rows publish no trailing volume; their exact pool ids join the existing CoinGecko onchain stale-pool refresh on chains CoinGecko covers (Arc rows stay without volume). Unreadable pool state, an oversized factory or the per-run read budget degrades the named chain instead of reading as clean absence. Direct Aero rows carry USD values only for untracked sides, so tracked reference prices never re-emit as Aero price evidence, and DefiLlama `aero-basic` rows share one basic-pool identity family with the direct stable and volatile rows so the same physical pool dedupes when the match is unique.
v6.98Oct 23, 2026Aero venue coverage after the Aerodrome/Velodrome merger
Aero, the 2026-10-22 merger of Aerodrome and Velodrome, is now read directly from its factories on Base, Ethereum, OP Mainnet, Arbitrum, Robinhood Chain and Arc, and its provider listings receive the same venue-quality tiers as Aerodrome instead of the generic 0.3x fallback. Component weights and legacy Aerodrome/Velodrome handling are unchanged; no budget constant changed, but the two new sources add serial RPC reads inside the existing direct-API phase.
- Aero Slipstream (concentrated liquidity) pools use the Aerodrome Slipstream tiers: 1.1x at a 1 bp base fee, 0.85x at 5 bp and 0.4x above. The tier is the tick spacing's base fee from the factory, not the hook-set dynamic per-swap fee. Aero basic pools score 0.85x when created by the stable-pool factory and 0.4x when created by the volatile-pool factory; DefiLlama `aero-basic` rows, which carry no stable label, stay at the volatile tier as Aerodrome classic rows do. GeckoTerminal `aero-slipstream-<chain>` rows use the 0.7x Slipstream weight.
- DefiLlama `aero-slipstream`/`aero-basic`, GeckoTerminal `aero-slipstream-<chain>` and DexScreener `aero` identifiers now group as one `aero` protocol, so the same physical pool deduplicates across direct, staged and provider rows instead of counting under unrelated generic protocol names.
- Two direct sources enumerate every Aero factory pool holding a tracked stablecoin. Concentrated-liquidity pools must prove factory binding and exact getPool membership at one pinned block; basic pools must prove factory, metadata decimals and getPool membership. Volatile basic pools derive an untracked side from the constant-product reserve ratio; stable-curve pools are admitted only when both sides are independently priced. Both sources are bounded samples and cannot veto staged pools. Ink is not read directly because the reviewed Worker RPC profile has no Ink endpoint.
- Aero rows publish no trailing volume; their exact pool ids join the existing CoinGecko onchain stale-pool refresh on chains CoinGecko covers (Arc rows stay without volume). Unreadable pool state, an oversized factory or the per-run read budget degrades the named chain instead of reading as clean absence. Direct Aero rows carry USD values only for untracked sides, so tracked reference prices never re-emit as Aero price evidence, and DefiLlama `aero-basic` rows share one basic-pool identity family with the direct stable and volatile rows so the same physical pool dedupes when the match is unique.
DetailsHide details
Aero venue coverage after the Aerodrome/Velodrome merger
Aero, the 2026-10-22 merger of Aerodrome and Velodrome, is now read directly from its factories on Base, Ethereum, OP Mainnet, Arbitrum, Robinhood Chain and Arc, and its provider listings receive the same venue-quality tiers as Aerodrome instead of the generic 0.3x fallback. Component weights and legacy Aerodrome/Velodrome handling are unchanged; no budget constant changed, but the two new sources add serial RPC reads inside the existing direct-API phase.
- Aero Slipstream (concentrated liquidity) pools use the Aerodrome Slipstream tiers: 1.1x at a 1 bp base fee, 0.85x at 5 bp and 0.4x above. The tier is the tick spacing's base fee from the factory, not the hook-set dynamic per-swap fee. Aero basic pools score 0.85x when created by the stable-pool factory and 0.4x when created by the volatile-pool factory; DefiLlama `aero-basic` rows, which carry no stable label, stay at the volatile tier as Aerodrome classic rows do. GeckoTerminal `aero-slipstream-<chain>` rows use the 0.7x Slipstream weight.
