Backtesting Volume Confirmation on 15,900 A-Share Stock-Days

Description and use

Backtesting the rule that a strong up day on weak volume predicts a good next‑day return on 15,900 A‑share days revealed that, without accounting for monthly drift, the rule appears profitable. After adjusting for drift, all seven setups flip sign, leaving the baseline win rate of 47.3 % with a negative mean as the only reliable metric.

Problem

Retail traders rely on the rule that a strong up day on weak volume signals a good next‑day return, but its validity in real data was unclear.

Outcome

The raw rule shows positive returns on limit‑ups and negative on weak‑volume up days, but when adjusted for monthly drift all seven setups flip sign. The baseline win rate is 47.3% with a negative mean, indicating no real edge.

Workflow

• 1. Pull daily bars for 584 stocks from free public endpoints.
• 2. Compute today’s return, volume ratio (vs 5‑day average), and next‑day return for each day with ≥6 prior bars.
• 3. Aggregate results by calendar month and by rule buckets (price move & volume level).
• 4. Calculate baseline mean next‑day return for all days.
• 5. For each bucket, compute mean next‑day return and subtract the month’s market drift to obtain excess return.
• 6. Compare excess signs across consecutive months; discard setups that flip sign.
• 7. Report baseline, excess, and filtered results.
• 8. Publish full script on GitHub for replication.

Evidence

Everything below is computed from that. No synthetic data, no curve fitting.

Source

All seven setups flip.

Source