vnpy/docs/cta_strategy.md

1333 lines
44 KiB
Markdown
Raw Normal View History

2019-04-02 08:50:43 +00:00
# CTA策略模块
## 模块构成
2019-04-02 08:50:43 +00:00
2019-04-12 08:01:58 +00:00
CTA策略模块主要由7部分构成如下图
2019-04-02 08:50:43 +00:00
2019-04-12 08:01:58 +00:00
- base定义了CTA模块中用到的一些基础设置如引擎类型回测/实盘、回测模式K线/Tick、本地停止单的定义以及停止单状态等待中/已撤销/已触发)。
- template定义了CTA策略模板包含信号生成和委托管理、CTA信号仅负责信号生成、目标仓位算法仅负责委托管理适用于拆分巨型委托降低冲击成本
- strategies: 官方提供的cta策略示例包含从最基础的双均线策略到通道突破类型的布林带策略到跨时间周期策略再到把信号生成和委托管理独立开来的多信号策略。
- backesting包含回测引擎和参数优化。其中回测引擎定义了数据载入、委托撮合机制、计算与统计相关盈利指标、结果绘图等函数。
- converter定义了针对上期所品种平今/平昨模式的委托转换模块对于其他品种用户也可以通过可选参数lock切换至锁仓模式。
- engine定义了CTA策略实盘引擎其中包括RQData客户端初始化和数据载入、策略的初始化和启动、推送Tick订阅行情到策略中、挂撤单操作、策略的停止和移除等。
- ui基于PyQt5的GUI图形应用。
2019-04-02 08:50:43 +00:00
2019-04-29 06:33:15 +00:00
![](https://vnpy-community.oss-cn-shanghai.aliyuncs.com/forum_experience/yazhang/cta_strategy/seix_elementos.png "enter image title here")
2019-04-02 08:50:43 +00:00
2019-04-12 08:01:58 +00:00
 
2019-04-02 08:50:43 +00:00
2019-04-29 06:33:15 +00:00
## 历史数据
2019-04-02 08:50:43 +00:00
2019-04-29 06:33:15 +00:00
### 回测历史数据
回测所需要的历史数据可通过运行getdata.py文件进行下载。该文件处于根目录下tests\backtesting文件夹内。
下载历史数据的原理是调用RQData的get_price()函数把数据下载到内存里面再通过generate_bar_from_row()函数,以固定格式把数据从内存载入到硬盘数据库中。
下面介绍具体流程:
- 填写RQData的账号密码初始化RQData
```
import rqdatac as rq
USERNAME = ""
PASSWORD = ""
FIELDS = ["open", "high", "low", "close", "volume"]
rq.init(USERNAME, PASSWORD, ("rqdatad-pro.ricequant.com", 16011))
```
 
- 定义数据插入格式。需要插入的数据包括合约代码、交易所、K线周期、开盘价、最高价、最低价、收盘价、成交量、数据库名称、vt_symbol注意K线周期可以是"1m"、"1h"、"d"、"w"。to_pydatetime()用于时间转换成datetime格式
```
def generate_bar_from_row(row, symbol, exchange):
""""""
bar = DbBarData()
bar.symbol = symbol
bar.exchange = exchange
bar.interval = "1m"
bar.open_price = row["open"]
bar.high_price = row["high"]
bar.low_price = row["low"]
bar.close_price = row["close"]
bar.volume = row["volume"]
bar.datetime = row.name.to_pydatetime()
bar.gateway_name = "DB"
bar.vt_symbol = f"{symbol}.{exchange}"
return bar
```
 
- 定义数据下载函数。主要调用RQData的get_price()获取指定合约或合约列表的历史数据包含起止日期日线或分钟线。目前仅支持中国市场的股票、期货、ETF和上金所现货的行情数据如黄金、铂金和白银产品。注意起始日期默认是2013-01-04结束日期默认是2014-01-04
```
def download_minute_bar(vt_symbol):
"""下载某一合约的分钟线数据"""
print(f"开始下载合约数据{vt_symbol}")
symbol, exchange = vt_symbol.split(".")
start = time()
df = rq.get_price(symbol, start_date="2018-01-01", end_date="2019-01-01", frequency="1m", fields=FIELDS)
with DB.atomic():
for ix, row in df.iterrows():
print(row.name)
bar = generate_bar_from_row(row, symbol, exchange)
DbBarData.replace(bar.__data__).execute()
end = time()
cost = (end - start) * 1000
print(
"合约%s的分钟K线数据下载完成%s - %s耗时%s毫秒"
% (symbol, df.index[0], df.index[-1], cost)
)
```
 
### 实盘历史数据
在实盘中RQData通过实时载入数据进行策略的初始化。该功能主要在CTA实盘引擎engine.py内实现。
下面介绍具体流程:
- 配置json文件在用户目录下.vntrader文件夹找到vt_setting.json输入RQData的账号和密码如图。
![](https://vnpy-community.oss-cn-shanghai.aliyuncs.com/forum_experience/yazhang/cta_strategy/RQData_setting.png "enter image title here")
- 初始化RQData客户端从vt_setting.json中读取RQData的账户、密码到rq_client.init()函数进行初始化
```
def init_rqdata(self):
"""
Init RQData client.
