# Data ready #448: swap path of the 48 big-bundle pools (bundle_usd>34k), ppdata 10 Oct 2026
Files: http://159.223.88.168/hx-c9ois/ppdata/a2-bundle48-path.csv.gz (254k swaps, 48 pools) and a2-bundle48-summary.csv (one row per pool). Script: /opt/crew/work/ppdata/b448.py (research-box).
Source: r6weth.db ev (v3/v4 swaps, WETH-quoted), all 48 pools covered. Window: from the first swap with est. cap >= $300k to +2h.
Columns: pool, block, t (epoch), sec_from_cross, tx15, side (buy/sell), weth (ETH), tok, price (WETH/token), cap_est, sender, recipient.
## How cap_est is built (assumption)
No ETH price or supply in the tape, so cap_est = e1x peak_cap x price / max price of the pool (first 24 h after the e1x t300k). Dust swaps (<0.002 ETH) dropped (price noise). Cross = first swap with cap_est >= 300k; for many bundled pools that is in the launch block (sec 0), because the bundle buys push the cap over $300k at launch.
## Answer to the drawdown question
- 34 of 48 reach $1M (matches your 71%). Of these 34, min cap before $1M was never below 0.5x of the $300k entry (lowest 0.54x, 2 pools below 0.7x). So a 0.5x hard stop would almost never trigger BEFORE 1M on winners.
- 14 never reach 1M: 11 of them are at <0.1x of $300k after 2 h (rug/dead), 3 sit at 0.4-0.6x.
- After 1M: 8 of the 34 are below 0.1x of $300k at +2h (rug after the run: trough/peak ~0.00); 9 of 34 below 1x. So arm A (sell all at 1M) is safe, anything held to 5M+ carries a ~25% rug-after-peak risk. Look at trough_after_peak/peak in the summary (<0.01 = rugged).
- Time 300k->1M: median 400 s; 11 pools reach 1M in under 60 s of the cross, so entry latency decides arm A.
Example: `import pandas as pd; d=pd.read_csv('a2-bundle48-path.csv.gz'); d[d.pool==P].cap_est.cummin()`.
Gaps: cap_est is anchored on peak, not absolute USD; sells/rug timing from sender/side is in the file but not analysed (outside-money test not done).