# Research Report · FIGHT (BSC)

- Token: `0xb2d97c4ed2d0ef452654f5cab3da3735b5e6f3ab`
- LLM reverse inference: **Enabled (LLM model openai-compatible:grok-4-1-fast-reasoning produced 3 hypothesis candidates.)**
- Current market-maker path: **DISCOVERY**
- Risk gate: **REVIEW** (HOLDER_CONCENTRATION_HIGH, NO_HARD_RISK_HIT)
- Generated at: `2026-02-15T15:19:17.830Z`

## 1) Executive Summary
- Bottom line: **Manual review required: observe only, avoid direct execution.**
- Final thesis: No immediate edge for execution; keep on watch with structured triggers.
- Why now: Top mechanism hypothesis: DISCOVERY (MED).
- What to do: Observe first; avoid chasing. Focus on net flow and whale behavior improvement.
- What to avoid: Avoid immediate entry before risk improves.

## 2) Reverse-engineered market-maker path
- Most likely path：**DISCOVERY**（confidence：**MED**）。Market is still in price discovery with mixed participation and unstable consensus.
- LLM reverse-inference summary：Recent swaps show balanced activity (13/15 baseline, 62% buys) and volume, consistent with organic early interest; however, 90% top10 concentration suggests limited distribution.
- 备选路径：DISTRIBUTION（LOW），用于防止单一路径误判。
- 筹码集中度：Top10 持仓占比约 89.95%。
- Interpretation:
  - Market is still in price discovery with mixed participation and unstable consensus.
  - 当前机会等级：**WATCH**（仅观察）

## 3) User action plan
- Observe first; avoid chasing. Focus on net flow and whale behavior improvement.
- Consider a small probe only if buy pressure strengthens and risk labels do not worsen.
- Risk boundaries:
  - If new evidence strongly contradicts the top hypothesis, downgrade confidence and re-run.
  - If risk labels or permission/proxy signals worsen, move to REVIEW/DENY.
  - If liquidity and activity collapse within short windows, invalidate opportunity assumptions.

## 4) Data quality and source tags
- Data completeness:
  - pairs: true
  - swaps: true
  - transfers: false
  - holders: true
  - lp: false
  - freshness: false
- Source tags:
  - swaps: RPC_T1
  - transfers: RPC_T2
  - holders: Covalent
  - lp: MISSING
  - freshness: MISSING

## 5) Why the model gave this conclusion
- Key evidence used by LLM and rule-verification:
  - stat: score=0.65
  - stat: inference=llm
  - address: 0x8Bf6CD197626e6D8EF9854be48eE6886D7115b80 (v2 pair)
  - stat: swapRatio=0.87
  - stat: buyPressure=0.62
- Independent risk gate result: **REVIEW**
- Risk notes: Precheck found risk signals. Proceeding with analysis but recommending review. No hard risk conditions detected.