- DefiLlama `aero-slipstream`/`aero-basic`, GeckoTerminal `aero-slipstream-<chain>` and DexScreener `aero` identifiers now group as one `aero` protocol, so the same physical pool deduplicates across direct, staged and provider rows instead of counting under unrelated generic protocol names.
- Two direct sources enumerate every Aero factory pool holding a tracked stablecoin. Concentrated-liquidity pools must prove factory binding and exact getPool membership at one pinned block; basic pools must prove factory, metadata decimals and getPool membership. Volatile basic pools derive an untracked side from the constant-product reserve ratio; stable-curve pools are admitted only when both sides are independently priced. Both sources are bounded samples and cannot veto staged pools. Ink is not read directly because the reviewed Worker RPC profile has no Ink endpoint.
- Aero rows publish no trailing volume; their exact pool ids join the existing CoinGecko onchain stale-pool refresh on chains CoinGecko covers (Arc rows stay without volume). Unreadable pool state, an oversized factory or the per-run read budget degrades the named chain instead of reading as clean absence. Direct Aero rows carry USD values only for untracked sides, so tracked reference prices never re-emit as Aero price evidence, and DefiLlama `aero-basic` rows share one basic-pool identity family with the direct stable and volatile rows so the same physical pool dedupes when the match is unique.
Impact notes
- Aero Slipstream (concentrated liquidity) pools use the Aerodrome Slipstream tiers: 1.1x at a 1 bp base fee, 0.85x at 5 bp and 0.4x above. The tier is the tick spacing's base fee from the factory, not the hook-set dynamic per-swap fee. Aero basic pools score 0.85x when created by the stable-pool factory and 0.4x when created by the volatile-pool factory; DefiLlama `aero-basic` rows, which carry no stable label, stay at the volatile tier as Aerodrome classic rows do. GeckoTerminal `aero-slipstream-<chain>` rows use the 0.7x Slipstream weight.
- DefiLlama `aero-slipstream`/`aero-basic`, GeckoTerminal `aero-slipstream-<chain>` and DexScreener `aero` identifiers now group as one `aero` protocol, so the same physical pool deduplicates across direct, staged and provider rows instead of counting under unrelated generic protocol names.
- Two direct sources enumerate every Aero factory pool holding a tracked stablecoin. Concentrated-liquidity pools must prove factory binding and exact getPool membership at one pinned block; basic pools must prove factory, metadata decimals and getPool membership. Volatile basic pools derive an untracked side from the constant-product reserve ratio; stable-curve pools are admitted only when both sides are independently priced. Both sources are bounded samples and cannot veto staged pools. Ink is not read directly because the reviewed Worker RPC profile has no Ink endpoint.
- Aero rows publish no trailing volume; their exact pool ids join the existing CoinGecko onchain stale-pool refresh on chains CoinGecko covers (Arc rows stay without volume). Unreadable pool state, an oversized factory or the per-run read budget degrades the named chain instead of reading as clean absence. Direct Aero rows carry USD values only for untracked sides, so tracked reference prices never re-emit as Aero price evidence, and DefiLlama `aero-basic` rows share one basic-pool identity family with the direct stable and volatile rows so the same physical pool dedupes when the match is unique.
- Grouping provider ids changes the retained pool population (a DefiLlama row and a DexScreener row of one Aero pool, previously counted under different protocol names, now dedupe), so the TVL-basis break list appends 6.98. Trend consumers compare stored measurement epochs rather than reading the step as a market move.
- Legacy Aerodrome/Velodrome Sugar sources stay active while their liquidity drains.
v6.97Oct 20, 2026Exact pool populations and source-bound measured evidence
Liquidity preserves distinct provider deployments, quarantines conflicting or unsupported fallback observations without clipping trusted TVL, and requires complete balance, valuation and execution provenance. Component weights and request budgets are unchanged.
