ths-advanced-analysis

THS Advanced Analysis Skill

Dialogue guidance specifications

Clarify intent (must ask when intent is ambiguous)

User input is often imprecise; determine intent before calling,do not guess and run directly。

User says	Possible intent	Must ask
"Help me look at stock XX"	Real-time quote? K-line trend? Large orders?	✅
"Analyze XX"	Technical? Capital flow? Compare with whom?	✅
"How is sector XX doing"	Overall sector gain/loss? Constituent stocks? Leading stocks?	✅
"Pick some good stocks"	Short-term? Value? Which industry? Conditions?	✅
"5-minute K-line for XX"	Intent clear	❌ Execute directly
"Today's limit-up stocks"	Intent clear	❌ Execute directly

Clarification wording examples:

用户："帮我分析一下宁德时代"
Claude："好的，请问你主要想看哪个方向？
  1. 今日实时行情 + 资金流向
  2. 分钟K线（盘中走势）
  3. 近期日K线趋势
  4. 和比亚迪、亿纬锂能等对比
  5. 用问财筛选相关概念股"

Follow-up prompts after calling (only mention when there is extended value)

Do not mechanically list "you can also do XYZ" every time. Only bring it up naturally in the following situations:

Scenario	Appropriate follow-up prompt
Displayed industry ranking	"Do you need to check the quotes of constituent stocks in a certain industry?"
Displayed minute K-line	"Do you need to also view large order flow or order book depth?"
Displayed multi-stock comparison table	"Do you need to display a normalized trend chart or correlation?"
Wencai selected candidate stocks	"Do you need to perform K-line technical validation on these stocks?"
Displayed auction anomalies	"Do you need to pull the pre-market time series for a certain anomalous stock to see the details?"

Complete calling examples (directly runnable)

See examples/ directory for 4 end-to-end scenarios:

File	Scenario
`examples/01_minute_kline.py`	Minute K-line + moving averages + volume anomaly annotation
`examples/02_sector_industry.py`	Industry ranking + concept sector constituent stocks + index quotes
`examples/03_multi_stock_compare.py`	Multi-stock batch comparison: table + normalized trend + correlation
`examples/04_bigorder_auction.py`	Large order flow + auction anomaly scan + time series/order book
`examples/05_wencai_nlp.py`	Wencai NLP: stock selection/quotes/financials/technical/complex combinations + combined with klines

Scenario quick reference

User need	How to use
Today's limit-up/consecutive limit-up/strong auction stocks	`wencai_nlp("今日涨停，非ST")`
Stock selection by financial indicators (ROE/PE/PB)	`wencai_nlp("连续3年ROE大于15%，非ST")`
Stock selection by technical pattern (MACD golden cross)	`wencai_nlp("均线多头排列，MACD金叉")`
Stock selection by complex combination conditions	`wencai_nlp("...多条件...")` See case 5
CATL 5-minute K-line	`klines(code, interval="5m", count=78)`
Moutai today's intraday chart	`intraday_data(code)`
Intraday chart for a historical day	`min_snapshot(code, date="20250101")`
Order book bid/ask five levels	`depth(code)` or `tick_level1(code)`
Large order flow	`big_order_flow(code)`
Today's auction anomalies	`call_auction_anomaly(market)`
SW Industry List	`ths_industry()`
Concept Sector List	`ths_concept()`
Sector Constituent Stocks	`block_constituents(link_code)`
Index Quotes	`market_data_index(ths_code)`
Multi-Stock Comparison	Batch `market_data_cn` + `klines`
Dividend/Ex-Rights Data	`corporate_action(code)`
Today's IPOs	`ipo_today()`

Step Zero: Installation

pip install --upgrade thsdk

Package source:PyPI

Connection

All calls use guest mode uniformly, no account configuration required:

from thsdk import THS

with THS() as ths:
    ...

