international-parity gold (ex-duty) — ground truth is Yahoo GC=F×INR÷31.1035×10, ~14.7% BELOW real duty-paid MCX. NOT a real Indian retail/MCX price.
Basis proxy error — near-band directional disagreement vs a domestic duty-paid proxy is ~53%, an upper bound (inflated by an ~11h phase mismatch). Do not read near-band directional accuracy without this caveat.
Structural up-tilt — the ~0.37–0.38 UP share is real drift under a zero-centered band, disclosed not corrected; it is why climatology’s majority is UP. Beating “always-up = climatology” means clearing the 3-class bar 0.365 (1d) / 0.384 (5d) — the directional 0.5322/0.5547 shares are diagnostic-only, never a skill bar.
5d autocorrelation — all 5d stats use effective-N = raw÷5 + block bootstrap; the raw 5d persistence 0.782 is an overlap artifact, not an edge.
Theme cells are diagnostic only (family-corrected, BH-FDR q=0.10); only Overall-1d can ever declare skill.
Skill claim embargoed · burn-in
Accumulating — n=6, embargoed until the criterion window closes. No skill is claimed until the Overall-1d Wilson lower bound exceeds 0.365 and the contemporaneous majority-class hit at 4 consecutive ISO-week (IST) checkpoints, each with cumulative N≥30. Consecutive passing checkpoints so far: 0/4.
Overall-1d hit-rate
—
insufficient · n=6 (<10)
Overall-5d hit-rate
—
insufficient · n=7 (<10)
Brier skill (1d)
+0.44
vs climatology-UP reference
Cumulative capture
-0.002
6 scored · 0L/4F/2S
Emission completeness
6/41
0 pending · 34 defect day(s)
Frozen outcomes
56
immutable records loaded
Calibration & skill charts
Reliability (top-label, 5 bins)
Overall-1d · ECE gated n≥50
Diagonal = perfect calibration. ECE: insufficient N.
Cumulative capture (Overall-1d)
∑ signed log-return in the direction taken
Long = +r, short = −r, flat = 0. Diagnostic of directional value; not a tradeable P&L (parity series, ex-costs).
Rolling Overall-1d accuracy vs baselines
cumulative 3-class hit-rate
Gold = cumulative hit-rate; dashed grey = contemporaneous always-up (adaptive guard); flat dashed line = frozen bar 0.365. A skill claim needs the Wilson LB above BOTH.
Scopes — 12 cells (overall + 5 themes × {1d,5d})
Scope
Horizon
N
Hit-rate
95% CI
Skill
Bar (max_ref)
Status
Overall
1d
6
insufficient (n<10)
—
—
0.365
insufficient
Overall
5d
7 (eff 1.4)
insufficient (n<10)
—
—
0.384
insufficient
Technical
1d
6
insufficient (n<10)
—
—
0.365
diagnostic · FDR
Technical
5d
6 (eff 1.2)
insufficient (n<10)
—
—
0.384
diagnostic · FDR
Macro
1d
6
insufficient (n<10)
—
—
0.365
diagnostic · FDR
Macro
5d
7 (eff 1.4)
insufficient (n<10)
—
—
0.384
diagnostic · FDR
Geopolitics
1d
3
insufficient (n<10)
—
—
0.365
diagnostic · FDR
Geopolitics
5d
5 (eff 1.0)
insufficient (n<10)
—
—
0.384
diagnostic · FDR
Demand
1d
4
insufficient (n<10)
—
—
0.365
diagnostic · FDR
Demand
5d
5 (eff 1.0)
insufficient (n<10)
—
—
0.384
diagnostic · FDR
Strategy
1d
5
insufficient (n<10)
—
—
0.365
diagnostic · FDR
Strategy
5d
6 (eff 1.2)
insufficient (n<10)
—
—
0.384
diagnostic · FDR
Only the two Overall rows are confirmatory; the 10 theme rows are permanently diagnostic (grey) and BH-FDR family-corrected. Skill = (hit − max_ref) / (1 − max_ref); shown only at full gate (n≥30) and, for Overall, only once the pre-registered criterion is met.
