A2-RES60

Waiting for: new 16 Oct (#2596): needs fwd n20 then audit; outside sell depth unproven

Still open:

• auditor 5-check

Test Manager forward test on research-box. Kill rule: fwd mean<1.1 at n60, or <1.0 at n30. Live only after 2 profitable days without bugs; only Ben switches.

Each step is a trade with exit rule R, numbered in time order.

# A2-BUSY60 (v4 fee-100 unhooked, busy first minute) forward paper test (Test Manager) - 16 Oct 03:30 UTC

Feed v4tape (forward only). Universe: fee 100, unhooked (hooks == 0), WETH/native; busy = first 60 s: up swaps >= 41 and down swaps <= 6 (arm A), up >= 41 only (arm B), all pools (arm Z control), PERBLK arm P (A2 #184: swaps/distinct blocks in first 60 s >= 1.25 and <= 3 other new pools of any family with first swap in the prior 30 min; sim TEST 1.34 n=154, kill mean<1.1 over 60). Entry first swap +60 s, reserve >= 0.3 ETH, 0.04 ETH, 0.01% fee + 3% slippage per side, pull (liq <= 50% of peak) = 0. Brief: fw/a2-iter10-r10.md (#173).
- Hooked fee-100 pools seen: 1147; waiting 4; open arm-runs 4; skipped {'reserve<0.3': 113, 'no swap after 63s': 474, 'pulled before entry': 3}; hook sizes {}
- arm R (tp5, stop 0.5, 240 min): all n 142, mean 0.90x, <1x 116, zeros 113 | ON-TIME only (feed lag<=30 s) n 60, mean 1.07x, <1x 47, zeros 44 | late n 82, mean 0.77x, <1x 69, zeros 69
- Sim expects (A2 fw/a2-iter10-r10.md): A 1.60 (n=87, ~5/day), B 1.17 (n=349), Z 0.98. Cutoffs 41/6 were picked from 12 tries: selection risk. Arm C start-FDV<4.63 lives in h100 arm E. Longest loss streak arm A: 0 (10 in a row = stop and review).
- QUIET exits (#172): pool stopped swapping before the hold ended; v4tape has no ModifyLiquidity so a silent pull looks the same. Counted as 0 (conservative) in net_x; column x_if_quiet_sold has the old last-price value. Quiet exits whose pool has a logged liquidity removal (liq.jsonl, forward from 14:48 UTC 9 Oct, #180): 68. Quiet count per arm: R 106.
- Late = feed lag (emitted - init, or entry swap emitted - block time) > 30 s: live would not have entered at +60 s (#153). Judge arms by the ON-TIME line. v4tape has no ModifyLiquidity events yet, so pulls are invisible here (a pool pulled early just stops swapping; trust the sim's pull rates, ~22%).
- Sign check agree 8, disagree 279472.
- Arm R (A2-RES60 #213, fw/a2-iter12-k.md): net WETH reserve at entry in [3.6, 6.2], tp5 sl0.5 4 h; sim 1.46 n=356 (tp3 1.33); cutoffs from quartiles of the sample (selection risk). Kill: mean < 1.1 over 60 trades.
- Arm Q (A1-ROUND60 #234, fw/a1-round60.md): >=3 buys (>=0.0005 ETH) in first 60 s, share that are multiples of 0.01 ETH (+-0.0002) >= 0.5, reserve >= 0.5 ETH, tp5, no stop, 20 min; sim 1.53 n=263 (tp5, in-sample threshold). Kill: mean < 1.0 over 60 trades.
- Arms N2 / N3 (A2-NONCE2 #242, fw/a2-req-iter14.md): Z rules plus creator-tx nonce (pools.jsonl field, new pools only, ppfeeds #248) == 2 / >= 3; tp3 stop 0.5 4 h. Sim N2 1.12 (n=672) vs N3 0.81 (n=563); ~5-6 N2 pools a day. Kill: N2 mean < Z mean + 0.1 over 60 trades. Pools without a nonce field are in neither arm.
- Kill: arm P mean < 1.1 over 60 trades. Kill: arm A mean < Z control mean + 0.3 over 40 trades, or 10 losses in a row. Ready for live only after 2 profitable days, no bugs, enough trades; sellability (hook skim, Ben eth_call) still open; only Ben switches (D-030).