Step One: Stock Code Resolution

For all Chinese names/abbreviations/short codes first use search_symbols to obtain the full ths_code:

with THS() as ths:
    resp = ths.search_symbols("宁德时代")
    # resp.data → [{'THSCODE': 'USZA300750', 'Name': '宁德时代', 'Code': '300750', 'MarketDisplay': '深A'}, ...]

Code selection rules:

Case	Handling
0 results	Inform the user that nothing was found
1 result	Use directly
Multiple results, only 1 A-share	Automatically select the A-share
Multiple results, multiple A-shares	Display the list and wait for the user to choose

Indices use dedicated market prefixes (no need for search_symbols):

Index	THSCODE
SSE Composite Index	`USHI000001`
SZSE Component Index	`USZI399001`
ChiNext Index	`USZI399006`
STAR 50	`USHI000688`
CSI 300	`USHI000300`
CSI 500	`USHI000905`
SSE 50	`USHI000016`

⚠️ The index prefix is USHI/USZI(not USHA/USZA), and requires calling market_data_index rather than market_data_cn

Market Code Description

Prefix	Meaning
`USHA`	Shanghai A-shares
`USZA`	Shenzhen A-shares
`USHI`	Shanghai Index
`USZI`	Shenzhen Index
`USTM`	Beijing Stock Exchange
`UHKG`	Hong Kong Stocks

K-Line Data

Full interval Parameters

"1m" / "5m" / "15m" / "30m" / "60m" / "120m" / "day" / "week" / "month" / "quarter" / "year"

⚠️ The correct syntax is "5m" rather than "5min"

Usage (choose either count or start/end, do not mix them)

from datetime import datetime
from zoneinfo import ZoneInfo

tz = ZoneInfo('Asia/Shanghai')

with THS() as ths:
    # 方式1：按条数（最常用）
    resp = ths.klines("USZA300750", interval="5m", count=78)

    # 方式2：按时间范围
    resp = ths.klines(
        "USZA300750",
        interval="day",
        start_time=datetime(2025, 1, 1, tzinfo=tz),
        end_time=datetime(2025, 3, 1, tzinfo=tz)
    )

    # 复权：前复权 forward / 后复权 backward / 不复权 ""（默认）
    resp = ths.klines("USHA600519", interval="day", count=250, adjust="forward")

    df = resp.df  # 列: 时间, 开盘价, 最高价, 最低价, 收盘价, 成交量 等
    # 分钟K线的"时间"已自动转为 datetime；日K的"时间"为 datetime(YYYYMMDD)

Minute K-Line Analysis Example

with THS() as ths:
    resp = ths.klines("USZA300750", interval="5m", count=78)
    df = resp.df
    df['ma5'] = df['收盘价'].rolling(5).mean()
    df['ma20'] = df['收盘价'].rolling(20).mean()
    # 成交量异动（超均量2倍）
    df['vol_avg'] = df['成交量'].rolling(20).mean()
    df['vol_spike'] = df['成交量'] > df['vol_avg'] * 2
    # 支撑/压力位
    support = df['最低价'].tail(20).min()
    resistance = df['最高价'].tail(20).max()

Order Book and Real-Time Data

Intraday Time-Sharing (Current Day)

with THS() as ths:
    resp = ths.intraday_data("USZA300750")
    df = resp.df  # 列: 时间(datetime), 价格, 成交量, 均价 等

Historical Time-Sharing (any date within the past year)

with THS() as ths:
    resp = ths.min_snapshot("USZA300750", date="20250301")
    df = resp.df

Five-Level Bid/Ask Order Book

with THS() as ths:
    resp = ths.depth("USZA300750")              # 单只
    resp = ths.depth(["USZA300750", "USHA600519"])  # 多只
    df = resp.df  # 含 买1~5价/量, 卖1~5价/量

