Yield, Pool Dynamics & Risk Signals

This page defines yield, structural pool dynamics, and risk-style signals derived from Solana DeFi activity.

Signals are computed from normalized pool snapshots and trade aggregates.

Risk signals are descriptive. They are built for monitoring and alerting, not advice.

---

Coverage (signal families)

• Yield and pool performance

  • Fee yield proxies, utilization, and stability.

• Pool dynamics

  • Liquidity flows, reserve shifts, and imbalance regimes.

• Risk and anomaly signals

  • Liquidity cliffs, abnormal impact sizing, and concentration spikes.

Common windows: 15m, 1h, 4h, 24h.

---

Yield & pool performance

Core yield metrics

• Fee revenue

  • Estimated fees per window (when fees are observable).

• Fee APR (proxy)

  • fee_apr ≈ (fees_24h / tvl) * 365

• Utilization

  • Volume-to-liquidity ratios as a usage proxy.

• Stability

  • Variance of fees and volume across adjacent windows.

APR values depend on pricing inputs and fee visibility. Always treat them as approximations.

---

Pool dynamics (liquidity behavior)

• Net LP flow

  • Net liquidity added/removed per window.

• Reserve volatility

  • Rapid reserve changes relative to baseline.

• Imbalance regime

  • Persistent skew between token reserves.

• Volume-to-liquidity divergence

  • High volume with dropping liquidity (stress proxy).

Example: pool dynamics (conceptual)

{
  "window": "1h",
  "start_time": "2025-12-17T09:00:00Z",
  "end_time": "2025-12-17T10:00:00Z",
  "protocol": "amm",
  "pool_address": "ExamplePoolAddress",
  "metrics": {
    "net_liquidity_change_usd": -48213.44,
    "reserve_volatility_score": 0.72,
    "imbalance_ratio": 0.18,
    "volume_to_liquidity": 0.94
  }
}

---

Risk & anomaly signals

Risk signals detect abnormal on-chain conditions. They are designed to be machine-consumable.

Common signal types

• Liquidity cliff

  • Large LP withdrawal relative to pool TVL in a short window.

• Impact-sized swapping

  • Trades sized unusually large relative to liquidity.

• Concentration spikes

  • Top trader / top pool share rises sharply vs baseline.

• Repeated high-impact interactions

  • Many large actions clustered in a tight time window.

Output conventions

Signals are emitted with:

• severity: low | medium | high

• confidence: 0..1

• reason_codes: stable machine labels

• dedupe_key: a stable identifier for cooldown + de-duplication

Example: risk signal payload

Example: risk signal (conceptual)

{
  "signal_type": "risk_anomaly",
  "detected_at": "2025-12-17T10:03:51Z",
  "protocol": "amm",
  "pool_address": "ExamplePoolAddress",
  "category": "liquidity_cliff",
  "window": "15m",
  "size": {
    "notional_usd": 221004.18,
    "pool_tvl_share": 0.27
  },
  "reason_codes": ["LP_WITHDRAWAL_SPIKE", "TVL_DROP"],
  "severity": "high",
  "confidence": 0.89,
  "dedupe_key": "risk_anomaly:ExamplePoolAddress:liquidity_cliff:15m"
}

---

Treat signals as inputs to a policy engine. Avoid acting on single datapoints.

Consumer guidance

Recommended patterns:

• Use completed windows for alert triggers.

• Combine size (relative to TVL) and rate (per window).

• Apply cooldown using dedupe_key to prevent alert storms.