Alpaca 1Hour caches, ~2025-07..2026-09, 1-bar XS momentum, dollar-neutral K=20, OOS = last 40% of bars.
| Grid | names | bars/yr | gross SR | net1 | net2 | net10 | turn/bar |
|---|
| RTH all-eq | 501 | 1764 | 1.97 | 0.42 | -1.13 | -13.53 | 3.53 |
| RTH+pre | 501 | 3024 | -2.95 | -5.21 | -7.48 | -25.65 | 2.65 |
| RTH+post | 501 | 2766 | -4.67 | -7.16 | -9.64 | -29.22 | 2.95 |
| RTH+all | 501 | 4026 | -9.84 | -12.83 | -15.81 | -39.37 | 2.57 |
Same grids with the sign flipped (1-bar XS REVERSAL):
| Grid | names | bars/yr | gross SR | net1 | net2 | net10 | turn/bar |
|---|
| RTH all-eq (reversal) | 501 | 1764 | -1.97 | -3.52 | -5.07 | -17.44 | 3.53 |
| RTH+pre (reversal) | 501 | 3024 | 2.00 | -0.17 | -2.34 | -19.30 | 2.65 |
| RTH+post (reversal) | 501 | 2766 | 6.69 | 4.37 | 2.06 | -16.26 | 2.95 |
| RTH+all (reversal) | 501 | 4026 | 10.22 | 7.39 | 4.57 | -17.65 | 2.57 |
Per-bar gross (bps) of the RTH+all book, and the Roll half-spread implied by 1h returns on that segment's bars.
| Segment | bars | gross bps/bar | Roll ½-spread bps |
|---|
| pre 04-08 | 1250 | -6.813 | 13.58 |
| RTH 09-15 | 1750 | -0.921 | 7.36 |
| post 16-19 | 994 | -20.407 | 18.12 |
| Universe | names | gross SR | net1 | net2 | turn/bar |
|---|
| live 101 | 99 | 0.44 | -2.07 | -4.59 | 3.08 |
| top150 by $vol | 150 | 0.33 | -1.14 | -2.62 | 3.25 |
| top250 by $vol | 250 | 0.19 | -1.34 | -2.88 | 3.42 |
| all 501 | 501 | 1.97 | 0.42 | -1.13 | 3.53 |
| Book | names | bars | gross SR | net1 | net2 | turn/bar |
|---|
| eq+crypto RTH grid | 509 | 1750 | 2.43 | 0.87 | -0.69 | 3.47 |
| eq+crypto 24/7 grid | 509 | 8671 | 1.28 | -0.44 | -2.16 | 0.63 |
Read: compare the pre/post segment gross bps/bar against its Roll ½-spread. Extended-hours bars only help if gross clears the much wider pre/post spread; watch the net columns for the same. The universe table shows whether the RTH edge survives a wider name set (top-N ranked by median RTH dollar volume).