- Token pairs are matching candidates, not pool identities. Chain-scoped opaque provider deployments remain distinct; exact repeats count once and conflicting exact observations are quarantined before attribution. One proven physical pool counts once globally across assets.
- External protocol totals are fallback admission evidence, never published valuations. Fallback groups above per-coin headroom are rejected before price admission, with physical rejection propagated across coin views. Trusted observed raw TVL is not scaled; headline, breakdown and volume statistics use one admitted graph. Excluded pool/protocol counts and raw TVL replace cap reductions.
- Provider values must be complete finite decimals rather than numeric prefixes. Orca token decimals must be integers in 0..255 for balance normalization. Curve requires complete USD balances and Fluid both reserve halves; missing evidence is unmeasured with neutral balance weighting, while complete zero-sided balances remain measured zero.
- Curve factory models require independent tracked-market references; missing references withhold exact capture. Bounded UniV3 identity carry does not supply current output USD valuation; independent references are still required.
DetailsHide details
Exact pool populations and source-bound measured evidence
Liquidity preserves distinct provider deployments, quarantines conflicting or unsupported fallback observations without clipping trusted TVL, and requires complete balance, valuation and execution provenance. Component weights and request budgets are unchanged.
- Token pairs are matching candidates, not pool identities. Chain-scoped opaque provider deployments remain distinct; exact repeats count once and conflicting exact observations are quarantined before attribution. One proven physical pool counts once globally across assets.
- External protocol totals are fallback admission evidence, never published valuations. Fallback groups above per-coin headroom are rejected before price admission, with physical rejection propagated across coin views. Trusted observed raw TVL is not scaled; headline, breakdown and volume statistics use one admitted graph. Excluded pool/protocol counts and raw TVL replace cap reductions.
- Provider values must be complete finite decimals rather than numeric prefixes. Orca token decimals must be integers in 0..255 for balance normalization. Curve requires complete USD balances and Fluid both reserve halves; missing evidence is unmeasured with neutral balance weighting, while complete zero-sided balances remain measured zero.
- Curve factory models require independent tracked-market references; missing references withhold exact capture. Bounded UniV3 identity carry does not supply current output USD valuation; independent references are still required.
Impact notes
- Token pairs are matching candidates, not pool identities. Chain-scoped opaque provider deployments remain distinct; exact repeats count once and conflicting exact observations are quarantined before attribution. One proven physical pool counts once globally across assets.
- External protocol totals are fallback admission evidence, never published valuations. Fallback groups above per-coin headroom are rejected before price admission, with physical rejection propagated across coin views. Trusted observed raw TVL is not scaled; headline, breakdown and volume statistics use one admitted graph. Excluded pool/protocol counts and raw TVL replace cap reductions.
- Provider values must be complete finite decimals rather than numeric prefixes. Orca token decimals must be integers in 0..255 for balance normalization. Curve requires complete USD balances and Fluid both reserve halves; missing evidence is unmeasured with neutral balance weighting, while complete zero-sided balances remain measured zero.
- Curve factory models require independent tracked-market references; missing references withhold exact capture. Bounded UniV3 identity carry does not supply current output USD valuation; independent references are still required.
- Measured-history constituents age at consumer assessment. Exact canonical hashes bind EVM proof and execution; unavailable probes do not prove method absence. Non-USD peg fallbacks require fresh non-future per-peg FX under original source cadence, reassessed at capture. These admission repairs do not alter aggregate TVL or volume by themselves.
- The retained-population policy adds TVL-basis break 6.97. Paired source-stage payload v6 rejects incompatible older identity/admission stages. Historical rows retain their original bytes and measurement epoch; policy steps are not market withdrawals.
- Planned release date is 2026-10-20. Fresh producer arithmetic, target admission and downstream acceptance remain distinct from fixed-input Safety replay.
Earlier versions
Compact summaries retain historical citation targets. Open a version archive for its full evidence and commit provenance.
