Records (VALID rows with 5y forward returns + all primitives): 342

==============================================================================================================
 HEAD-TO-HEAD: 8 value signals on identical data
 Tercile method: top-1/3 cheap → predict outperform, bottom-1/3 expensive → underperform
 Bonferroni correction: p_bonf = min(p_raw × 8, 1.0)
==============================================================================================================
  P/E             n= 220  acc=49.5%  naive CI [ 43%,  56%]  year-clustered CI [ 45%,  54%]  p_raw=0.5801  p_bonf=1.0000
  EV/EBITDA       n= 220  acc=50.5%  naive CI [ 44%,  57%]  year-clustered CI [ 44%,  57%]  p_raw=0.4731  p_bonf=1.0000
  P/B             n= 204  acc=46.6%  naive CI [ 40%,  53%]  year-clustered CI [ 40%,  54%]  p_raw=0.8532  p_bonf=1.0000
  FCF yield       n= 194  acc=53.6%  naive CI [ 47%,  61%]  year-clustered CI [ 47%,  60%]  p_raw=0.1753  p_bonf=1.0000
  Earnings yld    n= 220  acc=49.5%  naive CI [ 43%,  56%]  year-clustered CI [ 45%,  54%]  p_raw=0.5801  p_bonf=1.0000
  EV/NOPAT        n= 220  acc=51.4%  naive CI [ 45%,  58%]  year-clustered CI [ 45%,  56%]  p_raw=0.3681  p_bonf=1.0000
  OE-DCF MoS      n= 220  acc=51.4%  naive CI [ 45%,  58%]  year-clustered CI [ 47%,  56%]  p_raw=0.3681  p_bonf=1.0000
  Brina Gap       n= 220  acc=57.3%  naive CI [ 51%,  64%]  year-clustered CI [ 51%,  63%]  p_raw=0.0182**  p_bonf=0.1455

==============================================================================================================
 POWER ANALYSIS (minimum detectable accuracy above chance)
==============================================================================================================
  P/E             n= 220  acc=49.5%  MDE_at_80%_power=58.4%    above MDE
  EV/EBITDA       n= 220  acc=50.5%  MDE_at_80%_power=58.4%    above MDE
  P/B             n= 204  acc=46.6%  MDE_at_80%_power=58.7%    above MDE
  FCF yield       n= 194  acc=53.6%  MDE_at_80%_power=58.9%    above MDE
  Earnings yld    n= 220  acc=49.5%  MDE_at_80%_power=58.4%    above MDE
  EV/NOPAT        n= 220  acc=51.4%  MDE_at_80%_power=58.4%    above MDE
  OE-DCF MoS      n= 220  acc=51.4%  MDE_at_80%_power=58.4%    above MDE
  Brina Gap       n= 220  acc=57.3%  MDE_at_80%_power=58.4%    above MDE

==============================================================================================================
 SAME COMPARISON, split by long-side / short-side
==============================================================================================================
  P/E             total n=220  acc=49.5% [ 43%,  56%] p=0.5801      LONG (cheap) n=110 acc= 51%     SHORT (expensive) n=110 acc= 48%
  EV/EBITDA       total n=220  acc=50.5% [ 44%,  57%] p=0.4731      LONG (cheap) n=110 acc= 51%     SHORT (expensive) n=110 acc= 50%
  P/B             total n=204  acc=46.6% [ 40%,  53%] p=0.8532      LONG (cheap) n=102 acc= 51%     SHORT (expensive) n=102 acc= 42%
  FCF yield       total n=194  acc=53.6% [ 47%,  61%] p=0.1753      LONG (cheap) n= 97 acc= 56%     SHORT (expensive) n= 97 acc= 52%
  Earnings yld    total n=220  acc=49.5% [ 43%,  56%] p=0.5801      LONG (cheap) n=110 acc= 51%     SHORT (expensive) n=110 acc= 48%
  EV/NOPAT        total n=220  acc=51.4% [ 45%,  58%] p=0.3681      LONG (cheap) n=110 acc= 51%     SHORT (expensive) n=110 acc= 52%
  OE-DCF MoS      total n=220  acc=51.4% [ 45%,  58%] p=0.3681      LONG (cheap) n=110 acc= 53%     SHORT (expensive) n=110 acc= 50%
  Brina Gap       total n=220  acc=57.3% [ 50%,  64%] p=0.0182 **     LONG (cheap) n=110 acc= 61%     SHORT (expensive) n=110 acc= 54%

==============================================================================================================
 INTERPRETATION GUIDE
==============================================================================================================
 - Tercile-based signal test is the conventional factor-portfolio framework.
 - All signals tested on the same set of underlying observations.
 - p_raw: naive binomial p-value (treats observations as i.i.d.).
 - p_bonf: Bonferroni-corrected for 8 simultaneous tests.
 - naive CI: per-observation bootstrap (anticonservative; assumes independence).
 - year-clustered CI: per-year bootstrap (proper inference under within-year
   correlation in stock returns; the methodologically correct interval).
 - MDE = minimum detectable effect size at α=0.05 (one-sided) and β=0.20.
   Signals with acc < MDE are underpowered at this sample size.
 - Practitioner takeaway: a signal that survives Bonferroni AND has a
   year-clustered CI excluding 50% is robust to standard reviewer objections.