3-Second Tick Data

with THS() as ths:
    resp = ths.tick_level1("USZA300750")
    df = resp.df

Super Order Book (including ten-level orders)

with THS() as ths:
    resp = ths.tick_super_level1("USZA300750")                      # 实时
    resp = ths.tick_super_level1("USZA300750", date="20250301")     # 历史（近一年）
    df = resp.df

Large Orders and Call Auction

Large Order Flow

with THS() as ths:
    resp = ths.big_order_flow("USZA300750")
    df = resp.df
    # 含字段：主动买入特大单量/金额/笔数、主动卖出特大单量/金额/笔数、
    #         主动买入大单量/金额/笔数、资金流入/流出 等

Call Auction Anomalies (pre-market monitoring from 9:15 to 9:25)

with THS() as ths:
    resp = ths.call_auction_anomaly("USHA")   # 沪市
    resp = ths.call_auction_anomaly("USZA")   # 深市
    df = resp.df
    # 异动类型1 已自动映射中文：
    # 涨停试盘 / 跌停试盘 / 涨停撤单 / 竞价抢筹 / 竞价砸盘
    # 大幅高开 / 大幅低开 / 急速上涨 / 急速下跌
    # 买一剩余大 / 卖一剩余大 / 大买单试盘 / 大卖单试盘

Morning Call Auction Snapshot

with THS() as ths:
    resp = ths.call_auction("USZA300750")
    df = resp.df

Sectors and Indices

Industry Sector List

with THS() as ths:
    resp = ths.ths_industry()   # 同花顺行业（含 URFI 前缀的 link_code）
    df = resp.df  # 含板块名称、代码(link_code)、涨幅、成交量、上涨/下跌家数 等

Concept Sector List

with THS() as ths:
    resp = ths.ths_concept()
    df = resp.df  # 含概念名称、link_code、涨幅、领涨股 等

Sector Constituent Stocks

with THS() as ths:
    # 先获取行业/概念列表，找到 link_code（格式 URFIXXXXXX）
    industry_resp = ths.ths_industry()
    target_row = [r for r in industry_resp.data if '新能源' in str(r.get('名称', ''))][0]
    link_code = target_row.get('代码') or target_row.get('link_code')

    resp = ths.block_constituents(link_code)
    df = resp.df  # 含成分股代码、名称等

Sector Real-Time Quotes

with THS() as ths:
    # query_key: "基础数据"（涨幅/成交/市值）或 "扩展"（涨速/主力净流入）
    resp = ths.market_data_block("URFI881273", "基础数据")
    df = resp.df
    # 含: 价格, 涨幅, 成交量, 板块总市值, 板块流通市值, 上涨家数, 下跌家数, 领涨股

Index Real-Time Quotes

with THS() as ths:
    # 单只
    resp = ths.market_data_index("USHI000001", "基础数据")
    # 多只（必须同市场：同为 USHI 或同为 USZI）
    resp = ths.market_data_index(["USHI000001", "USHI000300", "USHI000905"])
    df = resp.df  # 含: 价格, 涨幅, 涨跌, 成交量, 总金额, 最高价, 最低价

    # 扩展（含量比、振幅等）
    resp = ths.market_data_index("USHI000001", "扩展")

Available query_key for market_data_cn

query_key	Meaning
`"基础数据"`	Price, change percentage, volume, turnover, open/high/low, speed of change, current volume
`"基础数据2"`	Compact version
`"基础数据3"`	Minimal (price, previous close, volume)
`"扩展1"`	Gain, change, turnover rate, volume ratio, main net inflow, bid/ask ratio
`"扩展2"`	Gain, turnover rate, total market cap, circulating market cap, bid/ask ratio, circulating market cap
`"汇总"`	Full fields (basic + extended merged, preferred for multi-stock comparison)

⚠️ market_data_cn requires the same market: Shanghai A-shares (USHA) and Shenzhen A-shares (USZA) cannot be mixed in the same call

