Read market temperature from advance/decline

1. Why look at advance/decline (30 seconds)

An index (e.g. Shanghai Composite) can be dragged by a few heavyweights; advance/decline reflects the whole market:

• Broad rally: far more advancers than decliners → warm sentiment (good money-making effect)
• Broad selloff: far more decliners → cold (loss-making effect)
• Split: roughly even → structural market (picking the right sector matters more than the index)
• Limit-ups: extreme sentiment — many limit-ups = active themes / euphoric money

2. Fetch the data (free endpoint)

`/v1/changedist` is a free endpoint (no Key). Returns two parts: breadth overview + percentage-change buckets, with `structured` for programmatic use.

import requests

r = requests.get("https://api.ashareapi.com/v1/changedist", timeout=30)
j = r.json()
s = j.get("structured")            # for code
print(s)                            # [{"上涨": "3006", "上涨占比": "53%", "下跌": "2303", ...}]

# data is Markdown (for humans); structured is a 2D array (for code)
tables = j.get("tables")            # [overview row, bucket rows...]
for t in tables:
    print(t)

• Free endpoint: no Key needed
• `structured` is a conditional field — only attached when data is Markdown (changedist qualifies)
• Use `structured` / `tables` for programmatic work, do not parse Markdown

3. Reading the fields

The buckets (>7% / 7-5% / 5-2% / 2-0% etc.) reveal strength — "3000 up" with lots of >5% gains is very different from "3000 up but mostly +1%".

4. How to read "market temperature"

• Never judge off one day — sentiment has inertia; track several days to find the turning point
• Overheat: high ratio + many limit-ups + rising turnover → possible sentiment top (pullback risk next day)
• Freezing: very low ratio + many limit-downs → often a sentiment bottom (bounce probability rises, but combine with other)

5. Live data (2026-09-29)

Read: 53% ratio + 42 limit-ups = "neutral-warm, some theme heat" — not freezing, not overheated. This lukewarm state especially needs the multi-day trend (next section).

6. Track several days and spot the turning point

import requests, time

def breadth():
    j = requests.get("https://api.ashareapi.com/v1/changedist", timeout=30).json()
    s = j["structured"][0]
    return int(s.get("上涨", 0)), int(s.get("下跌", 0)), int(s.get("涨停", 0))

# simulate logging (in practice store daily)
history = []
for day in range(5):
    up, down, zt = breadth()
    history.append((up, down, zt))
    time.sleep(1)

for i, (up, down, zt) in enumerate(history):
    ratio = up / (up + down) if (up + down) else 0
    print("day%d ratio %.0f%% limit-up %d" % (i + 1, ratio * 100, zt))

• Trend: ratio climbing → warming; falling → cooling
• Spike: jumps to >70% (with many limit-ups) → watch overheat; drops to <30% → watch freezing
• Discipline: breadth is a thermometer (describes state), not a predictor (does not guarantee reversal/continuation). Use it for environment, then set strategy preference (not a direct buy/sell)

7. Common misreads (important)

• ① Data basis: `/v1/changedist` is the current period. For a historical day use another means (this guide focuses on "today")
• ② Index vs breadth: the index may look fine due to heavyweights while the market is really split — breadth exposes that illusion
• ③ Many limit-ups ≠ necessarily euphoric: check persistence and theme concentration (scattered vs one sector clustering differ)