- v6.96 · One retained population for global TVL and volume coverage
Oct 19, 2026
Global DEX headline TVL, protocol and chain breakdowns, and volume-coverage TVL denominators now describe the same retained cross-coin deduped pools. The second global protocol clamp is removed; existing per-pool fraud exclusions, source admission and per-coin secondary-source caps remain unchanged.
- v6.95 · Reviewed retained-pool targets and missing-row discovery
Oct 14, 2026
Liquidity v6.95 expands the active reviewed exact-target cohort on existing venue families and repairs missing-row discovery scheduling and CHFSAFO Aquarius identity. Component weights, request budgets, cadence, connection limits and retained-TVL measurement semantics are unchanged; no TVL-basis break is added.
- v6.94 · Scope-matched pool quality and source-timed execution evidence
Oct 10, 2026
Curve metapool quality now uses the same base-pool-excluded TVL throughout, and balance measurements count only when they describe that retained scope. Exact EVM execution evidence must remain fresh on its original source clock. Pool admission also preserves independent valid observations instead of mistaking malformed or incomplete censuses for verified absence.
- v6.93 · Coherent registry evaluation and price-only CoinGecko tickers
Oct 8, 2026
Registry observations are evaluated whole against a clock captured when the registry read completes, closing the one-generation dead-pool-floor bypass for post-start refreshes. CoinGecko Tickers no longer supplies volume-derived synthetic liquidity, including legacy stored rows; it remains independent price evidence under the separately versioned Pricing policy.
- v6.92 · Dead-pool floor for zero-trade pools against untracked tokens
Sep 28, 2026
A retained pool with at least $1M of scoring TVL (DEX_DEAD_POOL_TVL_MIN_USD) no longer counts toward liquidity or DEX-implied prices when its admitted 24h reading is a trade-verified zero and none of its other tokens is a tracked stablecoin deployment on that chain. These pools are single-sided: an operator seeds a few dollars of a tracked stablecoin next to billions of an unlisted token, the seed fixes the pool price near $1, and the provider values the unlisted side at that price, so the reported reserve is not executable depth. A pool whose volume is missing, stale or not trade-verified is never screened, and a screened pool counts again on the first day CoinGecko reports a trade in it.
- v6.91 · NEAR Intents excluded as a non-AMM venue
Sep 28, 2026
NEAR Intents (`near-intents`) joins the blocked DEX ids. It is an intent-settlement verifier, not an AMM: GeckoTerminal and CoinGecko Onchain price every NEAR Intents pair from the `intents.near` contract's shared custody of the quote asset, so the reported pool TVL is not executable depth for the stablecoin. Its rows no longer count toward coin or global DEX liquidity, pool counts, exit routes, challenger snapshots or DEX-implied prices.
- v6.9 · Admitted-only DEX volume with a TVL coverage floor for Volume Activity
Sep 28, 2026
DEX volume now counts only admitted pool readings (observed at most 72 hours ago), never decayed, zero-filled or imputed. Volume Activity is admitted volume over admitted TVL, and the LiquidityScore is rated only when admitted pools cover at least 50% of the coin's retained scoring TVL; below that floor it is NR (DEC-19 as amended).
- v6.8 · Slipstream 100bp multiplier, PancakeSwap bounded-sample census, and cross-source price handling
Sep 23, 2026
Three DEX-liquidity corrections shipped together: the reviewed 100bp Slipstream fee tier regains its documented 0.4x quality multiplier, the PancakeSwap rotating capture loses its staged-pool veto authority, and an implausible cross-source price drops only the price instead of erasing the whole staged pool.
- v6.7 · Pair-price coherence gate at pool admission
Sep 23, 2026
GeckoTerminal and CoinGecko Onchain pool rows are admitted only when the tracked leg's USD price stays coherent with the pool's own pair ratio and the counter-leg's USD price; provider rows carrying the broken-price signature (leg USD prices published with null/zero pair-ratio inputs) are rejected before staging instead of contributing bogus TVL and price evidence.
