VelocityQuant · Phase 1 Shadow Run

VelocityQuant

where mathematics meets milliseconds

live · auto-refresh 15s

–

probing newark…

Inicio Shadow Run Projector Balance 📜 Evolución V1→V6

monitoring window

––h ––m ––s

started — · ends —

daemon · liquidator_rs.service

probing

pid — · uptime —

solana slot —

circuit breaker: probing

salud industrial · 4 pilares (R73 v2 Gemma)

TTPR · tip-to-profit ratio

–

tips_usd / gross_usd · target 15–30%

elite 0–15 silver 15–35 bronze 35–60 red >60

APPO · avg profit per opportunity

–

paper_gross_usd / would_send_true · ↑ rico, ↓ pobre

best bundle hoy: –

ODR · opportunity decay rate

n/a

tag WOULD_SEND→BUNDLE_SENT no instrumentado · target <120ms

LRV · landing rate variance (1h windows)

n/a

stddev del landing_rate por hora · agregador histórico pendiente

gating diagnosis (R72 estetoscopio)

gating breakdown · why we don't send

–

over total ws=false (gated bundles)

■ profit: – ■ stale: – ■ lag: – ■ cb/depeg: –

slot_lag distribution · bimodal check (R72 D2)

–

max slot_lag observed

shadow counters

hermes ok

–

poll every 1.5s

hermes failed

–

timeouts/5xx

cb tripped

–

slot_lag ≥ 5

cb warn

–

slot_lag 2–4

depeg blocked

–

tier-strict gate

decoded err

–

CLMM parse

cyclic scans

–

200ms tick

cyclic quotes

–

cycle_finder hits

tip target (lamports)

–

jito p50 clamped 20k–25M

paper p&l (since shadow start · 12h window)

paper net (gated)

–

would_send=true bundles only · what we'd actually pocket if execute path was wired

hypothetical net (all quotes)

–

if NO gates blocked · diagnostic only · gross is local CLMM math (sqrt_price), tip is jito floor

best bundle (theoretical)

–

single best net seen

total quotes

–

since shadow start

would_send true

–

— %

would_send false

–

gated by depeg/lag/cb/profit

paper gross

–

sum net_profit on ws=true

paper cost

–

sum tip+fees on ws=true

avg slot_lag

–

max –

pyth oracle (hermes v2 pull)

no feeds yet…

audit verdicts (gemma 4 31b)