Options-premium trading agent
Why
Selling options pays a little most days and loses a lot on a few. The question was whether it can be done with rules strict enough that the bad days cannot wipe you out, and honest enough that every decision can be checked afterwards.
How
I wrote the hypothesis down and tested it before any trading code existed. Then I built an agent that sells a defined-risk position only when volatility is measurably expensive, and refuses otherwise. Eleven numbered gates each have a test proving they halt the strategy, and every decision is written to a log that cannot be edited after the fact. It ran live on real market data for the hackathon week.
What came out
An agent that trades on a measured edge and sits out the rest, with a sealed record of every decision it made.
What broke
The pricing model failed its own accuracy gate, because the volatility index is not the same thing as the price of the options being sold. The tail hedge that was supposed to protect the worst days turned out to be inoperative live, since the data feed had no quotes in the window it needed. One result improved when I doubled the costs, which is impossible, so it was flagged as fragile rather than reported. All three are in the repo, next to the good numbers.