"""
username = SETTINGS["rqdata.username"]
password = SETTINGS["rqdata.password"]
if not username or not password:
return
import rqdatac
self.rq_client = rqdatac
self.rq_client.init(username, password,
('rqdatad-pro.ricequant.com', 16011))
```
- RQData载入实盘数据输入vt_symbol后首先会转换成符合RQData格式的rq_symbol通过get_price()函数下载数据,并且插入到数据库中。
```
def query_bar_from_rq(
self, vt_symbol: str, interval: Interval, start: datetime, end: datetime
):
"""
Query bar data from RQData.
"""
symbol, exchange_str = vt_symbol.split(".")
rq_symbol = to_rq_symbol(vt_symbol)
if rq_symbol not in self.rq_symbols:
return None
end += timedelta(1) # For querying night trading period data
df = self.rq_client.get_price(
rq_symbol,
frequency=interval.value,
fields=["open", "high", "low", "close", "volume"],
start_date=start,
end_date=end
)
data = []
for ix, row in df.iterrows():
bar = BarData(
symbol=symbol,
exchange=Exchange(exchange_str),
interval=interval,
datetime=row.name.to_pydatetime(),
open_price=row["open"],
high_price=row["high"],
low_price=row["low"],
close_price=row["close"],
volume=row["volume"],
gateway_name="RQ"
)
data.append(bar)
```
2019-04-02 08:50:43 +00:00
2019-04-12 08:01:58 +00:00
 
2019-04-02 08:50:43 +00:00
2019-04-29 06:33:15 +00:00
## 策略开发
2019-04-12 08:01:58 +00:00
CTA策略模板提供完整的信号生成和委托管理功能用户可以基于该模板自行开发策略。新策略可以放在根目录下vnpy\app\cta_strategy\strategies文件夹内也可以放在用户运行的文件内VN Station模式。注意策略文件命名是以下划线模式如boll_channel_strategy.py而策略类命名采用的是驼峰式如BollChannelStrategy。
2019-04-02 08:50:43 +00:00
2019-04-12 08:01:58 +00:00
下面通过BollChannelStrategy策略示例来展示策略开发的具体步骤
2019-04-02 08:50:43 +00:00
2019-04-29 06:33:15 +00:00
### 参数设置
2019-04-02 08:50:43 +00:00
2019-04-12 08:01:58 +00:00
定义策略参数并且初始化策略变量。策略参数为策略类的公有属性,用户可以通过创建新的实例来调用或者改变策略参数。
2019-04-02 08:50:43 +00:00
2019-04-12 08:01:58 +00:00
如针对rb1905品种用户可以创建基于BollChannelStrategy的策略示例如RB_BollChannelStrategyboll_window可以由18改成30。
2019-04-02 08:50:43 +00:00
2019-04-12 08:01:58 +00:00
创建策略实例的方法有效地实现了一个策略跑多个品种,并且其策略参数可以通过品种的特征进行调整。
```
boll_window = 18
boll_dev = 3.4
cci_window = 10
atr_window = 30
sl_multiplier = 5.2
fixed_size = 1
boll_up = 0
boll_down = 0
cci_value = 0
atr_value = 0
intra_trade_high = 0
intra_trade_low = 0
long_stop = 0
short_stop = 0
```
2019-04-29 06:33:15 +00:00
### 类的初始化
2019-04-12 08:01:58 +00:00
初始化分3步
- 通过super( )的方法继承CTA策略模板在__init__( )函数传入CTA引擎、策略名称、vt_symbol、参数设置。
- 调用K线生成模块:通过时间切片来把Tick数据合成1分钟K线数据然后更大的时间周期数据如15分钟K线。
- 调用K线时间序列管理模块基于K线数据如1分钟、15分钟来生成相应的技术指标。
```
def __init__(self, cta_engine, strategy_name, vt_symbol, setting):
""""""
super(BollChannelStrategy, self).__init__(
cta_engine, strategy_name, vt_symbol, setting
)
self.bg = BarGenerator(self.on_bar, 15, self.on_15min_bar)
self.am = ArrayManager()
```
2019-04-29 06:33:15 +00:00
### 策略的初始化、启动、停止
2019-04-12 08:01:58 +00:00
通过“CTA策略”组件的相关功能按钮实现。
注意函数load_bar(10)代表策略初始化需要载入10个交易日的历史数据。该历史数据可以是Tick数据也可以是K线数据。
```
def on_init(self):
"""
Callback when strategy is inited.
"""
self.write_log("策略初始化")
self.load_bar(10)
def on_start(self):
"""
Callback when strategy is started.
"""
self.write_log("策略启动")
def on_stop(self):
"""
Callback when strategy is stopped.
"""
self.write_log("策略停止")
```
2019-04-29 06:33:15 +00:00
### Tick数据回报
2019-04-12 08:01:58 +00:00
策略订阅某品种合约行情交易所会推送Tick数据到该策略上。
由于BollChannelStrategy是基于15分钟K线来生成交易信号的故收到Tick数据后需要用到K线生成模块里面的update_tick函数通过时间切片的方法聚合成1分钟K线数据并且推送到on_bar函数。
```
def on_tick(self, tick: TickData):
"""
Callback of new tick data update.
