Who is the top Hyperliquid builder?
30 complete UTC days (2026-09-17 → 2026-10-16), computed in-house from Hyperliquid's own public fills CDN. No key, no third-party stats API — the numbers below are re-derived from the raw per-fill stream, and the guard (check_builders.py --self-test) rejects a wrong implementation. Cohort: top 5 builders by 1D builder fees on the public loris.tools/builder-analytics leaderboard (the rail carrydesk Q400 consumes).
Answer
A builder earns {"b": addr, "f": tenths_of_bps} on notional routed with their code attached — the fee rides inside trades the user was making anyway. Fee rate = builder fees ÷ routed notional × 10⁴.
Top 5 builders by 30-day fees
| # | builder code | fees (USDC) | volume | bps | fills | traders | share |
|---|---|---|---|---|---|---|---|
| 1 | 0x5af1b5f44207784dcb850bbb4143c5dcd1885f71 | $845,778.63 | $881,074,583 | 9.60 | 368,288 | 10,185 | |
| 2 | 0x2a2b6b093a9813fbd8cddae800c3d17d46460d17 | $760,317.02 | $1,665,103,699 | 4.57 | 787,008 | 19,639 | |
| 3 | 0xb84168cf3be63c6b8dad05ff5d755e97432ff80b | $714,492.37 | $1,392,359,771 | 5.13 | 918,418 | 28,767 | |
| 4 | 0xe95a5e31904e005066614247d309e00d8ad753aa | $689,254.06 | $700,697,262 | 9.84 | 329,848 | 7,785 | |
| 5 | 0xe966a12bf7b93838096e4519a684519ab22df618 | $105,992.12 | $599,731,409 | 1.77 | 304,807 | 812 |
A day with no attributed fills (404 on the CDN) is a zero day, listed in missing_days — the window is never silently truncated. Generated 2026-10-17T10:49:52+00:00.
Machine-only: rails, schema, guard
{
"schema": "HL-BUILDERS-V1",
"rails": {
"cohort": "https://loris.tools/builder-analytics",
"fills": "https://stats-data.hyperliquid.xyz/Mainnet/builder_fills/{addr}/{day}.csv.lz4"
},
"cohort": [
{
"rank_1d": 1,
"builder": "0xb84168cf3be63c6b8dad05ff5d755e97432ff80b",
"fees_1d_usd": 106250.0
},
{
"rank_1d": 2,
"builder": "0xe95a5e31904e005066614247d309e00d8ad753aa",
"fees_1d_usd": 64510.00000000001
},
{
"rank_1d": 3,
"builder": "0x5af1b5f44207784dcb850bbb4143c5dcd1885f71",
"fees_1d_usd": 55380.0
},
{
"rank_1d": 4,
"builder": "0xe966a12bf7b93838096e4519a684519ab22df618",
"fees_1d_usd": 42010.0
},
{
"rank_1d": 5,
"builder": "0x2a2b6b093a9813fbd8cddae800c3d17d46460d17",
"fees_1d_usd": 36950.0
}
],
"window": {
"days": 30,
"start": "2026-09-17",
"end": "2026-10-16",
"basis": "last N complete UTC days (today excluded)"
},
"pool": {
"builders_tracked": 5,
"fees_usd": 3115834.1975302254,
"volume_usd": 5238966723.170113,
"trades": 2708369,
"avg_fee_rate_bps": 5.947421241959762
},
"missing_days": {},
"parse_stats": {
"skipped_rows": 0,
"fill_rows": 2708369
},
"generated_at": "2026-10-17T10:49:52+00:00"
}
Live fill rail: https://stats-data.hyperliquid.xyz/Mainnet/builder_fills/{addr}/{day}.csv.lz4 (lz4 frames, decoded by the pure-stdlib lz4f.py — verified byte-identical to lz4cat plus 12 negative cases; see PROOF.md). Cohort rail: https://loris.tools/builder-analytics (SSR leaderboard, the rail carrydesk Q400 consumes). Guard: python3 check_builders.py --self-test — fixture pin, independent hand re-derivation of fees/volume/bps/trades/pool/window, a hand-literal per-day bar, and 3 poisoned outputs each rejected by its dedicated bar alone.