Skip to main content
Every token on the leaderboard is scored 0–100 every five minutes. The score is called HP. At the hard cut (hour 96), the bottom six HP ranks are filtered. The same ranking decides finalists into settlement at hour 168.HP is a weighted composite of five on-chain signals. Each targets a different failure mode:
ComponentWhat it measures
velocityDecayed net buy inflow with sybil dampening
effectiveBuyerssqrt-dampened headcount of meaningful buyers
stickyLiquidityTime-weighted LP depth penalised by recent withdrawals
retentionLong+short two-anchor holder retention fraction
holderConcentrationHHI-mapped holder distribution
A sixth component, momentum, was tested in Track E v2–v4 and removed 2026-05-03 — its 10% weight was redistributed to stickyLiquidity. The component code remains behind a feature flag for v6 reconsideration. See HP methodology for the full derivation.

​
Active weights

Version 2026-05-03-v4-locked:
ComponentWeight
velocity0.30
effectiveBuyers0.15
stickyLiquidity0.30
retention0.15
holderConcentration0.10
For how these values were derived — what was tried, what was rejected, and what cross-validates against published research on adjacent launchpads — see HP methodology.

​
How weight changes work

Weights are off-chain config. Contracts read an oracle-signed Merkle root of rankings, not the coefficients — so a change takes effect only when the oracle reposts a new root.Updates require an empirical fit on at least 200 tokens, external review, seven days of public notice, and an indexer redeploy with a new version string. Every score the indexer writes is tagged with the version that produced it, so the active set is always verifiable.