"""
self.bg.update_tick(tick)
```
2019-04-29 06:33:15 +00:00
### K线数据回报
2019-04-12 08:01:58 +00:00
收到推送过来的1分钟K线数据后通过K线生成模块里面的update_bar函数以分钟切片的方法合成15分钟K线数据并且推送到on_15min_bar函数。
```
def on_bar(self, bar: BarData):
"""
Callback of new bar data update.
"""
self.bg.update_bar(bar)
```
2019-04-29 06:33:15 +00:00
### 15分钟K线数据回报
2019-04-12 08:01:58 +00:00
负责CTA信号的生成由3部分组成
- 清空未成交委托为了防止之前下的单子在上一个15分钟没有成交但是下一个15分钟可能已经调整了价格就用cancel_all()方法立刻撤销之前未成交的所有委托保证策略在当前这15分钟开始时的整个状态是清晰和唯一的。
- 调用K线时间序列管理模块基于最新的15分钟K线数据来计算相应计算指标如布林带通道上下轨、CCI指标、ATR指标
- 信号计算通过持仓的判断以及结合CCI指标、布林带通道、ATR指标在通道突破点挂出停止单委托buy/sell),同时设置离场点(short/cover)。
注意CTA策略具有低胜率和高盈亏比的特定在难以提升胜率的情况下研究提高策略盈亏比有利于策略盈利水平的上升。
```
def on_15min_bar(self, bar: BarData):
""""""
self.cancel_all()
am = self.am
am.update_bar(bar)
if not am.inited:
return
self.boll_up, self.boll_down = am.boll(self.boll_window, self.boll_dev)
self.cci_value = am.cci(self.cci_window)
self.atr_value = am.atr(self.atr_window)
if self.pos == 0:
self.intra_trade_high = bar.high_price
self.intra_trade_low = bar.low_price
if self.cci_value > 0:
self.buy(self.boll_up, self.fixed_size, True)
elif self.cci_value < 0:
self.short(self.boll_down, self.fixed_size, True)
elif self.pos > 0:
self.intra_trade_high = max(self.intra_trade_high, bar.high_price)
self.intra_trade_low = bar.low_price
self.long_stop = self.intra_trade_high - self.atr_value * self.sl_multiplier
self.sell(self.long_stop, abs(self.pos), True)
elif self.pos < 0:
self.intra_trade_high = bar.high_price
self.intra_trade_low = min(self.intra_trade_low, bar.low_price)
self.short_stop = self.intra_trade_low + self.atr_value * self.sl_multiplier
self.cover(self.short_stop, abs(self.pos), True)
self.put_event()
```
2019-04-29 06:33:15 +00:00
### 委托回报、成交回报、停止单回报
2019-04-12 08:01:58 +00:00
在策略中可以直接pass其具体逻辑应用交给回测/实盘引擎负责。
```
def on_order(self, order: OrderData):
"""
Callback of new order data update.
"""
pass
def on_trade(self, trade: TradeData):
"""
Callback of new trade data update.
"""
self.put_event()
def on_stop_order(self, stop_order: StopOrder):
"""
Callback of stop order update.
"""
pass
```
&nbsp;
2019-04-29 06:33:15 +00:00
## 回测研究
2019-04-12 08:01:58 +00:00
backtesting.py定义了回测引擎下面主要介绍相关功能函数以及回测引擎应用示例
2019-04-29 06:33:15 +00:00
### 加载策略
2019-04-12 08:01:58 +00:00
把CTA策略逻辑对应合约品种以及参数设置可在策略文件外修改载入到回测引擎中。
```
def add_strategy(self, strategy_class: type, setting: dict):
""""""
self.strategy_class = strategy_class
self.strategy = strategy_class(
self, strategy_class.__name__, self.vt_symbol, setting
)
```
&nbsp;
2019-04-29 06:33:15 +00:00
### 载入历史数据
2019-04-12 08:01:58 +00:00
负责载入对应品种的历史数据大概有4个步骤
- 根据数据类型不同分成K线模式和Tick模式
- 通过select().where()方法有条件地从数据库中选取数据其筛选标准包括vt_symbol、 回测开始日期、回测结束日期、K线周期K线模式下
- order_by(DbBarData.datetime)表示需要按照时间顺序载入数据;
- 载入数据是以迭代方式进行的数据最终存入self.history_data。
```
def load_data(self):
""""""
self.output("开始加载历史数据")
if self.mode == BacktestingMode.BAR:
s = (
DbBarData.select()
.where(
(DbBarData.vt_symbol == self.vt_symbol)
& (DbBarData.interval == self.interval)
& (DbBarData.datetime >= self.start)
& (DbBarData.datetime <= self.end)
)
.order_by(DbBarData.datetime)
)
self.history_data = [db_bar.to_bar() for db_bar in s]
else:
s = (
DbTickData.select()
.where(
(DbTickData.vt_symbol == self.vt_symbol)
& (DbTickData.datetime >= self.start)
& (DbTickData.datetime <= self.end)
)
.order_by(DbTickData.datetime)
)
self.history_data = [db_tick.to_tick() for db_tick in s]
self.output(f"历史数据加载完成,数据量:{len(self.history_data)}")
```
&nbsp;
2019-04-29 06:33:15 +00:00
### 撮合成交
2019-04-12 08:01:58 +00:00
载入CTA策略以及相关历史数据后策略会根据最新的数据来计算相关指标。若符合条件会生成交易信号发出具体委托buy/sell/short/cover并且在下一根K线成交。
根据委托类型的不同回测引擎提供2种撮合成交机制来尽量模仿真实交易环节
- 限价单撮合成交:(以买入方向为例)先确定是否发生成交,成交标准为委托价>= 下一根K线的最低价然后确定成交价格成交价格为委托价与下一根K线开盘价的最小值。
- 停止单撮合成交:(以买入方向为例)先确定是否发生成交,成交标准为委托价<= 下一根K线的最高价然后确定成交价格成交价格为委托价与下一根K线开盘价的最大值。
&nbsp;
下面展示在引擎中限价单撮合成交的流程:
- 确定会撮合成交的价格;
- 遍历限价单字典中的所有限价单,推送委托进入未成交队列的更新状态;
- 判断成交状态,若出现成交,推送成交数据和委托数据;
- 从字典中删除已成交的限价单。
```
def cross_limit_order(self):
"""
Cross limit order with last bar/tick data.
