# To: Analyst 1 (first minute). From: Manager (Claude), for Ben, 8 Oct 2026

Ben's decisions today: decisions.md D-030 to D-033. Three asks for you.

## 1. Dumb baseline in every result (D-032)
Every edge, internal test and handover must show the strategy next to a dumb baseline: the same exits on every coin in the same universe, no picking (and, where it fits, random picks of the same count). If it does not clearly beat the baseline, it does not go to the Test Manager.

## 2. Use the nightly review: what we missed and why
Page: http://159.223.88.168/hx-c9ois/review/ . Data: http://159.223.88.168/hx-c9ois/review/misses.csv (every runner a bot did not pick, with the reason; grows every night). Ben is most interested in what we missed and why. First night (runners born 7 Oct, ATH $1M+: 26 coins):
- **Exit lesson (biggest):** fmmc5 bought 6 coins that became runners; 4 of them were sold at the 0.5x stop before they ran (0x1ec25a19, 0x3796191a, 0x4edb7e29, 0x40a31c23). Its 2 winners (29x, 19.5x) held 4 h. Is the 0.5x stop selling runners on their first shakeout? What would a wider stop / time-only exit / re-entry have done on all 12 live trades and on the full band history?
- **Universe gap:** most runners were NOT SEEN by fmmc5 (it only watches v4 pools with native ETH); runtrack venue is "?" for 59 of 74 runners in 3 days. Which venues / pool types do the runners come from, and should fmmc5 watch them?
- **Cap band misses:** e.g. SAI ($1.8M ATH) had cap $58k at 5 min, under the $250k floor. How many runners were under / over the band at 5 min?
- Bad picks: 6 fmmc5 buys lost about half and never ran (cap $366k to $723k). What did they share?
Please reply with a brief (link) that Ben forwards to the Test Manager when something is ready.

## 3. Hypothesis template (from the article Ben shared)
For every idea add: (a) who is on the other side and why they keep losing, (b) the exact condition that proves it wrong. Plus the outside money test before any handover (Ben's rule).