Liquidity Score Changelog: v5 archive

Impact notes

• Activation is limited to exact hook-free PoolKeys on the pinned Ethereum PoolManager, StateView, and Quoter deployment; hooked pools and other V4 chains remain unsupported
• Every score-facing profile still requires current runtime hashes, immutable PoolManager bindings, pinned pool state, exact token and PoolId identity, and a fresh ABI-bound quote
• The existing 1,220-request admission ceiling, 80-request fragmentation reserve, 1,300-request hard ceiling, quote-subcall limit, memory containment, and shadow machinery are unchanged
• Fluid resolver, Solana native, and SunSwap V2 profiles remain shadow-only and score-ineligible

Impact notes

• Half-hourly Safety Score V9 preparations no longer derate measured DEX route confidence from high to medium on the runs that fall 60-120 minutes after an on-schedule liquidity publication, which had flipped several A-boundary grades on alternate score runs
• A missed liquidity publication still expires measured evidence: profiles older than two hours revert to medium confidence exactly as before
• Quote measurement, capacity curves, route selection, the Liquidity Score formula, and all other freshness gates are unchanged

Impact notes

• The Liquidity Score formula, source admission, retained-pool selection, protocol caps, and public row schema are unchanged
• Hourly price generations continue to use the exact retained pool graph and current published liquidity generation, while alternate hours avoid rewriting score, history, depth, and target rows
• Score-bearing EVM exact-route evidence retains its 30-minute quote cadence; score-ineligible EVM, Solana, and Tron evidence is isolated to a daily shadow lane and cannot affect scoring
• Synthetic budget deferrals are reconstructed from a count-and-digest-fenced generation manifest instead of being stored as one D1 row per target; fail-closed identity and last-known-good semantics are unchanged

Impact notes

• Classic Stellar CODE:ISSUER assets can contribute Horizon pool observations after exact reserve-identity validation; legacy bare-issuer metadata is combined with the tracked asset code without lowercasing
• A Horizon row receives price and TVL only when the counter-asset is another active tracked classic Stellar stablecoin with a usable peg reference and the implied tracked price passes the shared plausibility gate
• Horizon remains a capped secondary discovery family with 0.55 fallback-price confidence, the common retained-pool admission rules, and a bounded 200-row response rather than an exhaustive liquidity claim
• Soroban contract-token deployments fail closed as provider-degraded, and order-book or Soroban DEX liquidity remains outside this provider

Impact notes

• Assets intentionally omitted from the trust-filtered primary-price handoff can still publish their DEX-implied median and source aggregates, but primary price and primary-relative deviation fields remain null
• Untrusted or low-confidence primary references can no longer re-enter DEX price publication solely because an observed DEX asset was missing from the staged map
• DEX weighted medians, retained-pool price sources, aggregate liquidity, visible pool selection, target publication, V8 liquidity scoring, and Safety Score V9 route inputs are unchanged

Impact notes

• Uniswap V3 subgraph enrichment now covers Ethereum, Base, Arbitrum, and Polygon; Optimism no longer contributes source-stage UniV3 pools, price observations, or measured-execution targets
• The reviewed Optimism QuoterV2 deployment is no longer scheduled for evidence collection; stale persisted profiles fail deployment validation instead of remaining shadow evidence
• Other Optimism liquidity lanes, including Curve, Balancer, and Velodrome Slipstream, are unchanged
• Current score-eligible exact-route cohorts are unchanged: Ethereum, Polygon, and Arbitrum Uniswap V3; Base, BSC, and Ethereum PancakeSwap V3; and Base Aerodrome Slipstream

Impact notes

• The bounded payload guarantees the strongest executable-capacity route, the strongest exact reserve-model fallback, and an independent chain/protocol route when one exists
• Measured evidence no longer outranks a materially deeper exact route solely because its evidence tier is nominally stronger; per-chain, protocol, and adapter quotas prevent one common mode from occupying the route set
• If unexplained route-set churn cuts an asset's best stress capacity below half of a still-fresh prior set with at least $100K capacity, only that asset retains the prior route observations until confirmation or expiry
• SunSwap V2 returns to shadow-only collection while its production score impact is revalidated; aggregate liquidity, price consensus, visible pools, and V8 scoring remain unchanged