"""
if self.mode == BacktestingMode.BAR:
long_cross_price = self.bar.low_price
short_cross_price = self.bar.high_price
long_best_price = self.bar.open_price
short_best_price = self.bar.open_price
else:
long_cross_price = self.tick.ask_price_1
short_cross_price = self.tick.bid_price_1
long_best_price = long_cross_price
short_best_price = short_cross_price
for order in list(self.active_limit_orders.values()):
# Push order update with status "not traded" (pending)
if order.status == Status.SUBMITTING:
order.status = Status.NOTTRADED
self.strategy.on_order(order)
# Check whether limit orders can be filled.
long_cross = (
order.direction == Direction.LONG
and order.price >= long_cross_price
and long_cross_price > 0
)
short_cross = (
order.direction == Direction.SHORT
and order.price <= short_cross_price
and short_cross_price > 0
)
if not long_cross and not short_cross:
continue
# Push order udpate with status "all traded" (filled).
order.traded = order.volume
order.status = Status.ALLTRADED
self.strategy.on_order(order)
self.active_limit_orders.pop(order.vt_orderid)
# Push trade update
self.trade_count += 1
if long_cross:
trade_price = min(order.price, long_best_price)
pos_change = order.volume
else:
trade_price = max(order.price, short_best_price)
pos_change = -order.volume
trade = TradeData(
symbol=order.symbol,
exchange=order.exchange,
orderid=order.orderid,
tradeid=str(self.trade_count),
direction=order.direction,
offset=order.offset,
price=trade_price,
volume=order.volume,
time=self.datetime.strftime("%H:%M:%S"),
gateway_name=self.gateway_name,
)
trade.datetime = self.datetime
self.strategy.pos += pos_change
self.strategy.on_trade(trade)
self.trades[trade.vt_tradeid] = trade
```
&nbsp;
2019-04-29 06:33:15 +00:00
### 计算策略盈亏情况
2019-04-12 08:01:58 +00:00
基于收盘价、当日持仓量、合约规模、滑点、手续费率等计算总盈亏与净盈亏并且其计算结果以DataFrame格式输出完成基于逐日盯市盈亏统计。
下面展示盈亏情况的计算过程
- 浮动盈亏 = 持仓量 \*(当日收盘价 - 昨日收盘价)\* 合约规模
- 实际盈亏 = 持仓变化量 \* (当时收盘价 - 开仓成交价)\* 合约规模
- 总盈亏 = 浮动盈亏 + 实际盈亏
- 净盈亏 = 总盈亏 - 总手续费 - 总滑点
```
def calculate_pnl(
self,
pre_close: float,
start_pos: float,
size: int,
rate: float,
slippage: float,
):
""""""
self.pre_close = pre_close
# Holding pnl is the pnl from holding position at day start
self.start_pos = start_pos
self.end_pos = start_pos
self.holding_pnl = self.start_pos * \
(self.close_price - self.pre_close) * size
# Trading pnl is the pnl from new trade during the day
self.trade_count = len(self.trades)
for trade in self.trades:
if trade.direction == Direction.LONG:
pos_change = trade.volume
else:
pos_change = -trade.volume
turnover = trade.price * trade.volume * size
self.trading_pnl += pos_change * \
(self.close_price - trade.price) * size
self.end_pos += pos_change
self.turnover += turnover
self.commission += turnover * rate
self.slippage += trade.volume * size * slippage
# Net pnl takes account of commission and slippage cost
self.total_pnl = self.trading_pnl + self.holding_pnl
self.net_pnl = self.total_pnl - self.commission - self.slippage
```
&nbsp;
2019-04-29 06:33:15 +00:00
### 计算策略统计指标
2019-04-12 08:01:58 +00:00
calculate_statistics函数是基于逐日盯市盈亏情况DateFrame格式来计算衍生指标如最大回撤、年化收益、盈亏比、夏普比率等。
```
df["balance"] = df["net_pnl"].cumsum() + self.capital
df["return"] = np.log(df["balance"] / df["balance"].shift(1)).fillna(0)
df["highlevel"] = (
df["balance"].rolling(
min_periods=1, window=len(df), center=False).max()
)
df["drawdown"] = df["balance"] - df["highlevel"]
df["ddpercent"] = df["drawdown"] / df["highlevel"] * 100
# Calculate statistics value
start_date = df.index[0]
end_date = df.index[-1]
total_days = len(df)
profit_days = len(df[df["net_pnl"] > 0])
loss_days = len(df[df["net_pnl"] < 0])
end_balance = df["balance"].iloc[-1]
max_drawdown = df["drawdown"].min()
max_ddpercent = df["ddpercent"].min()
total_net_pnl = df["net_pnl"].sum()
daily_net_pnl = total_net_pnl / total_days
total_commission = df["commission"].sum()
daily_commission = total_commission / total_days
total_slippage = df["slippage"].sum()
daily_slippage = total_slippage / total_days
total_turnover = df["turnover"].sum()
daily_turnover = total_turnover / total_days
