POLYBOLTops
btc-updown-5m T− —
--:--:-- UTC

Bands

Underlying price

One line, chosen here, instead of three chips that drew on top of each other. Spot is the microprice whenever the book is fresh — measured identical on 100% of a session's rows — so the two were one line under another, and the simple mid sits about $0.004 away, which is 1/250th of a pixel on this axis. The crosshair is where those differences are readable, and it still carries all three.

Decision marks

A POST row is stamped when the trader drains the completed response, so as logged it sits at send plus its own round trip — measured p50 41 ms, p90 265, max 650 over 167 orders. Sent-time subtracts each row's rtt, which is exactly that offset. It matters in the tail, not the median: a slow ack is what happens when the market is moving, and the mark drags right while the band moves on beneath it. Measured at the answer, 11 of 155 posts appear priced above their own limit; measured at the send, none do and tick − limit is 0.00 on every one. Told (maker) fills never move — nothing was asked of the venue, so their timestamp is already when we were told.

Legend

waiting for the ring
start the trader with --ring /dev/shm/polybolt.log — a dry run is enough

traderoffline

monitoroffline

ring · drainoffline

recorderoffline

caps & risk

session
this window
contracts
orders
quotes livefills · rejectspacing

latency effort

wake p50 · 1m
wake p95 · 1m
spot age p95
t_wire p50 · 15m
venue rtt p50
ring depth
wake
t_wire
no data yet

P&L

day · settledno monitor
session · settled
makertaker
position
paid
worth now
open P&L
last window
day vs loss cap
no windows yet

monitor verdicts