Backtest Reporting Metrics for Futures and Stock Accounts
Code TqSdk
Summary
This report utility converts daily account snapshots and trade records into tables, then calculates summary statistics for simulated futures accounts or stock accounts. For both account types it derives daily profit and returns, cumulative profit and loss days, annualized return, maximum drawdown, and Sharpe, Sortino, and Calmar ratios. The futures path also counts opening and closing trades, averages account risk, and estimates win rate and profit-to-loss ratio from matched trades and contract multipliers; the stock path tracks buy and sell counts, fees, and winning and losing streak lengths.
The code also generates rolling risk-adjusted ratio chart data. Its stated scope is simulation reporting, and the available input records and quote metadata constrain what can be calculated. Metrics depend on the implementation’s assumptions, including annualization conventions and trade matching. The document is code rather than an empirical study: it provides no example results, validation, or guidance for interpreting the metrics, so users should confirm definitions and edge cases before relying on reports.
Key ideas
- Daily account snapshots are combined with trade records to calculate performance statistics.
- The report includes returns, drawdown, annualized risk-adjusted ratios, and streak counts.
- Futures trade statistics use matched openings and closings plus contract multipliers.
- Stock and futures accounts use different account fields and trade metrics.
- Rolling Sharpe, Sortino, and Calmar chart data are generated, but no validation results are shown.
Tags
Full text
# report.py
```py
#!/usr/bin/env python
# -*- coding: utf-8 -*-
__author__ = 'mayanqiong'
from typing import Dict, Optional
import numpy as np
from pandas import DataFrame, Series
from tqsdk.objs import Account, Trade, SecurityAccount, SecurityTrade
from tqsdk.tafunc import get_sharp, get_sortino, get_calmar, _cum_counts
TRADING_DAYS_OF_YEAR = 250
TRADING_DAYS_OF_MONTH = 21
class TqReport(object):
"""
天勤报告类，辅助 web_gui 显示回测统计信息和统计图表
1. 目前只针对 TqSim 账户回测有意义
2. 每份报告针对一组对应的账户截面记录和成交记录
"""
def __init__(self, report_id: str, trade_log: Optional[Dict] = None, quotes: Optional[Dict] = None, account_type: str = "FUTURE"):
"""
本模块为给 TqSim 提供交易成交统计
Args:
report_id (str): 报告Id
trade_log (dict): TqSim 交易结束之后生产的每日账户截面和交易记录
{
'2020-09-01': {
"trades": [],
"account": {},
"positions": {},
'2020-09-02': {....},
}
quotes (dict): 合约信息
Example::
TODO: 补充示例
"""
self.report_id = report_id
self.trade_log = trade_log
self.quotes = quotes
self.account_type = account_type
self.date_keys = sorted(trade_log.keys())
self.account_df, self.trade_df = self._get_df()
# default metrics
self.default_metrics = self._get_default_metrics() if self.account_type == "FUTURE" else self._get_stock_metrics()
def _get_df(self):
type_account = Account if self.account_type == "FUTURE" else SecurityAccount
type_trade = Trade if self.account_type == "FUTURE" else SecurityTrade
account_data = [{'date': dt} for dt in self.date_keys]
for item in account_data:
item.update(self.trade_log[item['date']]['account'])
account_df = DataFrame(data=account_data, columns=['date'] + list(type_account(None).keys()))
trade_array = []
for date in self.date_keys:
trade_array.extend(self.trade_log[date]['trades'])
trade_df = DataFrame(data=trade_array, columns=list(type_trade(None).keys()))
if type_trade == Trade:
trade_df["offset1"] = trade_df["offset"].replace("CLOSETODAY", "CLOSE")
return account_df, trade_df
def _get_default_metrics(self):
if self.account_df.shape[0] > 0:
result = self._get_account_stat_metrics()
result.update(self._get_trades_stat_metrics())
return result
else:
return {
"winning_rate": float('nan'), # 胜率
"profit_loss_ratio": float('nan'), # 盈亏额比例
"ror": float('nan'), # 收益率
"annual_yield": float('nan'), # 年化收益率
"max_drawdown": float('nan'), # 最大回撤
"sharpe_ratio": float('nan'), # 年化夏普率
"sortino_ratio": float('nan'), # 年化索提诺比率
"commission": 0, # 总手续费
"tqsdk_punchline": ""
}
def _get_stock_metrics(self):
if self.account_df.shape[0] > 0:
init_asset = self.account_df.iloc[0]['asset_his']
asset = self.account_df.iloc[-1]['asset']
self.account_df['profit'] = self.account_df['asset'] - self.account_df['asset'].shift(fill_value=init_asset) # 每日收益
self.account_df['is_profit'] = np.where(self.account_df['profit'] > 0, 1, 0) # 是否收益
self.account_df['is_loss'] = np.where(self.account_df['profit'] < 0, 1, 0) # 是否亏损
self.account_df['daily_yield'] = self.account_df['asset'] / self.account_df['asset'].shift(fill_value=init_asset) - 1 # 每日收益率