Multi-Stock Batch Comparison

Complete Workflow

import pandas as pd
from collections import defaultdict
from thsdk import THS

stock_names = ["贵州茅台", "五粮液", "泸州老窖"]

with THS() as ths:
    # Step 1: 批量解析代码
    stock_codes = []
    for name in stock_names:
        resp = ths.search_symbols(name)
        a_shares = [s for s in resp.data
                    if any(m in s.get('MarketDisplay', '') for m in ['沪A', '深A'])]
        if a_shares:
            stock_codes.append({'name': name, 'code': a_shares[0]['THSCODE']})

    # Step 2: 按市场分组（market_data_cn 要求同市场）
    by_market = defaultdict(list)
    for s in stock_codes:
        by_market[s['code'][:4]].append(s)

    # Step 3: 批量获取行情
    rows = []
    for market, stocks in by_market.items():
        codes = [s['code'] for s in stocks]
        resp = ths.market_data_cn(codes, "汇总")
        for i, row in enumerate(resp.data):
            row['股票名称'] = stocks[i]['name']
            rows.append(row)
    quote_df = pd.DataFrame(rows)

    # Step 4: 批量K线
    klines_data = {}
    for s in stock_codes:
        resp = ths.klines(s['code'], interval="day", count=30, adjust="forward")
        klines_data[s['name']] = resp.df

# Step 5: 归一化
for name, df in klines_data.items():
    df['归一化'] = df['收盘价'] / df['收盘价'].iloc[0] * 100

# Step 6: 相关性（量化场景）
returns = pd.DataFrame({
    name: df['收盘价'].pct_change()
    for name, df in klines_data.items()
})
corr_matrix = returns.corr()

Output specification (two steps)

Step 1: Table(rendered by show_widget)

Step 2: Chart(rendered by show_widget)

Normalized trend line chart (multiple lines, distinguished by color, starting point = 100)
Additional output for quantitative scenarios: correlation heatmap

Wencai natural language query (wencai_nlp)

Wencai is an AI stock selection platform under Hithink RoyalFlush (iwencai.com), supporting full-market scanning with natural language. wencai_nlp Directly connects to the same interface; for multiple conditions usecomma/semicolon/spaceas separators.

with THS() as ths:
    resp = ths.wencai_nlp("连续3日主力净流入，换手率大于5%，非ST")
    df = resp.df  # 每行一只股票，列为查询涉及的字段

⚠️ buffer_size is already set to 8MB; no manual adjustment is needed when the returned data volume is large

Quick reference for six query types

① Quotes & market conditions

"今日涨停，非ST"
"连续2日涨停，非一字板，非ST"
"今日涨停原因类别，涨停封单额，封单量"
"竞价涨幅大于3%，竞价量大于昨日成交量5%，非ST"
"主力净流入由大到小排名前20，非ST"
"近10日区间主力资金流向大于5000万，市值大于100亿，日成交额大于30亿"

② Sectors & industries

"今日申万行业涨跌幅排名"
"今日概念板块涨幅排名前20"
"人工智能概念股，今日涨跌幅，成交额，主力净流入"
"半导体行业股票，涨幅，换手率，市值"
"今日涨幅最大的5个概念板块，涨幅，成分股数量"

③ Financial indicators

"连续3年ROE大于15%，非ST，上市大于3年"
"净利润增长率大于30%，营业收入增长率大于20%，非ST"
"市盈率小于15，股息率大于3%，市净率小于2，非ST"
"市净率小于1，非ST，流通市值大于20亿"          # 破净股
"连续5年分红，股息率大于4%，资产负债率小于60%"

④ Technical patterns

"均线多头排列，MACD金叉，换手率大于3%，非ST"
"5日均线上穿20日均线，成交量放大，涨幅大于1%"
"均线粘合，平台突破，成交量大于5日均量1.5倍"
"仙人指路，非ST，非停牌"
"250日新高，非ST，沪深A，上市超过250天"

⑤ Complex combinations (short-term/quantitative)

