# #1266 round-number buy share on v4 nohook (native+weth), analyst2
Data: /opt/ppcopy/swapamt.db (ppdata #1218). Round share = share of first-300s buys with round ETH size (same isround as v3 test it56). Trades: pull-aware r6 tape sim run3 (analyst1 m1/s8.py: cross TH, lag 60s, stop .5, tp 3, 4h, 4% cost), joined by pool. Script it1266/r.py. USD-quote pools excluded (no ETH rounding).
| TH | n | cut | hi mean | lo mean |
|---|---|---|---|---|
| 150k | 677 | 0.10 | 1.02 (549) | 0.73 (128) |
| 150k | 677 | 0.20 | 1.24 (163) | 0.88 (514) |
| 200k | 435 | 0.10 | 1.00 (359) | 0.67 (76) |
| 200k | 435 | 0.20 | 1.07 (125) | 0.89 (310) |
| 300k | 196 | 0.10 | 0.78 | 0.63 |
Halves (cut .1, 150k): hi .89/1.14 vs lo .53/.93, same direction both halves. 
Verdict: PARK. Direction holds (hi > lo in all halves) but the lift is relative to a losing base; hi alone ~1.0 after cost, no absolute edge, at 300k cut .2 it flips. Most pools (81%) are already "hi" at 0.1, so the cut is weak. Same lens works on v3 band (1.67 vs 1.12); on v4 it only separates bad from mediocre. Revisit when the v3 forward (kitenrich round_buy_share300) has n>=30.