Impact notes

• The Smart Router calculation service remains the primary quote source; the fallback is eligible only when its successful response consists entirely of clean V2 multi-hop candidates, because the documented service returns only the three highest-output paths
• The fallback pins the documented SunSwap V2 Router address and runtime hash, proves its factory binding, and requires `getAmountsOut` for the exact two-token path to equal the independently read pair reserves under the reviewed 0.3% constant-product formula
• Any malformed, ambiguous, referral-bearing, output-mismatched, or otherwise invalid direct Smart Router candidate remains fail-closed and cannot be masked by the on-chain fallback
• The change closes the persistent USDT/PHONFT target gap without admitting multi-hop capacity or changing aggregate liquidity, price consensus, visible pools, target publication, or V8 liquidity scoring

Impact notes

• Only `factory-stable-ng` candidates with an exact EVM address, unique token identities, and one tracked input are eligible; legacy, metapool, CryptoSwap, unknown, stale, or unpinnable shapes remain capability-gated
• The capture rereads the pinned block header hash after bounded multicalls for `get_balances`, `stored_rates`, `A`, `fee`, `offpeg_fee_multiplier`, and `coins`; rate factors scale balances and inversely scale USD references, while Curve applies the captured static fee after the full-input invariant
• Missing, base-only, malformed, dynamic-fee, mismatched-order, or hash-drifted state clears the internal candidate and retains `curve-stableswap:rate-bearing-inputs` rather than using API balances, an unpinned amplification value, a guessed fee, or nominal parity
• This is route-only Safety Score V9 evidence: aggregate liquidity, price consensus, visible pools, target publication, and V8 liquidity scoring are unchanged

Impact notes

• Both cohorts are score-ineligible and retain an activation-pending capability gate while production scoring memory is remediated and revalidated
• Optimism deployment verification and quotes, plus the Raydium pool-account capture, direct quote, and exact single-segment replay, remain live as activation evidence
• Previously active Ethereum, Polygon, and Arbitrum Uniswap V3; Base, BSC, and Ethereum PancakeSwap V3; Base Aerodrome Slipstream; and Tron SunSwap V2 cohorts are unchanged
• No stale route, manual capacity, or reserve-derived substitute enters P4 while either reviewed cohort is paused

Impact notes

• Activation is limited to the case-sensitive wM-to-USDC direction in the reviewed Raydium CLMM pool; generic Raydium CLMM, Orca Whirlpool including the adaptive-fee HYUSD pool, Meteora, and all unlisted native targets remain shadow-only
• Each accepted wM profile binds the direct route, raw amounts, provider fee, pool owner and mint order, captured liquidity and sqrt price, post-swap sqrt price, bounded slot window, and exact single-segment integer replay; any mismatch, tick crossing, malformed state, route drift, or identity failure is capability-gated
• Only explicit operational collection failures may retain the newest still-fresh exact profile for the same target across bounded Solana rotation; semantic failures and malformed history are integrity barriers
• The route-only profile does not alter aggregate liquidity, price consensus, direct-source precedence, visible pool selection, target publication, or V8 liquidity scoring

Impact notes

• Activation is limited to the reviewed SunSwap V2 factory, pair runtime, and SUN Smart Router profile on Tron; Solana CLMM, Fluid, Uniswap V4, and unratified native cohorts remain shadow-only
• Every accepted profile must still prove the exact factory pair, token order, reserves, direct route, 0.3% constant-product output, bounded latest-state block bracket, freshness, and current target identity
• The first active scoring consumers remain a guarded production trial: memory termination, stale heartbeat, incomplete generations, or identity/proof drift require returning the registry to shadow while leaving evidence collection enabled
• SunSwap census rows remain excluded from liquidity scoring, price consensus, direct-source precedence, visible pool selection, and V8 liquidity scoring

Impact notes

• The SunSwap V2 adapter is score-ineligible and valid profiles retain the activation-pending capability gate while the scoring consumer's memory behavior is remediated and revalidated
• The complete 21-target producer remains enabled so direct-route reliability, identity, reserve, output, and block-bracket evidence continue to accumulate
• Missing, failed, stale, malformed, multi-hop-only, or identity-mismatched routes remain target-level capability gates; no stale proof or manual capacity substitutes for current direct evidence
• SunSwap census rows remain excluded from liquidity scoring, price consensus, direct-source precedence, visible pool selection, and V8 liquidity scoring

Impact notes

• Activation is limited to the reviewed SunSwap V2 factory, pair runtime, and SUN Smart Router profile on Tron; Raydium CLMM, Orca Whirlpool, and other native cohorts remain shadow-only
• Every accepted profile must prove the exact factory pair, token order, reserves, direct route, 0.3% constant-product output, bounded latest-state block bracket, freshness, and current target identity
• Missing, failed, stale, malformed, multi-hop-only, or identity-mismatched routes remain target-level capability gates; no stale proof or manual capacity substitutes for current direct evidence
• SunSwap census rows remain excluded from liquidity scoring, price consensus, direct-source precedence, visible pool selection, and V8 liquidity scoring

