quantflow/options/ssvi.py:160

• fit() uses np.interp(0.0, k_arr, w_obs) and then runs least-squares without checking that inputs are non-empty, same-shape, and sorted by k. np.interp requires an increasing x-grid, and empty/mismatched inputs will currently produce nan initial guesses or broadcast errors.

        k_arr = np.asarray(k, dtype=float)
        iv_arr = np.asarray(iv, dtype=float)
        w_obs = iv_arr**2 * ttm

        atm_var = float(np.interp(0.0, k_arr, w_obs)) if k_arr.size else w_obs.mean()
        x0 = [0.0, 1.0, max(atm_var, 1e-4)]

quantflow/options/ssvi.py:213

• fit_surface() collects slices without validating/sorting each (k, iv) pair. This can break the ATM interpolation (np.interp expects sorted k) and can yield nan initial thetas if a slice is empty.

        data = []
        thetas0 = []
        for k, iv, ttm in slices:
            k_arr = np.asarray(k, dtype=float)
            iv_arr = np.asarray(iv, dtype=float)
            w_obs = iv_arr**2 * ttm
            data.append((k_arr, w_obs))
            atm = float(np.interp(0.0, k_arr, w_obs)) if k_arr.size else w_obs.mean()
            thetas0.append(max(atm, 1e-4))

quantflow/options/ssvi.py:328

• fit_surface() doesn’t validate that each slice has non-empty k/iv arrays of matching shape. Empty slices will yield nan initial thetas and mismatched shapes can broadcast, producing incorrect residual vectors.

        for k, iv, ttm in slices:
            k_arr = np.asarray(k, dtype=float)
            iv_arr = np.asarray(iv, dtype=float)
            w_obs = iv_arr**2 * ttm
            data.append((k_arr, w_obs, float(ttm)))

quantflow/options/ssvi.py:276

• fit_surface() uses ttm directly without validating it is strictly positive, and uses np.interp(…, k_arr, …) without sorting k_arr. If ttm <= 0 this will divide by zero in the residual normalisation, and if k_arr is not monotone increasing np.interp can produce incorrect ATM variance seeds.

        for k, iv, ttm in slices:
            k_arr = np.asarray(k, dtype=float)
            iv_arr = np.asarray(iv, dtype=float)
            if k_arr.size == 0 or iv_arr.size == 0:
                raise ValueError("k and iv must contain at least one quote")
            if k_arr.shape != iv_arr.shape:

quantflow/options/ssvi.py:262

• This docstring uses a fully-qualified mkdocstrings cross-reference to VarianceCurve, but the target is in the same module. The repo instructions prefer relative cross-references ([...][.VarianceCurve]) when the target is in scope.

        monotone [VarianceCurve][quantflow.options.ssvi.VarianceCurve]. The

quantflow/options/ssvi.py:352

• fit_vol_surface() can return cls.fit_surface([]) when there are no converged options across all maturities, which raises a generic "at least one maturity slice" error. Adding an explicit check here yields a clearer error for callers (including the API endpoint).

        slices = []
        for index, maturity in enumerate(surface.maturities):
            options = list(surface.option_prices(index=index, converged=True))
            if not options:
                continue

quantflow/options/ssvi.py:209

• The iv() docstring says the result has the same shape as k, but the implementation returns the NumPy-broadcasted shape of k and ttm (e.g. scalar k with vector ttm returns a vector). This is misleading for API consumers.

        r"""Implied volatility $\sigma(k, \tau) = \sqrt{w(k, \tau) / \tau}$.

        Returns an array of the same shape as $k$. The SSVI total variance is
        strictly positive for $|\rho| < 1$, so no clipping is required.
        """