TWAP Desk

day stop $ size $ min fair min edge from s to s max open min lead bp max edge stable s

–

trades

–

win rate

–

paper p&l

–

fees paid

–

twin mismatches

–

live p&l

Open paper positions

Market	Side	Shares	Entry	Fee	Fair@entry	Fair now	Closes
none

Recent settles

Market	Side	Entry	Fair@entry	τ	Result	P&L
no settles yet

Missed windows — why no trade

Time	Market	τ	Why it was skipped
nothing missed yet — only windows that got near-certain and were blocked show here

Model calibration

Secs to close	Model said	Windows	Up actually	Verdict
collecting… every 5m/15m window feeds this, trades or not

The gate. A paper entry needs both: the model's probability for the side is at least min fair (the "only when it's obvious" rule), and that probability beats the book's ask plus the 10%-of-min(p,1−p) taker fee by min edge (so certainty the market already priced is skipped). The calibration table is the honesty check — if the 0.95 rows don't win ~95% of the time, the model gets fixed before any talk of live.