total_trade_count = df["trade_count"].sum()
daily_trade_count = total_trade_count / total_days
total_return = (end_balance / self.capital - 1) * 100
annual_return = total_return / total_days * 240
daily_return = df["return"].mean() * 100
return_std = df["return"].std() * 100
if return_std:
sharpe_ratio = daily_return / return_std * np.sqrt(240)
else:
sharpe_ratio = 0
```
&nbsp;
2019-04-29 06:33:15 +00:00
### 统计指标绘图
2019-04-12 08:01:58 +00:00
通过matplotlib绘制4幅图
- 资金曲线图
- 资金回撤图
- 每日盈亏图
- 每日盈亏分布图
```
def show_chart(self, df: DataFrame = None):
""""""
if not df:
df = self.daily_df
if df is None:
return
plt.figure(figsize=(10, 16))
balance_plot = plt.subplot(4, 1, 1)
balance_plot.set_title("Balance")
df["balance"].plot(legend=True)
drawdown_plot = plt.subplot(4, 1, 2)
drawdown_plot.set_title("Drawdown")
drawdown_plot.fill_between(range(len(df)), df["drawdown"].values)
pnl_plot = plt.subplot(4, 1, 3)
pnl_plot.set_title("Daily Pnl")
df["net_pnl"].plot(kind="bar", legend=False, grid=False, xticks=[])
distribution_plot = plt.subplot(4, 1, 4)
distribution_plot.set_title("Daily Pnl Distribution")
df["net_pnl"].hist(bins=50)
plt.show()
```
&nbsp;
2019-06-11 01:58:32 +00:00
### 单策略回测示例
2019-04-12 08:01:58 +00:00
- 导入回测引擎和CTA策略
- 设置回测相关参数品种、K线周期、回测开始和结束日期、手续费、滑点、合约规模、起始资金
- 载入策略和数据到引擎中,运行回测。
- 计算基于逐日统计盈利情况,计算统计指标,统计指标绘图。
```
from vnpy.app.cta_strategy.backtesting import BacktestingEngine
from vnpy.app.cta_strategy.strategies.boll_channel_strategy import (
BollChannelStrategy,
)
from datetime import datetime
engine = BacktestingEngine()
engine.set_parameters(
vt_symbol="IF88.CFFEX",
interval="1m",
start=datetime(2018, 1, 1),
end=datetime(2019, 1, 1),
rate=3.0/10000,
slippage=0.2,
size=300,
pricetick=0.2,
capital=1_000_000,
)
engine.add_strategy(AtrRsiStrategy, {})
engine.load_data()
engine.run_backtesting()
df = engine.calculate_result()
engine.calculate_statistics()
engine.show_chart()
```
&nbsp;
2019-06-11 01:58:32 +00:00
### 投资组合回测示例
投资组合回测是基于单策略回测的其关键是每个策略都对应着各自的BacktestingEngine对象下面介绍具体流程
- 创建回测函数run_backtesting()这样每添加一个策略就创建其BacktestingEngine对象。
```
from vnpy.app.cta_strategy.backtesting import BacktestingEngine, OptimizationSetting
from vnpy.app.cta_strategy.strategies.atr_rsi_strategy import AtrRsiStrategy
from vnpy.app.cta_strategy.strategies.boll_channel_strategy import BollChannelStrategy
from datetime import datetime
def run_backtesting(strategy_class, setting, vt_symbol, interval, start, end, rate, slippage, size, pricetick, capital):
engine = BacktestingEngine()
engine.set_parameters(
vt_symbol=vt_symbol,
interval=interval,
start=start,
end=end,
rate=rate,
slippage=slippage,
size=size,
pricetick=pricetick,
capital=capital
)
engine.add_strategy(strategy_class, setting)
engine.load_data()
engine.run_backtesting()
df = engine.calculate_result()
return df
```
&nbsp;
- 分别进行单策略回测得到各自的DataFrame(该DataFrame包含交易时间、今仓、昨仓、手续费、滑点、当日净盈亏、累计净盈亏等基本信息但是不包括最大回撤夏普比率等统计信息),然后把DataFrame相加并且去除空值后即得到投资组合的DataFrame。
```
df1 = run_backtesting(
strategy_class=AtrRsiStrategy,
setting={},
vt_symbol="IF88.CFFEX",
interval="1m",
start=datetime(2019, 1, 1),
end=datetime(2019, 4, 30),
rate=0.3/10000,
slippage=0.2,
size=300,
pricetick=0.2,
capital=1_000_000,
)
df2 = run_backtesting(
strategy_class=BollChannelStrategy,
setting={'fixed_size': 16},
vt_symbol="RB88.SHFE",
interval="1m",
start=datetime(2019, 1, 1),
end=datetime(2019, 4, 30),
rate=1/10000,
slippage=1,
size=10,
pricetick=1,
capital=1_000_000,
)
dfp = df1 + df2
dfp =dfp.dropna()
```
&nbsp;
- 创建show_portafolio()函数同样也是创建新的BacktestingEngine对象对传入的DataFrame计算如夏普比率等统计指标并且画图。故该函数不仅能显示单策略回测效果也能展示投资组合回测效果。
```
def show_portafolio(df):
engine = BacktestingEngine()
engine.calculate_statistics(df)
engine.show_chart(df)
show_portafolio(dfp)
```
&nbsp;
2019-04-29 06:33:15 +00:00
## 参数优化
2019-04-12 08:01:58 +00:00
参数优化模块主要由3部分构成
2019-04-29 06:33:15 +00:00
### 参数设置
2019-04-12 08:01:58 +00:00
- 设置参数优化区间如boll_window设置起始值为18终止值为24步进为2这样就得到了[18, 20, 22, 24] 这4个待优化的参数了。
- 设置优化目标字段:如夏普比率、盈亏比、总收益率等。
- 随机生成参数对组合:使用迭代工具产生参数对组合,然后把参数对组合打包到一个个字典组成的列表中
```
class OptimizationSetting:
"""
Setting for runnning optimization.