Impact notes

• Activation is limited to the reviewed Base Aerodrome Slipstream factory and QuoterV2 runtimes; Optimism Uniswap V3, Solana CLMM, SunSwap, Fluid, and other unratified cohorts remain shadow-only
• Every accepted profile must prove the retained pool through the factory's exact token and signed tick-spacing binding and pass independent generation, identity, price, freshness, curve, and retained-TVL validation
• Mature fresh last-known-good profiles can remain in the bounded route-only observation set across a temporary liquidity shortlist rotation without changing aggregate TVL, volume, visible pools, price consensus, target publication, or V8 liquidity scoring
• Missing, stale, price-drifted, malformed, or identity-mismatched profiles remain capability-gated; aggregate liquidity is never substituted as executable depth

Impact notes

• Only census-confirmed classic Base Aerodrome pools with stable=false are eligible; generic Solidly support and Avalanche, Linea, or Sonic deployments remain out of scope
• Each candidate requires reviewed factory and implementation runtimes, exact factory getPool(token0, token1, false) binding, pool stable=false, and an unpaused factory at one pinned block
• The model uses the pool's same-block dynamic fee; any runtime, identity, pause-state, fee, reserve, or price failure remains capability-gated instead of inheriting executable depth

Impact notes

• Incomplete paginated authoritative inventories now fail open, while complete PancakeSwap V3 authority remains scoped to concentrated-liquidity rows instead of suppressing V2 pools
• Ethereum Uniswap V2 and BSC PancakeSwap V2 candidates require a pinned canonical factory runtime, exact getPair binding, and same-block token-order, reserve, and decimals reads before becoming score-eligible
• Other V2 forks and any identity, factory, reserve, or price failure remain shaped or capability-gated evidence rather than inheriting executable depth

Impact notes

• Roughly 460 retained Raydium standard pools (USDC, USDT, USD1 pairs) move from capability-gated rows to exact constant-product execution models with score-eligible reserve-based simulations
• USD1 gains its first score-eligible Solana exit routes; USDC and USDT gain substantially deeper Solana route surfaces
• Counter-asset references derived this way are recorded as referencePriceSource "pool-implied"; token-identity and balance failures still fail closed to incomplete-exact-capture

Impact notes

• LUSD/3Crv-style metapools no longer treat the 3Crv quote side as an unknown token for effective TVL and pool-stress calculations
• Composite quote aliases continue to flow through the existing balance-health haircut, so one-sided pools remain penalized for imbalance while no longer receiving an additional unknown-token penalty
• The change applies to all configured composite pool aliases rather than special-casing one stablecoin

Impact notes

• Large malformed rows removed by the retained-pool anti-poisoning gate no longer strand the public dataset behind an inflated raw-TVl baseline
• The raw top-10 covered TVL remains in cron metadata, while the guard uses an additional quality-adjusted top-10 guard TVL for near/hard threshold decisions
• True top-asset coverage collapses still fail hard when the previous baseline had meaningful effective liquidity

Impact notes

• Pools above the large-pool TVL threshold are dropped when 24-hour volume is below the minimum-volume floor regardless of the volumeMeasured diagnostic flag
• Pool-state-only direct sources can still expand coverage with smaller eligible pools, but large zero-volume rows no longer bypass the retained-pool anti-poisoning guard
• Volume-to-TVL outlier checks, blocked-DEX filtering, protocol caps, and post-filter aggregate rebuilds continue to run around the stricter retained-pool gate

Impact notes

• GeckoTerminal and CoinGecko Onchain Curve pools remain skipped on Ethereum, Base, Arbitrum, and Polygon to avoid duplicate Curve API coverage
• Curve pools on unsupported native-API chains such as Plasma can now contribute retained liquidity, challenger-pool evidence, and DEX price observations after the normal TVL, price sanity, protocol-cap, and dedupe gates
• This lets single-chain assets such as Yuzu USD surface Plasma Curve liquidity evidence instead of depending only on aggregator prices when the native Curve API has no chain coverage

Impact notes

• Retained pools below the $50K DEX price-observation floor can still contribute to liquidity scoring when otherwise eligible, but no longer publish dex_price_usd or price_sources_json rows
• DEX price median weighting now uses canonical source families: DeFiLlama and direct API at 1.0, CoinGecko Onchain and GeckoTerminal at 0.85, DexScreener and CoinGecko tickers at 0.55
• Fallback rows that claim high-trust protocol names can no longer receive primary-source median weight solely from the protocol label