## 6) Evidence timeline
- **2026-02-15T15:19:00.878Z** · [system] Ingest: token requested: Token 0xb2d97c4ed2d0ef452654f5cab3da3735b5e6f3ab on BSC ingested. Source-pool bootstrapped.
- **2026-02-15T15:19:00.878Z** · [system] Source: sourcify: fail · not_verified_or_not_found
- **2026-02-15T15:19:00.878Z** · [system] Source: rpc-contract: ok · code_present
- **2026-02-15T15:19:00.878Z** · [system] Source: explorer: ok · contract metadata loaded
- **2026-02-15T15:19:00.878Z** · [system] Source: dexscreener: ok · pairs:1
- **2026-02-15T15:19:00.878Z** · [system] Source: geckoterminal: fail · http_429
- **2026-02-15T15:19:00.878Z** · [system] Source: subgraph-v3: fail · no_pools
- **2026-02-15T15:19:00.878Z** · [system] Source: goplus: ok · goplus:buy_tax_bps=0, goplus:sell_tax_bps=0
- **2026-02-15T15:19:00.878Z** · [system] Source: honeypot: ok · honeypot:is_honeypot=0, honeypot:buy_tax_bps=0, honeypot:sell_tax_bps=0
- **2026-02-15T15:19:00.878Z** · [system] Source: bitquery: fail · billing_or_quota_required
- **2026-02-15T15:19:00.878Z** · [system] Source: covalent: ok · top10=89.95%, whales=10, n=100
- **2026-02-15T15:19:00.878Z** · [system] Source: explorer-holders: ok · disabled_by_default
- **2026-02-15T15:19:00.878Z** · [system] Source: coingecko: fail · http_403
- **2026-02-15T15:19:00.878Z** · [system] Source: dexscreener-narrative: ok · boost=0, activeBoost=0, paidOrders=0
- **2026-02-15T15:19:00.878Z** · [system] Source: bridge-watch: ok · disabled_by_default_budget_priority
- **2026-02-15T15:19:00.878Z** · [system] Source: primary-pair: ok · 0x8bf6cd197626e6d8ef9854be48ee6886d7115b80 · dynamic_selected v2=0,v3=13
- **2026-02-15T15:19:00.878Z** · [system] Source: rpc-tier1-swaps: ok_full · pair=0x8bf6cd197626e6d8ef9854be48ee6886d7115b80, swaps=13, buy=8, sell=5, buyPressure=0.6154, ratio=0.8667, complete=1
- **2026-02-15T15:19:00.878Z** · [system] Source: rpc-tier2-transfers: ok · transfers=null, uniqueSenders=null, uniqueReceivers=null, whales=null
- **2026-02-15T15:19:00.878Z** · [system] Source: address-freshness: fail · no_wallets_or_budget_limited
- **2026-02-15T15:19:00.878Z** · [onchain] Source pool: market pairs: Loaded 1 pair candidates from public sources.
- **2026-02-15T15:19:00.878Z** · [risk] Source pool: security scan: goplus:buy_tax_bps=0, goplus:sell_tax_bps=0, honeypot:is_honeypot=0, honeypot:buy_tax_bps=0, honeypot:sell_tax_bps=0
- **2026-02-15T15:19:00.878Z** · [onchain] Source pool: holder concentration: Covalent top10=89.95% top50=98.89% whales=10 n=100 holders=100
- **2026-02-15T15:19:00.878Z** · [system] Source pool: narrative signals: DexScreenerNarrative hasBoost=0 activeBoost=0 paidOrders=0 links=0 web=0 x=0
- **2026-02-15T15:19:00.878Z** · [system] RPC: ingest metadata skipped: Skipped duplicate RPC metadata probes because source-pool RPC signals are already available.
- **2026-02-15T15:19:01.082Z** · [risk] Precheck: holder concentration risk: Top-10 holder concentration is high (89.95%). Manual review recommended.
- **2026-02-15T15:19:01.476Z** · [trm] Source: TRM: fail · not_configured
- **2026-02-15T15:19:01.476Z** · [chainalysis] Source: Chainalysis: fail · not_configured
- **2026-02-15T15:19:01.476Z** · [risk] Risk gate decision: REVIEW · HOLDER_CONCENTRATION_HIGH, NO_HARD_RISK_HIT · Precheck found risk signals. Proceeding with analysis but recommending review. No hard risk conditions detected.
- **2026-02-15T15:19:01.761Z** · [onchain] Feature: tiered microstructure loaded: source=RPC_T1 swaps=13 baseline=15 ratio=0.8667 buyPressure=0.6154
- **2026-02-15T15:19:17.312Z** · [system] Inference: LLM reasoning: LLM model openai-compatible:grok-4-1-fast-reasoning produced 3 hypothesis candidates.
- **2026-02-15T15:19:17.524Z** · [system] Verification: hypothesis sanity checks: Applied MVP verification rules (stage prerequisites + contradiction checks).

## 7) Full mechanism hypotheses (technical)
### h1
- Stage: **DISCOVERY**
- Confidence: **MED**
- Notes: Recent swaps show balanced activity (13/15 baseline, 62% buys) and volume, consistent with organic early interest; however, 90% top10 concentration suggests limited distribution.
- Evidence refs:
  - stat:score=0.65
  - stat:inference=llm
  - address:0x8Bf6CD197626e6D8EF9854be48eE6886D7115b80 (v2 pair)
  - stat:swapRatio=0.87
  - stat:buyPressure=0.62
- Counter-evidence refs:
  - stat:falsify:Whale sells exceed buys
  - stat:falsify:Holder count stagnates or top10 pct rises
- Next data requests:
  - transfers_T2: Check net flows, whale activity to confirm organic buys vs insider support
  - lp_holders: Identify LP providers; overlap with top holders would indicate bootstrapped liquidity

### h2
- Stage: **DISTRIBUTION**
- Confidence: **LOW**
- Notes: 89.95% top10 holdings signal concentrated control ripe for distribution; low socials/no taxes fit insider-driven setup.
- Evidence refs:
  - stat:score=0.35
  - stat:inference=llm
  - address:0x8Bf6CD197626e6D8EF9854be48eE6886D7115b80 (v2 pair)
  - stat:swapRatio=0.87
  - stat:buyPressure=0.62
- Counter-evidence refs:
  - stat:falsify:Holder base expands beyond top50
  - stat:falsify:Buy pressure sustains without whale inflows
- Next data requests:
  - transfers_T3: Detect whale outflows to pair or external dumps
  - holder_changes: Track top holder balances over time for redistribution signs