- v6.6 · Recorded concentration band thresholds
Sep 21, 2026
The DEX-liquidity card's `Concentration` verdict and the exit-route crowding bands now share one recorded threshold table: High at HHI >= 0.35, Medium at HHI >= 0.18, Low below. A non-finite HHI resolves to the broadest band instead of throwing at render time. The HHI computation itself is unchanged.
- v6.5 · Per-source pool registry
Sep 18, 2026
Staged pool memory moves from `dex_pool_staging` — one row per pool per coin carrying a single source, with lanes contending for that row — to `dex_pool_registry` keyed by (stablecoin, pool, source): every lane records its own observation, and the merge resolves one view per pool before scoring — value from the highest-trust observation refreshed within 24 hours (else the freshest observation of any source), family from the value's source, price from the highest-trust priced observation refreshed within 24 hours, and identity metadata (token pair) from any observation within the 14-day horizon, so derived dedupe is lane-symmetric; ownership-by-subtraction is removed.
- v6.4 · 14-day staged pool memory
Sep 10, 2026
Staged discovery pools are now remembered for 14 days instead of 24 hours — full confidence through the first day, then linear decay to zero at 336 hours — while price observations remain pinned to rows refreshed within 24 hours.
- v6.3 · Curve physical-pool alias normalization
Sep 1, 2026
Curve registry aliases for one physical pool now collapse by canonical chain and address before coin-set ambiguity is evaluated, allowing the reviewed LUSD/3Crv execution target to enter measurement without weakening collision handling.
- v6.2 · Exact LUSD/3Crv metapool execution
Sep 1, 2026
The reviewed LUSD/3Crv deployment now uses the existing exact Curve metapool adapter, replacing its unresolved execution gate with pinned on-chain get_dy_underlying measurements.
- v6.1 · Family-scoped authoritative confirmation for classic v2 pools
Aug 21, 2026
Authoritative staged-pool confirmation now follows the exact protocol family enumerated by each native source, so Slipstream-only inventories no longer veto classic Aerodrome or Velodrome v2 pools outside their coverage.
- v6.0 · Raydium double-count correction and native measured-lane retirement
Aug 20, 2026
DefiLlama Raydium pools are now classified from their pool metadata, collapsing the DefiLlama/direct-API double count of the same physical Solana CLMM pool, and the never-score-eligible Solana and Tron native measured-execution lanes plus the Fluid measured overlay are retired.
- v5.993 · Activated exact hook-free Ethereum Uniswap V4 exits
Aug 13, 2026
Reviewed hook-free Ethereum Uniswap V4 PoolKeys can now enter exact DEX exit-route scoring after three productive production shadow generations and 129 successful quotes across the latest 130-direction cohort.
- v5.992 · Aligned measured-quote freshness with the two-hour publication cadence
Aug 12, 2026
All measured DEX exit-route adapters now share the two-hour freshness ceiling the reviewed Curve StableSwap adapters already used, matching the two-hour score-bearing publication cadence introduced in 5.991.
- v5.991 · Separated DEX price, score, and shadow cadences
Aug 10, 2026
DEX prices now refresh hourly, the standalone Liquidity Score publishes every two hours, active exact-route quotes remain half-hourly, and score-ineligible shadow evidence moves to a daily diagnostic run.
- v5.99 · Stellar classic-AMM discovery
Aug 8, 2026
Stellar Horizon now supplies bounded, case-preserving classic-AMM discovery for active Stellar deployments, while Soroban contract-token identities remain explicitly unavailable.
- v5.98 · Trust-filtered DEX primary-price handoff
Jul 31, 2026
The DEX scoring consumer now treats the staged primary-price map as an authoritative trust-filtered input instead of backfilling omitted assets from the broader stablecoin cache.
- v5.97 · Retired Optimism Uniswap V3 shadow lane
Jul 29, 2026
The Optimism Uniswap V3 subgraph and QuoterV2 measured-execution lane are removed from the maintained DEX source roster after the owner accepted clean retirement.