"""
def __init__(self):
""""""
self.params = {}
self.target_name = ""
def add_parameter(
self, name: str, start: float, end: float = None, step: float = None
):
""""""
if not end and not step:
self.params[name] = [start]
return
if start >= end:
print("参数优化起始点必须小于终止点")
return
if step <= 0:
print("参数优化步进必须大于0")
return
value = start
value_list = []
while value <= end:
value_list.append(value)
value += step
self.params[name] = value_list
def set_target(self, target_name: str):
""""""
self.target_name = target_name
def generate_setting(self):
""""""
keys = self.params.keys()
values = self.params.values()
products = list(product(*values))
settings = []
for p in products:
setting = dict(zip(keys, p))
settings.append(setting)
return settings
```
&nbsp;
2019-04-29 06:33:15 +00:00
### 参数对组合回测
2019-04-12 08:01:58 +00:00
多进程优化时每个进程都会运行optimize函数输出参数对组合以及目标优化字段的结果。其步骤如下
- 调用回测引擎
- 输入回测相关设置
- 输入参数对组合到策略中
- 运行回测
- 返回回测结果,包括:参数对组合、目标优化字段数值、策略统计指标
```
def optimize(
target_name: str,
strategy_class: CtaTemplate,
setting: dict,
vt_symbol: str,
interval: Interval,
start: datetime,
rate: float,
slippage: float,
size: float,
pricetick: float,
capital: int,
end: datetime,
mode: BacktestingMode,
):
"""
Function for running in multiprocessing.pool
"""
engine = BacktestingEngine()
engine.set_parameters(
vt_symbol=vt_symbol,
interval=interval,
start=start,
rate=rate,
slippage=slippage,
size=size,
pricetick=pricetick,
capital=capital,
end=end,
mode=mode
)
engine.add_strategy(strategy_class, setting)
engine.load_data()
engine.run_backtesting()
engine.calculate_result()
statistics = engine.calculate_statistics()
target_value = statistics[target_name]
return (str(setting), target_value, statistics)
```
&nbsp;
2019-04-29 06:33:15 +00:00
### 多进程优化
2019-04-12 08:01:58 +00:00
- 根据CPU的核数来创建进程若CPU为4核则创建4个进程
- 在每个进程都调用apply_async( )的方法运行参数对组合回测其回测结果添加到results中 apply_async是异步非阻塞的即不用等待当前进程执行完毕随时根据系统调度来进行进程切换。
- pool.close()与pool.join()用于进程跑完任务后,去关闭进程。
- 对results的内容通过目标优化字段标准进行排序输出结果。
```
pool = multiprocessing.Pool(multiprocessing.cpu_count())
results = []
for setting in settings:
result = (pool.apply_async(optimize, (
target_name,
self.strategy_class,
setting,
self.vt_symbol,
self.interval,
self.start,
self.rate,
self.slippage,
self.size,
self.pricetick,
self.capital,
self.end,
self.mode
)))
results.append(result)
pool.close()
pool.join()
# Sort results and output
result_values = [result.get() for result in results]
result_values.sort(reverse=True, key=lambda result: result[1])
for value in result_values:
msg = f"参数:{value[0]}, 目标:{value[1]}"
self.output(msg)
return result_values
```
&nbsp;
2019-04-29 06:33:15 +00:00
## 实盘运行
在实盘环境用户可以基于编写好的CTA策略来创建新的实例一键初始化、启动、停止策略。
### 创建策略实例
用户可以基于编写好的CTA策略来创建新的实例策略实例的好处在于同一个策略可以同时去运行多个品种合约并且每个实例的参数可以是不同的。
在创建实例的时候需要填写如图的实例名称、合约品种、参数设置。注意实例名称不能重名合约名称是vt_symbol的格式如IF1905.CFFEX。
![](https://vnpy-community.oss-cn-shanghai.aliyuncs.com/forum_experience/yazhang/cta_strategy/add_strategy.png)
创建策略流程如下:
- 检查策略实例重名
- 添加策略配置信息(strategy_name, vt_symbol, setting)到strategies字典上
- 添加该策略要订阅行情的合约信息到symbol_strategy_map字典中
- 把策略配置信息保存到json文件内
- 在图形化界面更新状态信息。
```
def add_strategy(
self, class_name: str, strategy_name: str, vt_symbol: str, setting: dict
):
"""
Add a new strategy.
