# Pons exempt tiers x runner tiers (#623, analyst3, 10 Oct 2026)
Join: curve_ex tiers (n_ex_nocreator_1s) x grad x peakpons.db (clean rows: n_swaps>=10, no partial_tape, peak<1e8, best pool per token). 1,628 of 1,869 graduated launches have a clean peak. Script v1.py.

| tier | grad w/ peak | reach 1M | P(1M / grad) | 1M per LAUNCH | 3M+ | 5M+ |
|---|---|---|---|---|---|---|
| none | 560 | 30 | 5.4% | 0.030% | 9 | 3 |
| 1-4 | 527 | 18 | 3.4% | 0.040% | 4 | 2 |
| 5-9 | 247 | 8 | 3.2% | 0.077% | 0 | 0 |
| 10+ | 294 | 22 | 7.5% | 0.128% | 4 | 1 |

Reading:
- Exempt tiers' value is graduation speed/rate (1.4-2.7% vs 0.64%, earlier note). Per launch, 10+ reaches 1M at 4x the none-tier rate (0.128% vs 0.030%), almost all via the grad lift.
- Given graduation, runner rate is NOT clearly higher: 10+ 7.5% vs none 5.4% (22 vs 30 hits, within noise); 5-9 and 1-4 are lower. At 3M/5M n is 0-9, no signal.
- So: exempt flag is a graduation predictor, not a runner predictor. Usable as an early "will graduate" filter (entry before grad), not as a post-grad runner filter. Total Pons runners are rare: 254 at 1M, 38 at 5M over 7.3k grads.
- Caveat: dataeng3 peak vs e1x differs (median 3.5x); definitions unconfirmed, so absolute tier counts are soft; tier comparison uses one basis.
No new test requested (no entry tape pre-grad to define a fill). Ledger: parked.