- v5.96 · Capacity-dominant bounded route selection
Jul 28, 2026
The private executable-route candidate set is now evaluated at the $25M/200 bps V9 stress point before ten public observations are selected, with diversity quotas and a still-fresh asset-level churn hold.
- v5.95 · Pinned SunSwap V2 direct-route fallback
Jul 28, 2026
A canonical SunSwap V2 pool can retain exact direct-execution evidence when SUN's three best Smart Router paths are all multi-hop, but only through the reviewed on-chain V2 Router deployment.
- v5.94 · Pinned Curve StableSwap-NG rate state
Jul 27, 2026
Address-grade plain Curve StableSwap-NG pools with rate-bearing inputs can contribute a reserve model only after a fresh, same-block state capture proves the balances, amplification, stored rates, static fee state, and ordered coin layout.
- v5.93 · Paused v5.92 exact-route activations
Jul 27, 2026
Optimism Uniswap V3 and the reviewed Solana wM/USDC Raydium direction return to shadow-only collection after the first post-activation scoring consumers exceeded the Worker memory limit.
- v5.92 · Pinned Raydium CLMM on-state execution
Jul 27, 2026
The reviewed Solana wM/USDC Raydium CLMM direction can contribute exact route evidence only after its captured current liquidity segment reproduces the strict direct quote.
- v5.91 · Guarded SunSwap V2 reactivation
Jul 27, 2026
Reviewed Tron SunSwap V2 execution profiles are score-eligible again after current complete 21/21 shadow generations and healthy split scoring consumers.
- v5.90 · Paused SunSwap V2 scoring after rollout
Jul 25, 2026
SunSwap V2 exact-execution collection remains live, but its profiles return to shadow-only status after the first two active DEX-scoring invocations exceeded the Worker memory limit.
- v5.89 · Reviewed SunSwap V2 direct execution
Jul 25, 2026
Current direct SunSwap V2 routes can contribute factory-bound, reserve-reproduced execution profiles to the Safety Score V9 exact-route envelope after full-generation production validation.
- v5.88 · Exact-route continuity and Base Slipstream execution
Jul 24, 2026
Reviewed Base Aerodrome Slipstream pools can contribute factory-bound QuoterV2 execution profiles, while mature fresh routes remain available to the Safety Score V9 exact-route envelope across bounded pool-shortlist rotation.
- v5.87 · Base Aerodrome volatile exact execution
Jul 22, 2026
Classic volatile pools from the existing Base Aerodrome census can publish exact constant-product execution models after deployment-specific same-block verification.
- v5.86 · Canonical V2 pool retention and exact execution
Jul 22, 2026
PancakeSwap V2 discovery is no longer rejected by the V3 authoritative inventory, and reviewed Ethereum Uniswap V2 and BSC PancakeSwap V2 pools can publish factory-verified constant-product execution models.
- v5.85 · Raydium pool-implied counter-asset reference
Jul 20, 2026
Raydium standard AMM pools now complete exact reserve capture when the untracked counter asset's USD reference is implied from the pool's own captured spot price instead of gating to incomplete-exact-capture.
- v5.84 · Composite Curve LP pair-quality normalization
Jun 29, 2026
Composite Curve LP quote tokens such as 3Crv and FRAXBP now inherit the best pair-quality score of their underlying stablecoin basket instead of falling back to the unknown-token haircut.
- v5.83 · Top-asset recovery guard quality baseline
Jun 19, 2026
The top-asset coverage guard now discounts previously published rows whose raw TVL was dominated by near-zero effective liquidity before deciding whether a recovery run must fail hard.
- v5.82 · Large zero-volume pool retention hardening
Jun 19, 2026
Large retained pools must clear the minimum 24-hour volume floor even when a source marks volume as unmeasured.
