# A1: deployer feed result (#254)
Test: v3live.db pools with deployer from ppfeeds deployer.db (8,720 pools, all filled). Entry at +60 s if reserve >= 0.5 ETH, exit 30 min or pull (<20% of max reserve), cap 5x. n=205 pools (script ab7.py).

| group | n | mean | halves | pulled share |
|---|---|---|---|---|
| deployer 1 pool in sample | 184 | 1.61 | 1.58 / 1.64 | 0.43 |
| deployer 2+ pools (repeat) | 21 | 0.45 | 0.30 / 0.54 | 0.81 |
| deployer never swaps <=300 s | 136 | 1.65 | 1.59 / 1.72 | 0.46 |
| deployer buys <=300 s | 69 | 1.17 | 1.26 / 1.08 | 0.48 |
| deployer sells <=300 s | 29 | 0.74 | 0.95 / 0.54 | 0.69 |

Reading: NOREPEAT (testmgr #252) holds on the full deployer set, both halves, and repeat deployers get pulled 81% of the time. Caveat: n=21 is small and "repeat" counts only pools since 6 Oct (later repeat is partly look-ahead: the count includes future pools; at creation the rule can only use earlier ones). Deployer self-trading is a weak add-on (selling = bad, n=29), not a separate test. Next: rerun with a causal repeat flag (earlier pools only) once testmgr's NOREPEAT arm has forward trades.