# 短线强势选股
"均线多头排列，MACD金叉，DIFF上穿中轴，换手率大于1%且小于10%，30日内有2个交易日涨幅大于4%，非ST"

# 竞价选股（隔日打板）
"昨日非一字板涨停，今日竞价涨幅大于等于0%且小于等于9.9%，今日隔夜买单额小于10亿，非ST，非科创板"

# 连板选股
"最近5日有过涨停，最近5日没有跌停，今日成交量大于5日平均成交量，今日竞价涨幅在2%到3%之间，非北交所非科创板非ST"

⑥ Information queries (non-stock selection)

"涨停原因归类前20"                       # 今日涨停题材分布
"今日龙虎榜"                             # 龙虎榜数据
"今日大宗交易"                           # 大宗交易
"今日融资融券余额最大的前20只股票"
"近一周北向资金净买入前20"

Processing wencai_nlp returned data

with THS() as ths:
    resp = ths.wencai_nlp("连续3日主力净流入，换手率大于5%，非ST，市值大于30亿")
    if not resp:
        print(f"查询失败: {resp.error}")
    else:
        df = resp.df
        # 字段名来自查询语句，常见列：股票代码、股票简称、涨幅、成交额、主力净流入 等

        # 补全 ths_code（供后续调用 klines/market_data_cn）
        def to_ths_code(code_str):
            code_str = str(code_str).zfill(6)
            if code_str.startswith('6'):   return f"USHA{code_str}"
            if code_str.startswith(('0','3')): return f"USZA{code_str}"
            if code_str.startswith('8'):   return f"USTM{code_str}"
            return None

        df['ths_code'] = df.get('股票代码', df.get('代码', pd.Series())).apply(to_ths_code)

wencai_nlp vs wencai_base

Method	Purpose
`wencai_nlp(condition)`	Main usage. Complete natural language, returns stock list + field data
`wencai_base(condition)`	Simple condition queries, e.g. `"所属行业"` to look up a single stock's affiliation

See the complete example in examples/05_wencai_nlp.py

Other useful APIs

Rights and dividends data (ex-rights/ex-dividend history)

with THS() as ths:
    resp = ths.corporate_action("USHA600519")
    df = resp.df

Today's IPO / pending subscription

with THS() as ths:
    resp = ths.ipo_today()   # 今日上市新股
    resp = ths.ipo_wait()    # 待申购打新

Wencai natural language query

with THS() as ths:
    resp = ths.wencai_nlp("今日申万行业涨跌幅排名")
    resp = ths.wencai_nlp("今日概念板块涨跌幅排名前20")
    resp = ths.wencai_nlp("换手率大于10%且涨幅大于5%的股票")
    df_list = resp.data

Error handling

with THS() as ths:
    resp = ths.klines("USZA300750", interval="5m", count=60)
    if not resp:                  # resp.success == False
        print(f"调用失败: {resp.error}")
    elif resp.df.empty:
        print("数据为空，可能是非交易时间")
    else:
        df = resp.df

Quick reference for common errors:

Error message	Cause	Solution
`"未登录"`	Not connected	Make sure to use `with THS() as ths`
`"证券代码必须为10个字符"`	Incorrect code format	First pass through `search_symbols`
`"一次性查询多支股票必须市场代码相同"`	Mixed Shanghai and Shenzhen	Query separately grouped by market
`"无效的周期类型: 5min"`	Incorrect interval format	Change to `"5m"`
`"'count' 参数不能与 'start_time' 同时使用"`	Parameter conflict	Choose one of the two

Division of labor with ths-financial-data

Scenario	skill
Single A-share quotes/fund flows/daily K	`ths-financial-data`
Minute K-line / intraday monitoring	This skill
Order book depth / large orders / auction anomalies	This skill
Sector/index quotes and constituent stocks	This skill
Multi-stock batch comparison	This skill
Wencai natural language query	Either works