- v5.81 · Unsupported-chain Curve fallback coverage
Jun 14, 2026
Fallback pool discovery can now retain Curve pools on chains not covered by the native Curve API, while still skipping fallback Curve rows where native Curve enrichment already owns the chain.
- v5.8 · Retained-pool DEX price ownership hardening
Jun 6, 2026
DEX implied-price publication now reapplies the documented $50K retained-pool price floor and weights medians by source family instead of protocol labels.
- v5.7 · Peg-aware staged discovery price gate
May 20, 2026
Secondary discovery pools now need a plausible tracked-token price before their TVL can be staged or merged into Liquidity Score aggregates.
- v5.6 · Staged discovery TVL sanity ceiling
May 7, 2026
Secondary discovery rows now reject impossible pool TVL before staging and skip any already-staged over-cap row during the scoring merge.
- v5.5 · Absolute TVL Depth fallback recalibration and Slipstream sqrt_ratio price
Apr 17, 2026
Absolute TVL Depth fallback (used when circulatingUsd is unavailable) now shares the ratio formula's anchor via a $1B implied reference mcap. Aerodrome/Velodrome Slipstream price derivation now uses on-chain sqrt_ratio (Q64.96) instead of total-reserve ratios for concentrated liquidity pools.
- v5.4 · Curve enrichment scoping and staged UUID dedupe
Apr 14, 2026
Curve API enrichment is now scoped to Curve DeFiLlama rows, and staged exact pool-id rows can dedupe against a single identity-poor DeFiLlama UUID row through the same narrow optional-metadata wildcard used by primary dedupe.
- v5.3 · PancakeSwap trailing-hour volume window
Apr 8, 2026
PancakeSwap V3 direct volume now sums the official `poolHourDatas.volumeUSD` buckets across a bounded trailing 24-hour window instead of treating the latest `poolDayDatas` row as if it were a rolling 24h metric.
- v5.2 · Orderbook ticker contract refresh and Balancer exact-address identity
Apr 8, 2026
CoinGecko orderbook fallback now ignores the deprecated `trust_score` field and validates tickers by observable freshness/price/volume fields, while Balancer direct pools now use the API's exact pool address for identity instead of the 32-byte vault pool id.
- v5.1 · Authoritative protocol confirmation for staged discovery
Apr 7, 2026
Staged discovery rows can no longer invent new pools inside protocol families that already have a clean protocol-native direct source. When that authoritative fetch succeeds on a supported chain, staged rows must match one of its exact pool ids or they are excluded.
- v5.0 · Size-aware scoring: relative TVL depth, recalibrated volume, quality retention
Apr 5, 2026
All scoring dimensions are now size-independent. TVL Depth measures effective TVL relative to market cap instead of absolute dollar value. Volume Activity has a recalibrated curve with a realistic ceiling (tops out at ~32% V/T instead of ~500%). Pool Quality measures venue quality retention ratio (qualityAdjustedTvl / totalTvl, rescaled) instead of absolute quality-adjusted TVL. Weights rebalanced to 30/20/20/20/10.
- v4.9 · Blocked dead Bunni DEX inputs
Apr 3, 2026
Explicitly blocked Bunni from liquidity scoring and DEX implied-price publication after dead-venue rows kept surfacing as retained liquidity.
- v4.8 · Direct-source duplicate hardening for Balancer and staged exact ids
Apr 3, 2026
Direct-source dedupe now reserves every authoritative exact pool id for later staged matching, and Balancer stablecoin pools can still collapse against direct Balancer coverage when DefiLlama omits the subtype in `balancer-v3` metadata.
- v4.7 · Retained-pool DEX price publication
Apr 3, 2026
DEX implied-price publication now derives from the final retained pool surface after dedupe, caps, and scoring filters, instead of from the earlier raw observation stream.
- v4.6 · Protocol-native DEX coverage expansion
Mar 24, 2026
Liquidity scoring now ingests Meteora DLMM, PancakeSwap V3, and Aerodrome/Velodrome Slipstream pool-state data as protocol-native direct sources, expanding primary-grade coverage across Solana, BSC, Base, and Optimism.
