HKUDS/Vibe-Trading · error · ValueError

pd must be in (0.0, 1.0), got {pd}

Error message

pd must be in (0.0, 1.0), got {pd}

What it means

vasicek_credit_var evaluates the inverse normal CDF at pd (norm.ppf(pd)), so pd must be strictly inside (0.0, 1.0): at 0 or 1 the inverse is infinite and the Vasicek quantile formula is undefined.

Source

Thrown at agent/src/quantlib/credit.py:934

        dict with keys:
            * ``expected_loss`` (float): Base expected loss (EL).
            * ``wcdr`` (float): Worst-case conditional default rate at confidence.
            * ``worst_case_loss`` (float): Total portfolio loss at confidence (WCL).
            * ``unexpected_loss`` (float): Economic capital / Credit VaR (WCL - EL).
            * ``capital_ratio`` (float): Capital required as decimal fraction of EAD.

    Raises:
        ValueError: If parameters violate domain constraints.
    """
    ead = _require_finite(ead, "ead")
    pd = _require_finite(pd, "pd")
    lgd = _require_finite(lgd, "lgd")
    asset_correlation = _require_finite(asset_correlation, "asset_correlation")
    confidence = _require_finite(confidence, "confidence")
    if ead <= 0.0:
        raise ValueError(f"ead must be strictly positive, got {ead}")
    if not (0.0 < pd < 1.0):
        raise ValueError(f"pd must be in (0.0, 1.0), got {pd}")
    if not (0.0 <= lgd <= 1.0):
        raise ValueError(f"lgd must be in [0.0, 1.0], got {lgd}")
    if not (0.0 <= asset_correlation < 1.0):
        raise ValueError(f"asset_correlation must be in [0.0, 1.0), got {asset_correlation}")
    if not (0.0 < confidence < 1.0):
        raise ValueError(f"confidence must be in (0.0, 1.0), got {confidence}")

    rho = asset_correlation
    inv_pd = float(norm.ppf(pd))
    inv_conf = float(norm.ppf(confidence))

    numerator = inv_pd + np.sqrt(rho) * inv_conf
    denominator = np.sqrt(1.0 - rho)
    wcdr = float(norm.cdf(numerator / denominator))

    el = expected_loss(ead, pd, lgd)
    wcl = float(ead * lgd * wcdr)
    ul = float(max(0.0, wcl - el))

View on GitHub (pinned to 80ffdda44c)

Solutions

1. Convert units to decimals (percent/100, bps/10000)
2. Floor/ceiling borderline values, e.g. pd = min(max(pd, 1e-6), 1 - 1e-6)
3. Exclude pd == 0 (no default risk) or pd == 1 (already defaulted) names from the VaR batch

Example fix

# before
var = vasicek_credit_var(1e6, pd=0.0, lgd=0.6, asset_correlation=0.2, confidence=0.999)

# after
var = vasicek_credit_var(1e6, pd=1e-6, lgd=0.6, asset_correlation=0.2, confidence=0.999)

pd = min(max(pd, 1e-6), 1.0 - 1e-6)
if not 0.0 < pd < 1.0:
    raise ValueError(f"pd must be interior: {pd}")
var = vasicek_credit_var(ead, pd, lgd, rho, conf)

def is_interior_probability(p: float) -> bool:
    return isinstance(p, (int, float)) and 0.0 < float(p) < 1.0

try:
    var = vasicek_credit_var(ead, pd, lgd, rho, conf)
except ValueError as e:
    if 'pd' in str(e):
        var = 0.0 if pd <= 0 else ead  # degenerate cases
    else:
        raise

When it happens

Trigger: Calling vasicek_credit_var with pd = 0.0, pd = 1.0, pd = 2.5, or percent/bps values like 2 (for 2%).

Common situations: Default-free entities (pd = 0) in a portfolio batch; unit errors (2 instead of 0.02); already-defaulted names (pd = 1) passed through from a ratings table.

Related errors

• ead must be strictly positive, got {ead}
• asset_correlation must be in [0.0, 1.0), got {asset_correlat
• confidence must be in (0.0, 1.0), got {confidence}
• pd must be in [0.0, 1.0], got {pd}
• survival_prob must be in (0.0, 1.0], got {survival_prob}

AI-assisted analysis of HKUDS/Vibe-Trading@80ffdda44c (2026-08-28). Data as JSON: /api/errors/7a2b4e0e53fadda6. Report an issue: GitHub.