In Chinese A-shares, every indicator can be faked — except one.
K-lines can be drawn. MACD and KDJ can be painted. Even the "fundamentals" in a prospectus can be polished. But every single share of volume requires real money changing hands. You can't print liquidity.
I spent the last few months building an open-source A-Share scanner, and volume is the layer that keeps saving me from bad trades. Here's the playbook — built on what actually happened in the 2026 market, not textbook theory.
What volume actually measures
Most people think volume = supply and demand. It isn't. Volume is a measure of disagreement.
- Everyone agrees the stock is great → nobody sells → limit-up on tiny volume (price flies on almost no shares).
- Everyone is terrified and thrilled at the same time → massive turnover (someone who believes it's a 10-bagger selling to someone who believes it's a bankruptcy).
Every huge trade is one buyer and one seller, each convinced the other is an idiot. The question is never how much traded — it's which direction the chips moved.
Two historic blow-off days I watched
| Date | Turnover | What happened next |
|---|---|---|
| 2024-10-08 | 3.48 trillion yuan | Gap up, faded, violent correction |
| 2026-01-14 | 3.99 trillion yuan (record) | Closed red, volume dried up, months of sideways bleed |
Record volume is not a bull-market starting gun. It's a disagreement thermometer maxing out. When the whole market trades at record levels, the chips are changing hands at maximum speed — and someone with a huge position is usually on the sell side.
Where volume happens matters more than how much
High-position volume breakout = liquidity trap ⚠️
A main force that has held a stock for a year, doubled it, can't just dump — the stock would hit limit-down and they'd be stuck. So they wash-trade: account A places sell orders, account B eats them in seconds, drawing a giant "buying" volume bar on the daily chart. Retail and quant funds chase the breakout. The real chips get mixed into the exit.
If volume explodes after a stock has already doubled, ask: who is the seller?
High-position volume with no price progress = distribution ⚠️⚠️
Volume up 2–3x, price up 1% with a long upper shadow. Retail reads "shakeout." The math says otherwise: if real money were buying aggressively and the price can't rise, the sell side is effectively unlimited — major shareholders plus the bottom-position main force. This is the last distribution before the kite string breaks.
Low-position volume spike = accumulation ✅
Long decline, volume at ice-cold levels, then one day a volume bar far above the multi-month average. Retail is panic-selling; whoever can absorb hundreds of millions of yuan against the grain is almost certainly informed money. They accumulate quietly — that's why price often stays flat while volume builds (量增价平).
The counterintuitive side: shrinking volume
High-position shrink-up = locked chips (not always a bearish divergence)
Classic dogma says "a rally without volume is a scam." It misses the strongest stocks. When the main force controls 70–80% of the float, it takes just tens of millions of yuan to push a limit-up, and daily volume collapses. The tell is the trend slope: a steep 5/10-day MA with quickly recovered dips = locked chips; a choppy, one-step-back-three-step-forward drift = dying momentum.
Low-position shrink-down = the death spiral ⚠️⚠️⚠️
A volume-down decline at least has no buyers. A shrinking-volume decline means the market has completely given up. -1% a day, sawing your capital in half over six months. Never catch this knife — wait for the 地量 (extreme low volume) bottom to form first.
Turnover rate: volume, normalized
Absolute volume is meaningless. 50 million shares is a normal 5% turnover for a 10-billion-float stock and a 100% explosion for a 500-million float.
| Turnover | Zone | Meaning |
|---|---|---|
| <1% | Ice-cold | Forgotten stock; only actionable if chips are confirmed exhausted |
| 1–3% | Normal | Trend continues as-is; watch |
| 5–10% | Hot | Disagreement rising, main force footprints visible; start tracking |
| >15–20% | Extreme | Red alert at highs — 1 in 5 shares changed hands; distribution zone |
The eight classic volume-price patterns
| Pattern | At lows | At highs |
|---|---|---|
| Volume up, price flat | ✅ Accumulation | ⚠️⚠️ Distribution warning |
| Volume up, price up | ✅ Healthy (1–2x = golden) | ⚠️ 5–10x explosion = wash-trade bait |
| Volume flat, price up | ✅ Locked chips | ⚠️ Follow-on money drying |
| Volume down, price up | ✅ Controlled float | ⚠️ Liquidity exhaustion |
| Volume up, price down | ⚠️ Panic — exit signal anywhere | ❌ Huge red bar = trend over |
| Volume down, price down | ⚠️ Death spiral — no bottom-fishing | ⚠️ Wait for extreme low volume |
The one question to ask at every candle
Which direction are the chips moving?
- Main-force pocket → retail pocket (they sell, you buy) = danger
- Retail pocket → main-force pocket (they buy, you sell) = opportunity
Judge by the patterns above: high-position volume stalls → chips flowing to retail. Bottom 地量 followed by a low-position spike → chips flowing to the main force.
Turning this into code
Rules are only useful when they're enforced. I quantized these patterns into a scoring layer in the scanner:
def volume_score(high_pos: bool, vol_ratio: float, chg_pct: float, turnover: float):
s = 0
if high_pos and vol_ratio > 2 and chg_pct < 2:
s -= 4 # high-position volume stall = distribution
elif high_pos and vol_ratio > 3:
s -= 2 # liquidity-trap breakout
elif not high_pos and turnover < 0.5:
s += 2 # extreme low volume near bottom = setup
elif not high_pos and vol_ratio > 3:
s += 3 # low-position spike = accumulation
if turnover > 15:
s -= 3 # red-alert turnover
elif 5 <= turnover <= 10 and not high_pos:
s += 4 # main force ignition zone
return max(-5, min(5, s))
Combined with the technical (MACD/KDJ/RSI), fundamental (ROE/PEG) and sentiment layers of the three-pillar system, it's the difference between "the chart looks fine" and "the chips are flowing the wrong way."
Where to find it
The full scanner — including the volume scoring, a 5,000-stock daily screener, and the three-pillar framework — is open source: https://github.com/Felixwang007 (everything is MIT).
If you'd rather have it as a ready-made agent skill, search "A-Share Stock Analysis" on xiaping.coze.com — it runs the whole scan in 5 minutes, no SaaS subscriptions, no paid APIs.
Volume is the only honest participant in the market. Learn to read it, and the other indicators stop lying to you. Questions or a pattern you think I got wrong — drop it in the comments.
Top comments (0)