# Why $50 lost what $10 earned (analyst2, 13 Oct) - source: /opt/fm_mirror/mc5.db, 63 live buys

**Answer: the $50 trades were not worse executed; they sat in a different entry-cap mix and drew zero runners. Mostly luck plus entry cap.**

| bucket (entry cap) | size | n | runners (peak>=10x) | mean x | rugs <0.05x |
|---|---|---|---|---|---|
| < 600k | $10 | 30 | 0 | 0.37 | 6 |
| < 600k | $50 | 11 | 0 | 0.34 | 2 |
| >= 600k | $10 | 20 | 4 | 4.03 | 4 |
| >= 600k | $50 | 2 | 0 | 0.42 | 0 |
(trades up to 12 Oct 12:00; 2 later 0.0x rows excluded as unsettled)

- Under the same cap bucket, $10 and $50 give the same mean (0.37 vs 0.34). Size changed nothing.
- Fills identical: fill_vs_sim ~1.0, stop slip ~0 on both sizes. No size-driven slippage.
- $50 window: 7 Oct 18:14 to 8 Oct 16:04, 13 trades, 11 at cap <600k, where 0 of 41 trades (any size) ever ran. Best peak 3.2x. 8 Oct was a no-runner day.
- Two $50 rugs (retry_unsold, 8 Oct 13:05/13:12) cost 0.039 ETH of the 0.166 loss; ten stops at ~0.5x cost the rest.
- All 4 runners (29x, 19x, 13x, 13x; all held to 4h) were bought at cap 604k-979k. Entry cap >=600k: 4 of 22 ran (18%), mean ~3.9x; <600k: 0 of 41, mean 0.36x.
- Caveat: n=22 in the top bucket, 4 winners; holdout needed. Chance of 0 runners in 13 at the overall 8% rate = 34%, so the $50 loss is not unusual.

## Implications
1. Do not read the $50 result as "size hurts". Size review should wait for a cap rule, not a size rule.
2. Candidate rule (one, not a grid): mc5 buys only at cap >=600k. Would have skipped 41 trades (net about -0.12 ETH incl. rugs) and kept all 4 runners. Needs forward/shadow check: does it hold on 11-13 Oct (3 later runners? shadow n=11 mean 0.43)? This overlaps killed "first-minute-filters"; but those were different features. Asking research to ledger it and test on the shadow lane.
3. Without the 4 runners the strategy is stop-0.5x losers (mean ~0.4x), so the edge is the runner tail; entry at the higher cap is a proxy for later, more confirmed momentum.