### h3
- Stage: **EXIT_RISK**
- Confidence: **LOW**
- Notes: High concentration + low liq ($3.7k) poses rug risk, but no sell pressure yet; conservative flag pending transfer data. Verification: no strong contract/label risk found; lowered confidence for EXIT_RISK.
- Evidence refs:
  - stat:score=0.20
  - stat:inference=llm
  - address:0x8Bf6CD197626e6D8EF9854be48eE6886D7115b80 (v2 pair)
  - stat:swapRatio=0.87
  - stat:buyPressure=0.62
- Counter-evidence refs:
  - stat:falsify:LP locked/burned
  - stat:falsify:Diverse holder growth
  - stat:riskSignals=low (No high labels/proxy/unverified)
- Next data requests:
  - freshness_check: Verify age/activity recency
  - lp_locks: Confirm LP burn/lock status

## 8) Data collection details (appendix)
- What is collected:
  - Contract metadata and proxy/verification signals
  - Token security and tradability simulation signals
  - Holder concentration and whale behavior indicators
  - Contract-level market profile (mcap/fdv/volume/24h trend)
  - Narrative and promotion exposure signals
  - Liquidity microstructure (pool age/reserve/volume/tx)
  - Bridge inflow/outflow flow signals
  - Whale transfer directionality signals
  - Wallet freshness/age composition signals
  - Pool and liquidity venue presence
  - Swap/transfer behavior and counterparties
  - Risk labels and compliance flags
- Sources:
  - EVM RPC: eth_getCode, eth_call, eth_getLogs, eth_getStorageAt
  - Contract Metadata: Sourcify verification, RPC code/proxy hints, Explorer optional supplement
  - Public Source Pool: DexScreener pairs, GeckoTerminal token metadata, V3 subgraph pools
  - Security APIs: GoPlus token security, Honeypot.is simulation
  - Address Behavior APIs: Bitquery token holders, Covalent holders fallback
  - Explorer token stats: Etherscan/BscScan holder count
  - Market APIs: CoinGecko contract market profile, CoinGecko tickers venue split (CEX/DEX share), CoinGecko market chart trend/volatility
  - Narrative APIs: DexScreener boosts and paid orders
  - Liquidity APIs: GeckoTerminal pool reserve/tx/age profile
  - Bridge Flow: On-chain bridge watch (known bridge addresses + ERC20 Transfer logs)
  - Whale Flow: On-chain whale flow (address-cluster netflow aggregation)
  - Behavior Pattern: Clustered seller behavior from transfer graph and repeated edges
  - Address Freshness: On-chain nonce-based wallet freshness from sampled counterparties
  - Risk labels store: TRM/Chainalysis/manual labels in D1
- Methods:
  - Windowed log collection
  - State snapshot reads
  - Structured normalization into dossier/features
- Snapshot: chain=BSC, token=0xb2d97c4ed2d0ef452654f5cab3da3735b5e6f3ab, pair=0x8Bf6CD197626e6D8EF9854be48eE6886D7115b80, riskLabels=0

## 9) Analysis system proof (appendix)
- Methods:
  - LLM reasoning over structured chain data
  - Feature-conditioned hypothesis ranking
  - Falsification-oriented confidence framing
- Tools:
  - LLM API model: openai-compatible:grok-4-1-fast-reasoning
  - EVM JSON-RPC feature extraction
  - Risk gate and verification modules
- Proof checklist:
  - Hypothesis carries score and confidence mapping
  - CounterEvidence includes falsification cues
  - Inputs are structured and traceable to on-chain features
  - Risk gate remains independent safety boundary
- Hypothesis ranking:
  - h1 · DISCOVERY · MED · score=0.65
  - h2 · DISTRIBUTION · LOW · score=0.35
  - h3 · EXIT_RISK · LOW · score=0.20

## 10) Risk profile
- No risk labels found in local store.

## 11) Observation checklist
- Watch LP add/remove events and sudden liquidity drops.
- Watch contract permission changes / proxy upgrades.
- Watch holder concentration changes and cluster movements.

## 12) Boundaries / falsification
- If new evidence contradicts the stage hypothesis, re-run analysis.
- Any high-risk label hit will move decision to DENY/REVIEW.