"""
if strategy_name in self.strategies:
self.write_log(f"创建策略失败,存在重名{strategy_name}")
return
strategy_class = self.classes[class_name]
strategy = strategy_class(self, strategy_name, vt_symbol, setting)
self.strategies[strategy_name] = strategy
# Add vt_symbol to strategy map.
strategies = self.symbol_strategy_map[vt_symbol]
strategies.append(strategy)
# Update to setting file.
self.update_strategy_setting(strategy_name, setting)
2019-04-02 08:50:43 +00:00
2019-04-29 06:33:15 +00:00
self.put_strategy_event(strategy)
```
&nbsp;
### 初始化策略
- 调用策略类的on_init()回调函数,并且载入历史数据;
- 恢复上次退出之前的策略状态;
- 调用接口的subcribe()函数订阅指定行情信息;
- 策略初始化状态变成True并且更新到日志上。
```
def _init_strategy(self):
"""
Init strategies in queue.
"""
while not self.init_queue.empty():
strategy_name = self.init_queue.get()
strategy = self.strategies[strategy_name]
if strategy.inited:
self.write_log(f"{strategy_name}已经完成初始化,禁止重复操作")
continue
self.write_log(f"{strategy_name}开始执行初始化")
# Call on_init function of strategy
self.call_strategy_func(strategy, strategy.on_init)
# Restore strategy data(variables)
data = self.strategy_data.get(strategy_name, None)
if data:
for name in strategy.variables:
value = data.get(name, None)
if value:
setattr(strategy, name, value)
# Subscribe market data
contract = self.main_engine.get_contract(strategy.vt_symbol)
if contract:
req = SubscribeRequest(
symbol=contract.symbol, exchange=contract.exchange)
self.main_engine.subscribe(req, contract.gateway_name)
else:
self.write_log(f"行情订阅失败,找不到合约{strategy.vt_symbol}", strategy)
# Put event to update init completed status.
strategy.inited = True
self.put_strategy_event(strategy)
self.write_log(f"{strategy_name}初始化完成")
self.init_thread = None
```
&nbsp;
### 启动策略
- 检查策略初始化状态;
- 检查策略启动状态,避免重复启动;
- 调用策略类的on_start()函数启动策略;
- 策略启动状态变成True并且更新到图形化界面上。
```
def start_strategy(self, strategy_name: str):
"""
Start a strategy.
"""
strategy = self.strategies[strategy_name]
if not strategy.inited:
self.write_log(f"策略{strategy.strategy_name}启动失败,请先初始化")
return
if strategy.trading:
self.write_log(f"{strategy_name}已经启动,请勿重复操作")
return
self.call_strategy_func(strategy, strategy.on_start)
strategy.trading = True
self.put_strategy_event(strategy)
```
&nbsp;
### 停止策略
- 检查策略启动状态;
- 调用策略类的on_stop()函数停止策略;
- 更新策略启动状态为False
- 对所有为成交的委托(市价单/限价单/本地停止单)进行撤单操作;
- 在图形化界面更新策略状态。
```
def stop_strategy(self, strategy_name: str):
"""
Stop a strategy.
"""
strategy = self.strategies[strategy_name]
if not strategy.trading:
return
# Call on_stop function of the strategy
self.call_strategy_func(strategy, strategy.on_stop)
# Change trading status of strategy to False
strategy.trading = False
# Cancel all orders of the strategy
self.cancel_all(strategy)
# Update GUI
self.put_strategy_event(strategy)
```
&nbsp;
### 编辑策略
- 重新配置策略参数字典setting
- 更新参数字典到策略中;
- 在图像化界面更新策略状态。
```
def edit_strategy(self, strategy_name: str, setting: dict):
"""
Edit parameters of a strategy.
"""
strategy = self.strategies[strategy_name]
strategy.update_setting(setting)
self.update_strategy_setting(strategy_name, setting)
self.put_strategy_event(strategy)
```
&nbsp;
### 移除策略
- 检查策略状态,只有停止策略后从可以移除策略;
- 从json文件移除策略配置信息(strategy_name, vt_symbol, setting)
- 从symbol_strategy_map字典中移除该策略订阅的合约信息
- 从strategy_orderid_map字典移除活动委托记录
- 从strategies字典移除该策略的相关配置信息。
```
def remove_strategy(self, strategy_name: str):
"""
Remove a strategy.
