What our signals tell you.

Ranking signals are historical proxies.

The current engine combines mempool.space public node rankings for liquidity (45%), connectivity (40%), and age (15%), with capacity and channel-count adjustments. The liquidity ranking does not reveal usable directional channel balances. The engine ranks node candidates; it does not construct an end-to-end route between your sender and recipient.

Reliability signal · 0–100

Route confidence is adjusted by status, risk category, confidence trend, stability, and optional payment-size bands, then clamped to 0–100. A score of 100 is the upper end of this heuristic. It does not mean a payment has a 100% chance of succeeding. Payment-size adjustments are heuristic penalties, not tested path liquidity.

Historical ranking consistency

This score measures dominance, appearances, confidence, volatility, recency, and leadership runs within our recorded winners and alternates. It is partly dependent on our own ranking algorithm. Low scores or missing coverage do not establish that a node is unreliable, insecure, or unable to route payments.

Coverage and freshness

Public history contains collected winners and alternates, not every Lightning node. The dashboard labels its selected snapshot count and observed time span. Snapshot times are interpreted as UTC; server and collector clocks must be configured accordingly. Cached public responses can lag collection by 90–120 seconds. A selected historical window may contain gaps.

Observed payment performance

Private workspace metrics summarize the payment outcomes submitted by that customer. Success rate is successful outcomes divided by all submitted terminal outcomes in the window. Average latency includes both successes and failures. Retries sum attempts minus one. Fees are the totals supplied by the collector. Missing events or selective reporting change these metrics.

How to use the evidence

Investigate changes using your node logs and payment outcomes. Network rankings alone do not establish a cause of failure. Predictions require independent validation against held-out payment outcomes, calibrated by amount, sender, destination, and time.

Learn how LND uses payment history and liquidity information →