HKUDS/Vibe-Trading · error · ValueError

no event could be measured; reasons: {reasons}

Error message

no event could be measured; reasons: {reasons}

What it means

After per-event processing, if every event was dropped (window falls outside the data, estimation window has too few observations, etc.), event_study has no outcomes to aggregate and raises this error, enumerating each dropped event with its reason.

Source

Thrown at agent/src/quantlib/eventstudy.py:448

        outcomes.append(
            (
                EventOutcome(
                    symbol=symbol,
                    event_date=index[position],
                    abnormal_returns=pd.Series(abnormal, index=relative_days, name=symbol),
                    car=car,
                    car_std_error=car_se,
                    standardised_car=standardised,
                    fit=fit,
                ),
                est_residuals,
                abnormal,
            )
        )

    if not outcomes:
        raise ValueError(
            "no event could be measured; reasons: "
            + "; ".join(f"{s}@{d}: {r}" for s, d, r in dropped)
        )

    n_events = len(outcomes)
    event_outcomes = [o[0] for o in outcomes]
    ar_matrix = np.vstack([o.abnormal_returns.to_numpy() for o in event_outcomes])
    aar = pd.Series(ar_matrix.mean(axis=0), index=relative_days, name="aar")
    caar = pd.Series(np.cumsum(aar.to_numpy()), index=relative_days, name="caar")

    cars = np.array([o.car for o in event_outcomes])
    if n_events > 1:
        car_sd = float(cars.std(ddof=1))
        t_stat = float(cars.mean() / (car_sd / np.sqrt(n_events))) if car_sd > 0 else float("nan")
        t_p = float(2 * student_t.sf(abs(t_stat), df=n_events - 1)) if np.isfinite(t_stat) else float("nan")
    else:
        t_stat, t_p = float("nan"), float("nan")

View on GitHub (pinned to 80ffdda44c)

Solutions

1. Read the per-event reasons in the message: symbol@date: reason tells you exactly which constraint failed.
2. Trim events whose date is earlier than estimation_window + estimation_gap + |start| periods into the sample, or extend the returns history backwards.

Example fix

# before
events = {"AAPL": ["2010-01-05"]}  # too early for the sample
# after
min_days = estimation_window + estimation_gap + abs(event_window[0])
events = {s: [d for d in ds if returns.index.get_loc(d) >= min_days] for s, ds in events.items()}

min_days = estimation_window + estimation_gap + abs(event_window[0])
events = {s: [d for d in ds if returns.index.get_loc(d) >= min_days] for s, ds in events.items()}

try:
    res = event_study(...)
except ValueError as e:
    if "no event could be measured" in str(e):
        logger.warning("all events dropped: %s", e)
    else:
        raise

When it happens

Trigger: Event dates near the start of the sample such that the estimation window (event-relative days minus gap and window length) predates the first observation, or NaNs wiping out an estimation window for every event.

Common situations: Events at the very beginning of a backtest, a returns frame that starts later than expected after data-cleaning, or timezone shifts moving event dates outside the index.

Related errors

• estimation window needs at least {MIN_ESTIMATION_OBSERVATION
• market returns are constant over the estimation window, so b
• event_window start must be <= end, got {event_window}
• estimation_gap must be >= 0, got {estimation_gap}
• estimation_window must be at least {MIN_ESTIMATION_OBSERVATI

AI-assisted analysis of HKUDS/Vibe-Trading@80ffdda44c (2026-08-28). Data as JSON: /api/errors/667ea8d80a22b56f. Report an issue: GitHub.