Forecast log
Issued
Scope
H
Call
Conf
Outcome
19 Jul
Technical
1d
flat
0.35
✗ up · +1.72% prov
19 Jul
Technical
5d
down
0.40
✗ up · +1.88% prov
19 Jul
Strategy
1d
flat
0.35
✗ up · +1.72% prov
19 Jul
Strategy
5d
flat
0.35
✗ up · +1.88% prov
19 Jul
Overall
1d
flat
0.37
✗ up · +1.72% prov
19 Jul
Overall
5d
up
0.38
✓ up · +1.88% prov
19 Jul
Macro
1d
flat
0.40
✗ up · +1.72% prov
19 Jul
Macro
5d
up
0.40
✓ up · +1.88% prov
19 Jul
Geopolitics
5d
up
0.40
✓ up · +1.88% prov
19 Jul
Demand
5d
up
0.35
✓ up · +1.88% prov
18 Jul
Technical
1d
up
0.38
✓ up · +1.67%
18 Jul
Technical
5d
down
0.40
✗ up · +1.22%
18 Jul
Strategy
1d
flat
0.40
✗ up · +1.67%
18 Jul
Strategy
5d
down
0.38
✗ up · +1.22%
18 Jul
Overall
1d
flat
0.45
✗ up · +1.67%
18 Jul
Overall
5d
down
0.40
✗ up · +1.22%
18 Jul
Macro
1d
flat
0.50
✗ up · +1.67%
18 Jul
Macro
5d
down
0.42
✗ up · +1.22%
18 Jul
Geopolitics
5d
up
0.33
✓ up · +1.22%
17 Jul
Technical
1d
down
0.48
✗ flat · +0.01%
17 Jul
Technical
5d
down
0.42
✗ up · +1.23%
17 Jul
Strategy
1d
down
0.38
✗ flat · +0.01%
17 Jul
Strategy
5d
down
0.35
✗ up · +1.23%
17 Jul
Overall
1d
down
0.43
✗ flat · +0.01%
17 Jul
Overall
5d
down
0.39
✗ up · +1.23%
17 Jul
Macro
1d
flat
0.35
✓ flat · +0.01%
17 Jul
Macro
5d
up
0.38
✓ up · +1.23%
17 Jul
Demand
1d
flat
0.40
✓ flat · +0.01%
17 Jul
Demand
5d
flat
0.35
✗ up · +1.23%
16 Jul
Technical
1d
flat
0.42
✗ up · +0.84%
16 Jul
Technical
5d
up
0.48
✓ up · +1.95%
16 Jul
Strategy
1d
flat
0.45
✗ up · +0.84%
16 Jul
Strategy
5d
up
0.50
✓ up · +1.95%
16 Jul
Overall
1d
flat
0.42
✗ up · +0.84%
16 Jul
Overall
5d
up
0.51
✓ up · +1.95%
16 Jul
Macro
1d
flat
0.50
✗ up · +0.84%
16 Jul
Macro
5d
up
0.55
✓ up · +1.95%
16 Jul
Geopolitics
1d
flat
0.35
✗ up · +0.84%
16 Jul
Geopolitics
5d
flat
0.33
✗ up · +1.95%
16 Jul
Demand
1d
flat
0.40
✗ up · +0.84%
16 Jul
Demand
5d
flat
0.45
✗ up · +1.95%
15 Jul
Technical
5d
up
0.40
✓ up · +2.06%
15 Jul
Strategy
5d
up
0.38
✓ up · +2.06%
15 Jul
Overall
5d
up
0.41
✓ up · +2.06%
15 Jul
Macro
5d
up
0.45
✓ up · +2.06%
14 Jul
Technical
1d
down
0.52
✗ flat · +0.14%
14 Jul
Technical
5d
down
0.55
✗ flat · +0.71%
14 Jul
Strategy
1d
down
0.45
✗ flat · +0.14%
14 Jul
Strategy
5d
down
0.48
✗ flat · +0.71%
14 Jul
Overall
1d
down
0.46
✗ flat · +0.14%
14 Jul
Overall
5d
down
0.50
✗ flat · +0.71%
14 Jul
Macro
1d
down
0.42
✗ flat · +0.14%
14 Jul
Macro
5d
down
0.48
✗ flat · +0.71%
14 Jul
Geopolitics
1d
flat
0.35
✓ flat · +0.14%
14 Jul
Geopolitics
5d
flat
0.35
✓ flat · +0.71%
14 Jul
Demand
1d
flat
0.35
✓ flat · +0.14%
14 Jul
Demand
5d
flat
0.35
✓ flat · +0.71%
13 Jul
Technical
1d
flat
0.40
✗ up · +1.81%
13 Jul
Overall
1d
flat
0.38
✗ up · +1.81%
13 Jul
Overall
5d
flat
0.32
✓ flat · +0.85%
13 Jul
Macro
1d
flat
0.45
✗ up · +1.81%
13 Jul
Macro
5d
flat
0.30
✓ flat · +0.85%
13 Jul
Geopolitics
1d
flat
0.35
✗ up · +1.81%
13 Jul
Geopolitics
5d
down
0.40
✗ flat · +0.85%
13 Jul
Demand
1d
flat
0.30
✗ up · +1.81%
13 Jul
Demand
5d
flat
0.35
✓ flat · +0.85%
Matured rows show ✓/✗ + realized class and % move; pending rows show their maturity date; defects/drops are shown, not hidden. “prov” = live-provisional (no frozen record yet).
Methodology & integrity
Frozen constants & checks
Flat band
FLAT_K=0.36
τ(1d)=0.36·σ1d(ref) rolling · τ(5d)=0.008933 const
σ1d window
60d sample (n−1)
σ5d=0.024814 (empirical, not σ1d·√5=0.026782)
Interval
skip-issuance-bar
1d close[ref+1]→[ref+2] · 5d close[ref+1]→[ref+6]
max_ref (bar)
0.365 / 0.384
3-class climatology = always-up (1d/5d)
Gates
low 10 · full 30
ECE at n≥50, 5 bins · ref_price tol 0.001
Criterion
4× ISO-week
cumulative Wilson LB > 0.365 & adaptive, cum N≥30, reset-on-fail
Trading-day integrity
OK
3 row(s) rejected by the trading-day integrity assertion (incl. 2019-10-27 weekend)
Leakage check
OK
8 live anchor(s) checked, none peeking
Recommended upgrade (not in frozen v2): switch ground truth to a real MCX ~23:30-IST settlement series under a new METHODOLOGY_VERSION. Research & education only.
Generated 10 Sep 2026, 23:37 IST · vedant.lodha.cloud