Underwriter declines
SpecimenSelling an option is underwriting insurance. This agent writes defined-risk credit spreads on a fixed ETF universe — and refuses far more often than it writes. The refusals are the product, so they lead.
What it refused to do
SpecimenEight stages stand between a candidate and an order, and every one of them can stop the trade. A refusal is recorded with the code that caused it, so a decision nobody made is as visible as a decision somebody did. The gates should be judged on whether they fire, not on the P&L.
Reading the refusal ledger
Open book
SpecimenEvery position is a two-leg vertical: a short leg that collects the premium and a long leg that caps the loss. Maximum loss is width minus credit, known before entry and bounded. Nothing here can lose more than the figure in its own row.
Decision log
SpecimenChronological, newest first: what was considered, what happened to it, and the codes that decided. A strategy whose rejections are invisible is a strategy nobody can check.
Order tape
Every submission carries a unique client order ID, and a credit spread is submitted at a negative net limit price — the broker's sign convention for money received.
Two P&L series, kept apart
SpecimenThe official figure is what the paper broker reports. The shadow figure is what the same trades are worth after crossing the quoted spread. They measure different things, so they are drawn separately and never added together.
Official Alpaca paper account
Account equity as the broker reports it. Paper multi-leg fills simulate against modified indicative quotes on the Basic plan, and the fill model is undocumented — this is a number we report, not one we trust.
Shadow Conservative, spread-crossing
The same fills repriced with explicit slippage: exits marked across the quoted bid/ask rather than at the paper engine's price. Lower by construction. This is the honest figure, and the one to judge.
Why two numbers. Underwriter runs on the free Basic plan for the whole contest: IEX equity data rather than consolidated SIP, option trades delayed fifteen minutes, and indicative option quotes rather than OPRA NBBO. Alpaca's paper engine checks a spread's marketability against those altered quotes, and the multi-leg fill model is not documented anywhere.
A paper P&L built on that is a plausible number with an unknown error bar. Publishing it alone would be a claim we cannot support, so the conservative series sits beside it and the gap between them is left visible rather than reconciled away.