95% block interval −0.002606 to +0.005568.
MODEL 2 · STEP 5Stress
Stress
the result.
Apply the same nine frozen gates to both phases. The signed challenger improves calibration and fold consistency, but clears only two gates.
MARKET CONTEXT · OUTSIDE MODEL FREEZEHOOD
These readings update independently. They are not model inputs, deployment evidence, or a trading recommendation.
- Latest price
- $122.11 · fallback
- 20D momentum
- +30.9% Close-to-close · 20 completed sessions
- RSI 14
- 65.9 neutral · Wilder method
- Relative volume
- 1.2× Versus the prior 20 completed sessions
TECHNICALS THROUGH · TIMESTAMPED FALLBACK
Does either development phase earn deployment consideration?
Phase A records the original five-candidate search. Phase B compares three bounded challengers and applies the same nine predeclared gates to selected Signed Core Ridge. Passing two isolated gates is progress, not deployment permission.
PHASE A · ORIGINAL DEVELOPMENT SEARCH
Cross Asset Ridge ranks first, but does not validate.
Selection means lowest observed Brier score among the five inspected candidates. It does not override uncertainty, discrimination, or calibration evidence.
Below the 0.50 no-skill reference.
Equals the always-up OOS class rate; fold-prior accuracy is 58.102%.
Slope; intercept 0.822 and five-bin ECE 0.111.
Benchmark-only recommendation.
Active prior controller; development candidate is flat at 57.508%.
PHASE A · PREDECLARED EVIDENCE GATES
Zero passes in the original search.
Block resampling respects overlapping five-session outcomes. The family-wise null also accounts for selecting the strongest of five Phase A development candidates.
| Gate | Rule | Observed | Required | Result |
|---|---|---|---|---|
| G1 | Brier improvement and its 20-session lower bound are positive | +0.001459; 95% CI −0.002606 to +0.005568 | Point > 0; lower > 0 | FAIL |
| G2 | ROC AUC and its block-bootstrap lower bound clear no-skill | 0.428325; lower 0.3457 | AUC ≥ 0.52; lower > 0.50 | FAIL |
| G3 | Family-wise circular-shift result | p = 0.8180 | p < 0.05 | FAIL |
| G4 | Selected candidate beats frozen Core Ridge robustly | +0.000051; lower −0.000260 | Point > 0; lower > 0 | FAIL |
| G5 | Calibration slope | −1.399207 | 0.70 to 1.30 | FAIL |
| G6 | Absolute calibration intercept | 0.822288 | ≤ 0.10 | FAIL |
| G7 | Five-bin expected calibration error | 0.110998 · 11.10% | < 5% | FAIL |
| G8 | Outer folds with positive Brier improvement | 6 / 11 · 54.5% | ≥ 60% | FAIL |
| G9 | Block-length and supported-regime sensitivity | 10d −0.002467; 20d −0.002606; 40d −0.002876 | All lower bounds > 0; no regime skill < −5% | FAIL |
0 of 9 gates pass. The probability ordering is reversed in aggregate, calibration is unstable, and the apparent Brier gain is not separated from zero under block uncertainty.
PHASE B · DEVELOPMENT CHALLENGER
Signed Core Ridge lowers retrospective Brier, not the bar.
The challenger reuses the same 685 dates already examined in Phase A. Its results are development evidence only—not a fresh holdout or deployment test.
Improvement +0.004739; 95% lower bound −0.002580.
Block-bootstrap lower bound 0.4347.
Slope; intercept 0.312902 and five-bin ECE 0.042856.
Gate G8 passes with 72.7% of folds positive.
G7 calibration error and G8 fold consistency pass.
Experimental challenger; active prior controller remains 55.745%.
PHASE B · SAME NINE GATES
Two pass. Seven still fail.
G7 confirms five-bin calibration error below 5%. G8 confirms positive Brier improvement in 8 of 11 folds. Uncertainty, discrimination, the family-wise null, calibration shape, and sensitivity still block deployment.
| Gate | Rule | Observed | Required | Result |
|---|---|---|---|---|
| G1 | Brier improvement and its 20-session lower bound are positive | +0.004739; 95% lower −0.002580 | Point > 0; lower > 0 | FAIL |
| G2 | ROC AUC and its block-bootstrap lower bound clear no-skill | 0.503030; lower 0.4347 | AUC ≥ 0.52; lower > 0.50 | FAIL |
| G3 | Family-wise circular-shift result | p = 0.6880 | p < 0.05 | FAIL |
| G4 | Challenger beats frozen Core Ridge robustly | +0.003331; lower −0.003398 | Point > 0; lower > 0 | FAIL |
| G5 | Calibration slope | 0.266303 | 0.70 to 1.30 | FAIL |
| G6 | Absolute calibration intercept | 0.312902 | ≤ 0.10 | FAIL |
| G7 | Five-bin expected calibration error | 0.042856 · 4.29% | < 5% | PASS |
| G8 | Outer folds with positive Brier improvement | 8 / 11 · 72.7% | ≥ 60% | PASS |
| G9 | Block-length and supported-regime sensitivity | 10d −0.002417; 20d −0.002580; 40d −0.001978 | All lower bounds > 0; no regime skill < −5% | FAIL |
Signed Core Ridge passes 2 of 9 gates. Its +0.004739 Brier improvement has a −0.002580 lower bound, AUC is 0.503030 with a 0.4347 lower bound, and the family-wise p-value is 0.6880. The same 685 dates remain development-only. G3 corrects the three-member family, while G4 is still only a pairwise comparison with frozen Core, so it cannot erase post-selection risk.
The high-volatility split uses the full out-of-sample median for evaluation only. It does not enter a forecast, but a future deployment protocol must freeze that threshold inside training history before reuse.
Alpaca consolidated stock bars · Alpaca US crypto bars · Alpaca corporate-action announcements · Robinhood SEC filings · Robinhood investor relations. The Excel file preserves the source ledger and frozen outputs.
NEXTStep 6 · Summary and recommendation
Continue through the frozen Model 2 workflow.