Scorecard
Graded by ORACLE's own grader: first touch on hourly highs/lows, costs included, expiry at the last close. Unvalidated (shadow), N=306.
9 graded calls excluded: target at or below zero (opened before the levels guard; owner decision 2026-10-12).
By asset class
| group | n | target hit | avg net |
|---|---|---|---|
| CRYPTO | 50 | 0% | -1.79R |
| EQUITY | 256 | 18% | -0.59R |
By direction
| group | n | target hit | avg net |
|---|---|---|---|
| LONG | 219 | 15% | -0.24R |
| SHORT | 87 | 14% | -2.17R |
By regime at the call
| group | n | target hit | avg net |
|---|---|---|---|
| pre-ledger | 19 | 0% | -1.01R |
| risk_on | 188 | 16% | -1.30R |
| neutral | 99 | 15% | +0.23R |
Is the prior calibrated?
| prior | n | mean prior | target hit |
|---|---|---|---|
| 50-60% | 145 | 56% | 12% |
| 60-70% | 142 | 63% | 19% |
The prior is a hand-set number from July 2026. This table is the evidence trial T-0001 uses to decide whether a calibrated mapping replaces it.
Benchmarks (M-0001)
selection shown · v1 stock longs since 2026-10-09: 148 calls on 7 open days beat a random same-day pick by +0.22R ±0.20 (t 2.23, permutation p 0.004). Registered first look; later figures are monitoring only.
| calls | vs random pick | vs average symbol | vs SPY / BTC |
|---|---|---|---|
| v1 crypto long 28 calls on 9 days | -0.15R ±0.24 | -0.82% ±2.26 | -3.75% ±2.67 |
| v1 crypto short 3 calls on 2 days | +0.20R ±0.11 | +1.36% ±0.25 | +2.61% ±0.57 |
| v1 stock long 176 calls on 9 days | +0.16R ±0.20 | +1.74% ±1.67 | +1.82% ±1.68 |
| v1 stock short 78 calls on 3 days | +0.06R ±0.39 | -0.43% ±5.07 | +3.14% ±8.59 |
| v2 crypto long 21 calls on 9 days | -0.05R ±0.11 | -1.13% ±2.79 | -5.70% ±3.58 |
| v2 crypto short 36 calls on 13 days | +0.04R ±0.15 | -0.03% ±1.37 | +1.26% ±2.64 |
| v2 stock long 245 calls on 14 days | +0.03R ±0.12 | +0.10% ±0.44 | +0.96% ±0.80 |
| v2 stock short 112 calls on 14 days | +0.12R ±0.25 | +0.13% ±1.36 | -0.56% ±1.09 |
- Gate, stocks: 254 taken, 324 blocked. Taken minus blocked on the 6 days with both: +0.38R (t 1.03).
- Gate, crypto: 31 taken, 57 blocked. No day has both yet.
Each finished call is re-graded with the engine's own bracket and compared with the same bracket on every other symbol scanned that day (a random pick), with the average scanned symbol, and with SPY or BTC, over its own holding window. The ± is two standard errors, counting each open day as one draw. Few days means wide error bars: treat every figure as fragile.