Data folder: /home/hatch/workspace/maineblock-products/data 

Step 1. The reduced grid of forty variants, checked against the registry
Variants run: 40. Every row's month counts match the registry exactly.
Largest absolute difference against the registry: 4.95e-09 in a monthly mean, 4.80e-05 in a t statistic.
V191 (survivor, long-only top quartile, reversal, lookback 2, no skip): in sample 58 months, mean +0.4114 percent a month, t = 2.46, Sharpe 1.12; holdout mean +0.0433 percent, t = 0.18.

Step 2. Full-grid counts from the registry file
Variants positive in sample: 110 of 200. Point-in-time family: 47 of 100. Survivor family: 63 of 100.
Point-in-time family: largest in-sample |t| is 1.34 (V081).
Survivor family: largest in-sample |t| is 2.46 (V191).
Passes at the 5 percent level in sample: 1 (V191). Passes at the 1 percent level: 0 (the null expectation is two).

Step 3. The winner and its point-in-time twin
V191 holdout: mean +0.0433 percent a month, t = 0.18, Sharpe 0.07; holdout t rank 109 of 200.
V091, the same rule on the point-in-time universe: in sample +0.1615 percent, t = 0.78, Sharpe 0.35; holdout -0.0023 percent, t = -0.01.
Correlation of in-sample and holdout Sharpe ratios across the 200: -0.08.
Median variant: in-sample Sharpe 0.04, holdout Sharpe 0.09. Holdout positives: 124 of 200.
The twenty best in-sample variants average a holdout Sharpe of +0.04; the remaining 180 average +0.11.

Step 4. Multiplicity corrections
Expected false winners under the null: 200 x 0.05 = 10 (the point-in-time family alone: 5).
Bonferroni per-test level 0.00025; critical |t| at 57 degrees of freedom: 3.91. Variants passing: 0.
V191 raw p = 0.0169; Bonferroni-adjusted p = 1.0.
Benjamini-Hochberg at 5 percent: rejections 0. Closest approach at rank 1: the smallest p is 0.0169 against a line of 0.00025, short by 0.0166.

Step 5. Deflated Sharpe ratio for V191
Winner monthly Sharpe 0.3232; skewness -0.039; kurtosis 2.777 (not excess).
Cross-sectional standard deviation of the 200 monthly Sharpe ratios: 0.1000. Expected maximum under the null: 0.2764 monthly, 0.96 annualized.
Deflated Sharpe ratio: 0.634. Unadjusted probabilistic Sharpe against zero: 0.991.

Part B. Solution to the R companion lane: extending the grid by a second skip month
Before the extension: 40 tests, 1 passes at 5 percent, expected false winners 2.0.
The extension adds 40 tests and 0 passes. After: 80 tests, 1 passes, expected false winners 4.0.

Chapter 13 lab complete.