"""
strategy = self.strategies[strategy_name]
if strategy.trading:
self.write_log(f"策略{strategy.strategy_name}移除失败,请先停止")
return
# Remove setting
self.remove_strategy_setting(strategy_name)
# Remove from symbol strategy map
strategies = self.symbol_strategy_map[strategy.vt_symbol]
strategies.remove(strategy)
# Remove from active orderid map
if strategy_name in self.strategy_orderid_map:
vt_orderids = self.strategy_orderid_map.pop(strategy_name)
# Remove vt_orderid strategy map
for vt_orderid in vt_orderids:
if vt_orderid in self.orderid_strategy_map:
self.orderid_strategy_map.pop(vt_orderid)
# Remove from strategies
self.strategies.pop(strategy_name)
return True
```
2019-06-13 07:53:57 +00:00
&nbsp;
### 锁仓操作
用户在编写策略时可以通过填写lock字段来让策略完成锁仓操作即禁止平今通过反向开仓来代替。
- 在cta策略模板template中可以看到如下具体委托函数都有lock字段并且默认为False。
```
def buy(self, price: float, volume: float, stop: bool = False, lock: bool = False):
"""
Send buy order to open a long position.
"""
return self.send_order(Direction.LONG, Offset.OPEN, price, volume, stop, lock)
def sell(self, price: float, volume: float, stop: bool = False, lock: bool = False):
"""
Send sell order to close a long position.
"""
return self.send_order(Direction.SHORT, Offset.CLOSE, price, volume, stop, lock)
def short(self, price: float, volume: float, stop: bool = False, lock: bool = False):
"""
Send short order to open as short position.
"""
return self.send_order(Direction.SHORT, Offset.OPEN, price, volume, stop, lock)
def cover(self, price: float, volume: float, stop: bool = False, lock: bool = False):
"""
Send cover order to close a short position.
"""
return self.send_order(Direction.LONG, Offset.CLOSE, price, volume, stop, lock)
def send_order(
self,
direction: Direction,
offset: Offset,
price: float,
volume: float,
stop: bool = False,
lock: bool = False
):
"""
Send a new order.
"""
if self.trading:
vt_orderids = self.cta_engine.send_order(
self, direction, offset, price, volume, stop, lock
)
return vt_orderids
else:
return []
```
&nbsp;
- 设置lock=True后cta实盘引擎send_order()函数发生响应并且调用其最根本的委托函数send_server_order()去处理锁仓委托转换。首先是创建原始委托original_req然后调用converter文件里面OffsetConverter类的convert_order_request来进行相关转换。
```
def send_order(
self,
strategy: CtaTemplate,
direction: Direction,
offset: Offset,
price: float,
volume: float,
stop: bool,
lock: bool
):
"""
"""
contract = self.main_engine.get_contract(strategy.vt_symbol)
if not contract:
self.write_log(f"委托失败,找不到合约:{strategy.vt_symbol}", strategy)
return ""
if stop:
if contract.stop_supported:
return self.send_server_stop_order(strategy, contract, direction, offset, price, volume, lock)
else:
return self.send_local_stop_order(strategy, direction, offset, price, volume, lock)
else:
return self.send_limit_order(strategy, contract, direction, offset, price, volume, lock)
def send_limit_order(
self,
strategy: CtaTemplate,
contract: ContractData,
direction: Direction,
offset: Offset,
price: float,
volume: float,
lock: bool
):
"""
Send a limit order to server.
"""
return self.send_server_order(
strategy,
contract,
direction,
offset,
price,
volume,
OrderType.LIMIT,
lock
)
def send_server_order(
self,
strategy: CtaTemplate,
contract: ContractData,
direction: Direction,
offset: Offset,
price: float,
volume: float,
type: OrderType,
lock: bool
):
"""
Send a new order to server.
"""
# Create request and send order.
original_req = OrderRequest(
symbol=contract.symbol,
exchange=contract.exchange,
direction=direction,
offset=offset,
type=type,
price=price,
volume=volume,
)
# Convert with offset converter
req_list = self.offset_converter.convert_order_request(original_req, lock)
# Send Orders
vt_orderids = []
for req in req_list:
vt_orderid = self.main_engine.send_order(
req, contract.gateway_name)
vt_orderids.append(vt_orderid)
self.offset_converter.update_order_request(req, vt_orderid)
# Save relationship between orderid and strategy.
self.orderid_strategy_map[vt_orderid] = strategy
self.strategy_orderid_map[strategy.strategy_name].add(vt_orderid)
return vt_orderids
```
&nbsp;
- 在convert_order_request_lock()函数中,先计算今仓的量和昨可用量;然后进行判断:若有今仓,只能开仓(锁仓);无今仓时候,若平仓量小于等于昨可用,全部平昨,反之,先平昨,剩下的反向开仓。
```
def convert_order_request_lock(self, req: OrderRequest):
""""""
if req.direction == Direction.LONG:
td_volume = self.short_td
yd_available = self.short_yd - self.short_yd_frozen
else:
td_volume = self.long_td
yd_available = self.long_yd - self.long_yd_frozen
# If there is td_volume, we can only lock position
if td_volume:
req_open = copy(req)
req_open.offset = Offset.OPEN
return [req_open]
# If no td_volume, we close opposite yd position first
# then open new position
else:
open_volume = max(0, req.volume - yd_available)
req_list = []
if yd_available:
req_yd = copy(req)
if self.exchange == Exchange.SHFE:
req_yd.offset = Offset.CLOSEYESTERDAY
else:
req_yd.offset = Offset.CLOSE
req_list.append(req_yd)
if open_volume:
req_open = copy(req)
req_open.offset = Offset.OPEN
req_open.volume = open_volume
req_list.append(req_open)
return req_list
```