# analyst3: Pons exempt-wallet launches vs outcome (#486 #488), 10 Oct 2026
Scripts: q1.py q2.py (work/analyst3). Data: ponsexempt.db curve_ex (172,789 launches) + pons.db grad (7,292 graduations).

## Blocker: runner tiers cannot be computed yet
Pons graduated tokens are not in seen_peakcap.csv (only 37 of 7,292 by token, 0 by pool; peakcap pools are 42-char v2/v3 addresses, Pons grads are 66-char v4 pool ids). h4e_usd.db holds 0 of them (it is hook 0x4e34 only). r6usd matches 68 pools. So there is no peak-cap for the Pons graduation pools. Asked dataeng (board) for a peak-cap table on pons.db grad pools (pool, token, peak_cap, t_100k/300k/1m/5m).

## What the data does say (graduation = curve filled, the only outcome we have for all 172k launches)
Tier = exempted non-creator addresses within 1 s of launch:
| tier | launches | grad | grad % | median sec launch->grad |
|---|---|---|---|---|
| none | 100,520 | 642 | 0.64 | 957 |
| 1-4 | 44,747 | 632 | 1.41 | 570 |
| 5-9 | 10,351 | 274 | 2.65 | 400 |
| 10-19 | 11,309 | 201 | 1.78 | 93 |
| 20-29 | 3,788 | 83 | 2.19 | 55 |
| 30+ | 2,074 | 37 | 1.78 | 130 |
- Any exemption roughly doubles to quadruples graduation (1.4-2.7% vs 0.64%); no dose rise above 5-9 wallets. Holds in all four launch-time quartiles (ex>=10: 1.4-2.7% vs 0.9-1.1%), so not a one-batch artifact.
- Exempt launches graduate much faster (1.5-3 min at 10+ vs 16 min for none): insider-filled curves, consistent with a decoy/insider pump rather than organic demand.
- 58% of launches have zero exemptions, yet the none-tier still supplies 642 of 1,869 matched graduates (34%), so exemption is NOT required to graduate.
- Not a crew fingerprint: 115,596 creators; only 82 have >=10 launches with ex10; among creators with >=5 launches the median graduation rate is 0. Top "serial exempt wallets" (1,000-1,900 curves each) include 0xca11...ca11 (Multicall3) and look like routers/helpers, so exempt-wallet counts need that infra filtered before building a crew list.

## Tiny overlap (n=26 with peak, 37 tokens; biased, illustration only)
Tokens with $1M+ peak: none-tier 3-4 of 11, exempt tiers ~8 of 15 (runtrack_ath rows are runners by construction). No signal at this n; e1x peaks on thin pools are unreliable (some absurd 1e12+ values).

## Verdict / next
- Unproven as a runner signal. Graduation lift 2-4x is real but graduation is not a runner (0.64% -> ~2% of launches still means 98% never graduate).
- To finish: peak-cap for grad pools (dataeng), then tier x runner rate and entry multiple at 300-400k; compare insider sell timing using ex table + post-grad sells (tx.from pass requested by scout).
- Filter serial infra addresses (Multicall3 etc.) before using exempt counts as features.