1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14

import ccxt
import pandas as pd

# 取 binance 的 BTC/USDT 1分钟线,公共数据不用 API key
# 连不上?确保代理在跑,或把 'binance' 换成 'okx'/'gate'
ex = ccxt.binance()
ohlcv = ex.fetch_ohlcv('BTC/USDT', '1m', limit=500)  # 500 根

# ccxt 返回的是 list of list:[时间戳, 开, 高, 低, 收, 量]
df = pd.DataFrame(ohlcv, columns=['ts', 'open', 'high', 'low', 'close', 'vol'])
df['ts'] = pd.to_datetime(df['ts'], unit='ms')   # 毫秒时间戳 → 时间
df = df.set_index('ts')                            # 时间当索引
df.head()                                          # 看前 5 行
df.dtypes                                          # 看每列数据类型

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16

# 切片:按时间取一段(用 loc + 字符串时间)
recent = df.loc['2026-09-05':]

# 重采样:1m → 4h 的 OHLC(金融标准聚合方式)
df_4h = df.resample('4h').agg({
    'open': 'first',   # 区间开盘价 = 第一根
    'high': 'max',      # 区间最高 = 每根最高的最大值
    'low': 'min',
    'close': 'last',   # 区间收盘 = 最后一根
    'vol': 'sum'       # 区间成交量 = 求和
})
df_4h.head()

# 按位置切片(iloc,第几行)
df.iloc[:10]      # 前 10 行
df.iloc[-5:]      # 最后 5 行

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20

# 移动平均线 MA(20):过去 20 根的均值滚动
df['ma20'] = df['close'].rolling(20).mean()

# RSI(14):先算涨跌幅,再用 ewm 平滑
delta = df['close'].diff()                     # diff = 当前减前一根
gain = delta.clip(lower=0)                      # 只留正的部分(涨)
loss = -delta.clip(upper=0)                     # 只留负的(跌),取绝对值
avg_gain = gain.ewm(alpha=1/14, adjust=False).mean()
avg_loss = loss.ewm(alpha=1/14, adjust=False).mean()
rs = avg_gain / avg_loss
df['rsi'] = 100 - 100 / (1 + rs)

# MACD:两条 EMA 的差
ema12 = df['close'].ewm(span=12, adjust=False).mean()
ema26 = df['close'].ewm(span=26, adjust=False).mean()
df['macd'] = ema12 - ema26
df['signal'] = df['macd'].ewm(span=9, adjust=False).mean()
df['hist'] = df['macd'] - df['signal']

df[['close', 'ma20', 'rsi']].tail()             # 看最近几行

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13

# 先分别取 ETH 的收盘价 series(同样套路)
eth = ex.fetch_ohlcv('ETH/USDT', '1m', limit=500)
eth_df = pd.DataFrame(eth, columns=['ts','o','h','l','c','v'])
eth_df['ts'] = pd.to_datetime(eth_df['ts'], unit='ms')
eth_close = eth_df.set_index('ts')['c']

# 横向拼:把 BTC 收盘和 ETH 收盘并成两列(axis=1 = 按列拼)
both = pd.concat({'BTC': df['close'], 'ETH': eth_close}, axis=1)
both = both.dropna()            # 丢掉没对齐上的行

both.corr()                     # 相关性矩阵
both['spread'] = both['BTC'] - both['ETH']   # 新列 = 两列相减
both.plot()                     # 直接画图(pandas 自带)

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14

df['ret'] = df['close'].pct_change()          # 收益率 = (今-昨)/昨
df['ret'].describe()                          # 均值/标准差/分位数 速览

# 按小时分组:看哪个小时波动最大
df['hour'] = df.index.hour
df.groupby('hour')['ret'].agg(['mean', 'std', 'count'])

# 按天聚合日收益
df['day'] = df.index.date
daily_ret = df.groupby('day')['ret'].sum()

# 画图(在 notebook 里内联显示)
df[['close','ma20']].plot(title='BTC close + MA20')
df['ret'].hist(bins=50)                        # 收益率分布直方图

 1
 2
 3
 4
 5
 6
 7
 8
 9
10

import vectorbt as vbt

# 信号:收盘价上穿 MA20 买入,下穿卖出(布尔 Series)
entries = df['close'] > df['ma20']
exits   = df['close'] <= df['ma20']

# vectorbt 接收 close(Series)+ 进出场布尔信号
pf = vbt.Portfolio.from_signals(df['close'], entries, exits, init_cash=10000)
pf.stats()                  # 回测统计:总收益/胜率/最大回撤等
pf.plot()                   # 回测曲线