Impact notes

• CoinGecko Onchain and GeckoTerminal pool rows whose tracked-token price fails the existing peg-aware DEX observation sanity gate are rejected before staging
• Already-staged rows with implausible tracked-token prices are skipped during scoring merge, so malformed discovery rows cannot inflate coin or global TVL until their staging TTL expires
• Carbon DeFi chain-suffixed secondary-source ids now normalize to the DefiLlama `carbon-defi` protocol cap, adding a cap-level backstop for Carbon rows
• Rows without a measured token price still flow through the existing TVL, volume, dedupe, protocol-cap, and retained-pool quality gates

Impact notes

• CoinGecko Onchain, GeckoTerminal, DexScreener, and CoinGecko ticker staging can no longer persist non-finite, negative, or over-cap TVL values
• The scoring cron skips legacy staged rows above the same sanity ceiling before they can affect global TVL, coverage drift, or DEX price observations
• Valid high-liquidity pools below the ceiling continue through the existing dedupe, protocol-cap, and retained-pool quality gates

Impact notes

• Absolute TVL Depth fallback: `20 * log10(tvl / 100_000) + 20` → `35 * log10(tvl / 700_000)`; coins without market cap data no longer gain ~24 points of unearned TVL Depth
• Aerodrome/Velodrome Slipstream price observations now derive from on-chain sqrt_ratio instead of reserve ratios; concentrated liquidity pools no longer emit biased spot prices when one side lacks a tracked USD price
• Slipstream pools where sqrt_ratio is unusable and one side has no tracked price are now dropped entirely (no reserve-ratio fallback derivation)

Impact notes

• Non-Curve DeFiLlama pools that share token symbols with a Curve pool no longer inherit Curve registry metadata, balance ratios, token prices, or metapool-adjusted TVL
• CoinGecko/GeckoTerminal provider ids with underscores or provider suffixes normalize to the same canonical protocol family as DeFiLlama ids during pool-identity construction
• Staged discovery now skips a staged exact pool-id row when it uniquely matches one primary DeFiLlama UUID row by chain, protocol, token set, and pool-shape family, while ambiguous same-pair staged buckets still remain separate

Impact notes

• Intraday PancakeSwap volume no longer collapses toward zero until UTC rollover just because the current day bucket has only accumulated partial activity
• Fresh non-swap day buckets can no longer zero out yesterday's still-relevant trading activity, because trailing volume now comes from summed hourly swap buckets
• The PancakeSwap direct fetch keeps bounded batching under The Graph row cap while staying on official subgraph entities instead of adding a new historical block lookup dependency

Impact notes

• CoinGecko tickers fallback and discovery staging no longer drop every post-March-2026 ticker row just because CoinGecko now returns `trust_score = null`
• Orderbook fallback ticker intake now requires finite USD price/volume plus a stable exchange identifier, improving sanitization without depending on deprecated metadata
• Balancer direct API pools now reserve and dedupe by the true pool address, restoring exact-id confirmation against staged discovery and overlap checks

Impact notes

• GeckoTerminal, CoinGecko Onchain, and DexScreener staging rows that claim Balancer, Fluid, Raydium, Orca, Meteora, PancakeSwap, Aerodrome, or Velodrome liquidity now require authoritative exact-id confirmation when the matching direct fetch succeeded cleanly on that chain
• The guard fails open when the authoritative source is unavailable or degraded, so discovery sources still recover coverage during native-source incidents instead of hard-zeroing the row
• Liquidity cron metadata now records `stagedPoolsSkippedByAuthoritativeProtocol` separately from exact-id and derived-identity dedupe skips

Impact notes

• TVL Depth uses effective-TVL-to-market-cap ratio on a log scale (35 × log10(ratio / 0.0007)), with absolute fallback for coins without market cap data
• Volume Activity recalibrated: 38 × (log10(V/T) + 3). Zero line at 0.1% V/T, tops at ~32% V/T. USDC/USDT now score 86-90 instead of 52-56
• Pool Quality measures quality retention (qualityAdjustedTvl / totalTvl, rescaled from 15-80% range to 0-100). Fully size-independent
• Weights rebalanced from 35/20/22.5/15/7.5 to 30/20/20/20/10: structural quality (Pool Quality + Durability = 40%) now matches depth + activity (50%)
• Coins like BOLD and LUSD with high relative depth ratios see significant score improvements; large-cap coins with low relative depth see depth dimension scores decrease but compensate through volume, durability, and diversity