HKUDS/Vibe-Trading · error · ValueError

style drift needs at least 2 dates, got {exposure_history.sh

Error message

style drift needs at least 2 dates, got {exposure_history.shape[0]}

What it means

style_drift measures change in exposures over time and needs a time series; with fewer than 2 rows in exposure_history there is no possible delta, so it raises.

Source

Thrown at agent/src/quantlib/factormodel.py:542

def style_drift(exposure_history: pd.DataFrame) -> StyleDrift:
    """Summarise how a portfolio's exposures moved over time.

    Args:
        exposure_history: Rows indexed by date in chronological order, one
            column per factor, each cell a portfolio-level exposure from
            :func:`portfolio_style_exposure`.

    Returns:
        A :class:`StyleDrift`.

    Raises:
        ValueError: If fewer than two dates are supplied -- drift is a change,
            and one observation cannot express one.
    """
    if exposure_history.shape[0] < 2:
        raise ValueError(
            f"style drift needs at least 2 dates, got {exposure_history.shape[0]}"
        )

    frame = exposure_history.drop(columns=["unmatched_weight"], errors="ignore")
    return StyleDrift(
        mean_exposure=frame.mean(),
        std_exposure=frame.std(ddof=1),
        first_exposure=frame.iloc[0],
        last_exposure=frame.iloc[-1],
        total_change=frame.iloc[-1] - frame.iloc[0],
        max_abs_change=frame.diff().abs().max(),
    )


def factor_return_attribution(
    portfolio_exposures: pd.Series,
    factor_returns: pd.Series,
    portfolio_return: float,

View on GitHub (pinned to 80ffdda44c)

Solutions

1. Check exposure_history.shape[0] >= 2 before calling
2. Widen the date window used to build the history
3. If only one date exists, report zero drift instead of calling the API

Example fix

// before
drift = style_drift(hist)
// after
drift = style_drift(hist) if len(hist) >= 2 else None

assert exposure_history.shape[0] >= 2, 'need >= 2 dates for drift'

When it happens

Trigger: Calling style_drift with a one-row (or zero-row) history DataFrame, e.g. style_drift(exposure_history.iloc[[-1]]).

Common situations: Backfill produced only one rebalance date; date filtering (.loc[date:]) sliced history to a single point; new portfolio with one day of history.

Related errors

• holdings is empty
• exposures has no factor columns
• portfolio_weights cannot be empty
• portfolio_weights contains non-finite values
• exposures contains non-finite values

AI-assisted analysis of HKUDS/Vibe-Trading@80ffdda44c (2026-08-28). Data as JSON: /api/errors/5c597c2343876644. Report an issue: GitHub.