- v4.5 · Coverage recall hardening and measurement-aware confidence
Mar 24, 2026
DEX liquidity now paginates deeper through GeckoTerminal and CoinGecko Onchain discovery, enriches weak partial coverage instead of only zero-coverage rows, and scores coverage confidence from measured-vs-synthetic retained TVL rather than a fixed source-family ladder.
- v4.4 · Chain-aware pool identity dedupe and challenger snapshot publishing
Mar 19, 2026
DEX liquidity now resolves tracked tokens chain-aware, deduplicates pools with conservative identity keys instead of coarse fingerprints, collapses duplicate DEX price observations before aggregation, and publishes dedicated challenger snapshots from the full retained pool set.
- v4.3 · Fluid DexReservesResolver balance integration
Mar 18, 2026
Fluid pools on Ethereum, Arbitrum, Base, and Polygon now read balances and fee detail from the official DexReservesResolver instead of staying on neutral placeholders.
- v4.2 · Measured direct-API balance health and normalized pool-detail metadata
Mar 18, 2026
Balancer, Raydium, and Orca direct-API pools now preserve measured balance and fee metadata through scoring instead of merging with neutral placeholders. Pool-detail fee tiers are normalized to basis points for all sources.
- v4.1 · Direct API precedence, primary-grade coverage, and fetcher contract hardening
Mar 18, 2026
Direct API sources now replace overlapping DeFiLlama pools before scoring, run ahead of staged/fallback sources, and count as primary-grade coverage. Raydium and Orca contract handling was hardened against live API drift, Fluid volume normalization moved to one-sided USD volume, and Balancer intake now excludes unsupported pool types.
- v4.0 · Log-scale volume, cross-chain removal, durability rebalance
Mar 10, 2026
Volume activity switched from linear to log-scale. Cross-chain component removed and weight redistributed to TVL depth and pool quality. Durability sub-weights rebalanced: locked liquidity removed, organic fraction reduced to 15% with sqrt curve, history-measured signals raised to 85%.
- v3.4 · Retained-pool score recomputation and trusted staged-price hardening
Mar 9, 2026
Liquidity scoring now rebuilds every aggregate from the retained pool set after filtering/caps, while staged discovery preserves pool-quality metadata and stricter DEX-price trust rules.
- v3.3 · Separated discovery pipeline with staged pool confidence decay
Mar 9, 2026
Discovery sources (CG Onchain, GeckoTerminal, DexScreener, CG Tickers) now run on an independent 20-minute cron with 3x more budget. Staged pools merged into scoring with freshness confidence decay and explicit defaults contract.
- v3.2 · Effective TVL symbol-fallback inflation fix
Mar 2, 2026
Corrected effective TVL inflation when symbol fallback matched non-Curve pools to Curve entries.
- v3.1 · Anti-duplication and protocol TVL cap normalization
Feb 28, 2026
Introduced fingerprint-based deduplication and DeFiLlama-anchored cap logic to suppress inflated secondary-source TVLs.
- v3.0 · Coverage expansion with fallback sources
Feb 28, 2026
Expanded zero-pool recovery with DexScreener and CoinGecko tickers fallbacks for orderbook-heavy assets.
- v2.2 · No-pool rows moved to NR semantics
Feb 27, 2026
Coins without DEX pools switched from score=0 placeholders to NULL (NR) semantics.
- v2.1 · Onchain source upgrade and locked-liquidity durability term
Feb 25, 2026
Primary pool discovery moved to CoinGecko Onchain with locked-liquidity data integrated into durability scoring.
- v2.0 · Six-component v2 liquidity model
Feb 19, 2026
Moved from a five-component heuristic to a six-component model with effective TVL and durability decomposition.
- v1.0 · Initial DEX liquidity score release
Feb 19, 2026
Launched baseline DEX liquidity scoring, API surface, and dashboard integration.