BUG: Formula used (d - sqrt_rho*z) but correct is (d + sqrt_rho*z)
- Our thresholds are cumulative ascending (AAA→CCC→D)
- Higher Z should push probability mass left (better ratings)
- Previous: Z+ = higher PD = bad economy (WRONG)
- Fixed: Z+ = lower PD = good economy (matches paper)
Verification:
- 1998 IMF crisis: Zt=-2.12 (negative = bad ✅)
- 2006 boom: Zt=+1.53 (positive = good ✅)
- Pipeline 8/8 validation pass
