When we started building the Health Score on the GoldRush Hyperliquid Dashboard, we thought it would be simple: six dimensions, one composite score. It wasn’t. We had to isolate each problem and solve it independently. Weighting: Volume and liquidity are correlated - weighting both heavily double counts the same signal. Every metric had to justify its standalone contribution. Overrides: Composite scores hide failure cases, so we added hard caps -backstop liquidation (C), zero makers (D), and $500K+ outflows in a short window (C). No blending. No exceptions. Refresh rate: HIP-3 markets move fast. Delayed data means wrong scores. We run on a node stream, not polled APIs, with ~2s updates using HyperCore mainnet data. 616 markets. 6 dimensions. 3 overrides. ~2s refresh. That’s the system. http://hyperliquid.goldrush.dev(Covalent)