Frozen presets, live replacement ON. USD net c50 PnL; numbers in parentheses are accepted legs. Windows use entry time. Each member stops at 6 Oct 12:00 UTC minus its own frozen horizon; exits of those entries are included. The pre-25 Sep maturity boundary enters seen OOS. Seen OOS and forward are descriptive only. The last window holds only entries that matured by the cut: each member stops at 6 Oct 12:00 UTC minus its own horizon (up to 10 h), so 5 Oct 14:05→6 Oct 12 is a partial paper-period cohort. Live-fleet reference (Replace live fleet) window: 2026-09-16 00:00:00 → 2026-10-06 22:00:00 UTC from the RH replica, last fill 2026-10-06 20:44:01 UTC. Continuity against the published 4 Oct page: all 178,978 previously complete physical legs reappear with identical entries; 32 of them (five tokens, entries 1–3 Oct) have a different frozen exit in the longer replay, which is measured, not refit; its mechanism is not proven. Prefix PnL change per preset: f_primary -19, f_secondary -27, caps_best40 -110, caps_simple40 -350, fee0_best40_nocaps -30, equal_40k -51 USD; closed windows up to 3 Oct 10:45 move by at most $12 with identical trade counts. 1,878 legs that were in wind-down at the 4 Oct 06:00 cut are now complete cohorts, which is why the 3 Oct 10:45→4 Oct 06 column grew. Source: RH tape rh-20261006T115959Z (replica-only, FINAL-free; final-24h parity passed initial and delayed with the query cache off).
| Preset | IS · mature by 25 Sep | seen OOS · boundary→29 Sep 07 | forward · 29 Sep 07→6 Oct 12 | 30 Sep 18→1 Oct 07 | 1 Oct 07→18 | 1 Oct 18→2 Oct 01 | 2 Oct 01→3 Oct 10:45 | 3 Oct 10:45→4 Oct 06 | 4 Oct 06→5 Oct 14:05 · pre paper reset | 5 Oct 14:05→6 Oct 12 · paper period |
|---|---|---|---|---|---|---|---|---|---|---|
| f_primary | +37,678 (12735) | +12,763 (7568) | +15,822 (7707) | +2,920 (621) | +3,149 (520) | +918 (282) | +6,568 (1737) | +5,442 (958) | +976 (1473) | -1,900 (648) |
| f_secondary | +38,474 (42905) | +14,147 (25473) | +14,617 (25491) | +2,632 (2063) | +3,109 (1786) | +1,049 (961) | +6,182 (5897) | +5,441 (3183) | +1,525 (4610) | -1,647 (2061) |
| caps_best40 | +42,406 (22327) | +7,731 (13854) | +5,316 (15315) | +2,563 (1341) | +3,369 (1152) | +1,227 (645) | +3,311 (3378) | +6,278 (2092) | +531 (2781) | -4,566 (1324) |
| caps_simple40 | +44,670 (4598) | +34 (2596) | +3,281 (2706) | +1,345 (225) | +1,957 (196) | +991 (110) | +4,861 (631) | +5,541 (328) | -2,476 (490) | -2,514 (234) |
| fee0_best40_nocaps | +44,562 (20655) | +9,409 (12031) | +10,416 (12083) | +2,581 (956) | +2,490 (809) | +903 (436) | +5,277 (2762) | +4,588 (1584) | +114 (2332) | -1,652 (1086) |
| equal_40k | +42,563 (46498) | +11,786 (27304) | +12,867 (27442) | +2,317 (2190) | +2,816 (1828) | +913 (1007) | +5,477 (6272) | +5,071 (3544) | +885 (5226) | -1,675 (2432) |
Scale multiplies each leg's 005 notional. Sleeve cap = that member's own live-capital budget (blank = none). Solid lines = u1 (all tokens), dashed = u2 (hooked only).
Live trades are the real on-chain swaps our production RH wallets made (traders 44–51, read from the research replica), grouped into cells of (pool, UTC minute, side). If this portfolio were deployed it would replace that fleet, so those swaps would not happen next to ours.
On (replacement): in every cell, the first L dollars of our same-side flow (L = what the live fleet already bought, or sold, in that pool and minute) pay no new own impact — the price had already moved by that much in the history the backtest reads. Only impact above L is charged. Off (additive): our impact is charged in full on top of the live fleet's (the 005 convention) — that double counts the pool move the live fleet caused.
Why it matters: the backtest's prices already contain the live fleet's footprint; charging our full impact on top would bill the same pool move twice. Caveats: same pool and same minute only (other pools of the token get no credit); opposite-side live flow never offsets; the order of our fill vs theirs inside the minute is unknown (replacement assumes ours replaces theirs); hooked-v4 pools, where most of these sleeves trade, are rarely touched by the live fleet, so the credit is small. The KPI "live offset" shows PnL(on) − PnL(off) for the current settings.
Joint = this member's legs inside the portfolio (crowding, caps and live offset applied). 005 solo = the member's own 005 book at scale 1 (test window, no router fee; the page reproduces it exactly at λ any / offset off / fee 0 with only that member on).