Strategies
Week 6 — Cheap leftover asks
This day’s lineage note is the cleanest picture of what I changed, and of what failed when I tried it live.
| v1 | Favorite buying. High win rate, tiny profit per winning trade. |
|---|---|
| StratA | $0.40–$0.72, 11–15 seconds left, gap floor around $15. Not treated as proven. |
| StratB | Wider search. Champion-style configs sat nearer $0.30–$0.55. |
| UC / penny | $0.01–$0.20 leftover asks on the about-to-win side. A book condition, not a security hole. |
A paper sweep loved the 1c–5c bucket. Pasted live on 5–6 May: 22 orders, 3 wins, −$12.16. Queue, delay, and non-fills are why that paper bucket was the worst place to trust a replay. The 4GB books.jsonl capture was not on the machine. This was logged candidates, not a fillability simulation.
Moving the entry band was the engineering work. The cheap-fill paper profit did not survive that live sample.
← Blog