How a strategy earns capital
Edge in short-dated options is small and perishable. Most of the work is checking that an effect is real and that the live system trades it the way the backtest did.
Eight stages
A strategy moves to the next stage only after it clears the gate.
-
01
Hypothesis
A written, testable claim, with the evidence that would refute it.
Gate · Written criteria -
02
Definition
Defined once, in a format the backtester and the live system both read.
Gate · One definition -
03
Backtest
Replayed over archived quotes, paying the spread.
Gate · Net of costs -
04
Validation
Held-out periods, controls, and fills worse than modeled.
Gate · Edge survives -
05
Promotion
Only the exact configuration that was tested can go live.
Gate · Exact config -
06
Paper
Live data and a simulated broker. Decisions have to match replay.
Gate · Matches replay -
07
Staged capital
Size goes up in steps. Each step takes a run of clean sessions, and limits are recalculated before the next.
Gate · Clean sessions -
08
Retirement
If live results drift outside what replay predicts, size is cut or the strategy is retired.
Gate · Pre-committed
Limits that block orders
Set in advance and checked in code before an order goes out.
Loss limits
Caps on loss and exposure per position, per strategy and per day, sized on the position’s maximum loss.
Fail closed
If a limit can’t read fresh inputs, new entries stop.
Failure drills
Failsafes are drilled outside market hours against process crashes, feed loss and broker outages.
Reconciliation
Positions and fills are reconciled against the broker. A retried order can never double a position.
We compete on selection and cost, not speed
We take liquidity at our own pace. Fill costs are broken out by entry and exit, side, order type and leg count, so each strategy is judged on what it actually pays.