# A3.11 Smart wallet cohort, earlier entry (Analyst 3)

## Read this first (07 Oct 10:52 Bangkok)
- **Verdict: KILL.** The chosen cell (picked on half A, as pre-registered) fails every leg. Fixed cohort in half B: -$407 a day. Rolling cohort: -$559 (A), +$2,599 (B). Beats random: 49% (fixed B), 40% (rolling A), 89% (rolling B), against a 95% bar. Added to the live book it lowers $ a day in half B fixed (-$1,357) and in half A rolling (-$677).
- **Why (mechanism):** "bought 2 winners under $300k" is not rare. 2,541 wallets qualify from only 69 half A winners. These are busy early buyers who buy almost everything, so the list mostly says "this coin got early buyers". Every K=1 and K=2 cell loses or breaks even. At K=3 the half B numbers look good, but all 9 cells lose in half A rolling, so that is the half B tape lifting everything (random half B coins also average 0.96x vs 0.80x in A), not the wallets.
- **Half B big coins caught (fixed cohort, chosen cell):** 5 of the 26 coins that hit $3M, at entry caps of $622k to $3.9M (kit coins open high), with results of 0.22x to 4.66x. It also caught 10 of 111 "10x from the 30 minute cap" coins. Most of those are tiny coins that peaked under $300k and were stopped out.
- Nothing to build. If anyone retries wallet following, it needs a much tighter list (for example 3+ winners and a hit rate above the base rate), which half A is too short to supply.

## Pre-registered rule (published 09:50 Bangkok, before any result)
- Winner (half A only) = cap reached $3M with $20k of buys at $3M or more, before 1 Oct 00:00 Bangkok. 69 winners.
- Buyer = swap recipient; a router recipient is resolved to the tx signer (56,409 router buys resolved). Cohort = wallets that bought 2+ winners under $300k. Dropped: contracts (24), the coin creator and wallets fed by the creator outside the pool (56), snipers who sold or sent out inside 2 min (612), bots that were early buyers in more than 30 half A launches (269 drops; 3,109 bot wallets in all).
- Signal: K distinct cohort wallets buy within N minutes of the first swap; entry = first swap at least 5 s later, $1k, in-range depth, 1% fee. Live book exits (0.3x stop, 5x, KIT HOLD for Kit A/B, pull, 12 h).
- Fixed cohort tested on half B; rolling cohort (only winners whose $3M moment came before the trigger) tested on both halves. Cell = best half A rolling $ a day: **K3 N5**.

## Grid ($ a day, $1k tickets; days A 18.3, B 6.4)
| cell | fixed, half B: trades, mean x, $ a day | rolling A: trades, mean x, $ a day | rolling B: trades, mean x, $ a day |
|---|---|---|---|
| K1 N5 | 582, 1.00x, +$170 | 429, 0.70x, -$6,922 | 810, 0.91x, -$10,890 |
| K1 N15 | 615, 0.99x, -$695 | 575, 0.79x, -$6,481 | 848, 0.91x, -$12,048 |
| K1 N30 | 641, 0.98x, -$1,732 | 621, 0.81x, -$6,385 | 869, 0.90x, -$13,319 |
| K2 N5 | 93, 0.79x, -$3,005 | 88, 0.71x, -$1,404 | 274, 1.01x, +$290 |
| K2 N15 | 165, 0.86x, -$3,595 | 244, 0.72x, -$3,734 | 348, 0.95x, -$2,609 |
| K2 N30 | 228, 0.83x, -$6,043 | 313, 0.80x, -$3,382 | 399, 0.93x, -$4,133 |
| **K3 N5 (chosen)** | 26, 0.90x, -$407 | 36, 0.72x, -$559 | 77, 1.22x, +$2,599 |
| K3 N15 | 81, 1.42x, +$5,366 | 139, 0.67x, -$2,516 | 148, 1.18x, +$4,184 |
| K3 N30 | 138, 1.06x, +$1,385 | 228, 0.68x, -$3,957 | 214, 1.36x, +$11,882 |

- Worst days for the chosen cell: fixed B -$2,194, rolling A -$3,354, rolling B -$8,102.
- Random: 1,000 draws of the same count from coins with 3 distinct non-bot buyers inside 5 minutes (A 1,280 coins, mean 0.80x; B 1,071, mean 0.96x).
- Add to the live book (two tickets on overlap, earlier entry kept; book data to 6 Oct 13:40): fixed B $10,681 to $9,324; rolling A $2,580 to $1,903; rolling B $10,681 to $12,925.

## Half B $3M coins caught (fixed cohort, K3 N5)
| token | peak | entry cap | minutes after launch | result |
|---|---|---|---|---|
| 0x56fa3670e19396c4ee8f7775b49ceaede04a84ef | $86.5M | $2.81M | 1 | 4.66x (5x sale) |
| 0x60042b65a0c0a90ba75d17879850b6f99d9422eb | $31.6M | $622k | 4 | 4.62x (5x sale) |
| 0x0f526bdc56d1d764bd744412c1aba039c3728b76 | $18.9M | $1.35M | 2 | 3.03x (kit trail) |
| 0x74ccd8416cbc4e4a864176c2707bc1434819ef0f | $8.2M | $1.00M | 4 | 3.74x (kit trail) |
| 0xc60ba256b44334a0cd2c7242e98b88f031abb006 | $7.2M | $3.91M | 1 | 1.37x (12 h) |
- The rolling cohort caught 11 of 26 (incl. 0x224a794031d54e49b46919ade2c55ff255dfdd97 14.28x open, 0xb18a4bcf9018bedb1fcb90ed80f87087151c09f4 5.10x, 0xf9d7691a78b96b7a1e4bb2d1f15fe63fabef885b 6.55x), but paid for them with 60+ stopped small coins.

## Data and limits
- My rebuild: raw Uniswap v3 WETH/USDG Swap/Mint/Burn from the public Robinhood RPC, 8,630 pools, 12 Sep 16:29 to 7 Oct 09:39 Bangkok; supply from totalSupply now (burns after launch would shift caps a little).
- Sniper test uses token transfers for 75 of 96 winners, sell swaps for the rest. Bot list counts all half A launches, so early half A rolling triggers see a bot list built with later launches (behaviour, not outcomes).
- Trades: http://159.223.88.168/hx-c9ois/sw/sw-a3-11-trades-fixed.csv , http://159.223.88.168/hx-c9ois/sw/sw-a3-11-trades-roll.csv
- Compute ran on Ben's Mac, not on main.

Analyst 3, 07 Oct 2026 10:52 Bangkok. Latest: http://159.223.88.168/hx-c9ois/sw/sw-a3-11-smart-wallets.md