self.account_df['max_asset'] = self.account_df['asset'].cummax() # 当前单日最大权益
self.account_df['drawdown'] = (self.account_df['max_asset'] - self.account_df['asset']) / self.account_df['max_asset'] # 回撤
_ror = asset / init_asset
return {
"start_date": self.account_df.iloc[0]["date"],
"end_date": self.account_df.iloc[-1]["date"],
"init_asset": init_asset,
"asset": init_asset,
"start_asset": init_asset,
"end_asset": asset,
"ror": _ror - 1, # 收益率
"annual_yield": _ror ** (TRADING_DAYS_OF_YEAR / self.account_df.shape[0]) - 1, # 年化收益率
"trading_days": self.account_df.shape[0], # 总交易天数
"cum_profit_days": self.account_df['is_profit'].sum(), # 累计盈利天数
"cum_loss_days": self.account_df['is_loss'].sum(), # 累计亏损天数
"max_drawdown": self.account_df['drawdown'].max(), # 最大回撤
"fee": self.account_df['buy_fee_today'].sum() + self.account_df['sell_fee_today'].sum(), # 总手续费
"buy_times": self.trade_df.loc[self.trade_df["direction"] == "BUY"].shape[0], # 买次数
"sell_times": self.trade_df.loc[self.trade_df["direction"] == "SELL"].shape[0], # 卖次数
"max_cont_profit_days": _cum_counts(self.account_df['is_profit']).max(), # 最大连续盈利天数
"max_cont_loss_days": _cum_counts(self.account_df['is_loss']).max(), # 最大连续亏损天数
"sharpe_ratio": get_sharp(self.account_df['daily_yield']), # 年化夏普率
"calmar_ratio": get_calmar(self.account_df['daily_yield'], self.account_df['drawdown'].max()), # 年化卡玛比率
"sortino_ratio": get_sortino(self.account_df['daily_yield']), # 年化索提诺比率
"tqsdk_punchline": self._get_tqsdk_punchlines(_ror - 1)
}
else:
return {
"profit_loss_ratio": float('nan'), # 盈亏额比例
"ror": float('nan'), # 收益率
"annual_yield": float('nan'), # 年化收益率
"max_drawdown": float('nan'), # 最大回撤
"sharpe_ratio": float('nan'), # 年化夏普率
"sortino_ratio": float('nan'), # 年化索提诺比率
"fee": 0, # 总手续费
"tqsdk_punchline": ""
}
def _get_account_stat_metrics(self):
init_balance = self.account_df.iloc[0]['pre_balance']
balance = self.account_df.iloc[-1]['balance']
self.account_df['profit'] = self.account_df['balance'] - self.account_df['balance'].shift(fill_value=init_balance) # 每日收益
self.account_df['is_profit'] = np.where(self.account_df['profit'] > 0, 1, 0) # 是否收益
self.account_df['is_loss'] = np.where(self.account_df['profit'] < 0, 1, 0) # 是否亏损
self.account_df['daily_yield'] = self.account_df['balance'] / self.account_df['balance'].shift(fill_value=init_balance) - 1 # 每日收益率
self.account_df['max_balance'] = self.account_df['balance'].cummax() # 当前单日最大权益
self.account_df['drawdown'] = (self.account_df['max_balance'] - self.account_df['balance']) / self.account_df['max_balance'] # 回撤
_ror = self.account_df.iloc[-1]['balance'] / self.account_df.iloc[0]['pre_balance']
return {
"start_date": self.account_df.iloc[0]["date"],
"end_date": self.account_df.iloc[-1]["date"],
"init_balance": init_balance,
"balance": balance,
"start_balance": init_balance,
"end_balance": balance,
"ror": _ror - 1, # 收益率
"annual_yield": _ror ** (TRADING_DAYS_OF_YEAR / self.account_df.shape[0]) - 1, # 年化收益率
"trading_days": self.account_df.shape[0], # 总交易天数
"cum_profit_days": self.account_df['is_profit'].sum(), # 累计盈利天数
"cum_loss_days": self.account_df['is_loss'].sum(), # 累计亏损天数
"max_drawdown": self.account_df['drawdown'].max(), # 最大回撤
"commission": self.account_df['commission'].sum(), # 总手续费
"open_times": self.trade_df.loc[self.trade_df["offset1"] == "OPEN"].shape[0], # 开仓次数
"close_times": self.trade_df.loc[self.trade_df["offset1"] == "CLOSE"].shape[0], # 平仓次数
"daily_risk_ratio": self.account_df['risk_ratio'].mean(), # 提供日均风险度
"max_cont_profit_days": _cum_counts(self.account_df['is_profit']).max(), # 最大连续盈利天数
"max_cont_loss_days": _cum_counts(self.account_df['is_loss']).max(), # 最大连续亏损天数
"sharpe_ratio": get_sharp(self.account_df['daily_yield']), # 年化夏普率
"calmar_ratio": get_calmar(self.account_df['daily_yield'], self.account_df['drawdown'].max()), # 年化卡玛比率
"sortino_ratio": get_sortino(self.account_df['daily_yield']), # 年化索提诺比率
"tqsdk_punchline": self._get_tqsdk_punchlines(_ror - 1)
}
def _get_trades_stat_metrics(self):
"""
根据成交手数计算 胜率，盈亏额比例
self.quotes 主要需要合约乘数，用于计算盈亏额
"""
trade_array = []
for date in self.date_keys:
for trade in self.trade_log[date]['trades']:
# 每一行都是 1 手的成交记录
trade_array.extend([{
"symbol": f"{trade['exchange_id']}.{trade['instrument_id']}",
"direction": trade["direction"],
"offset": "CLOSE" if trade["offset"] == "CLOSETODAY" else trade["offset"],
"price": trade["price"]
} for i in range(trade['volume'])])
trade_df = DataFrame(data=trade_array, columns=['symbol', 'direction', 'offset', 'price'])
profit_volumes = 0 # 盈利手数
loss_volumes = 0 # 亏损手数
profit_value = 0 # 盈利额
loss_value = 0 # 亏损额
all_symbols = trade_df['symbol'].drop_duplicates()
for symbol in all_symbols:
for direction in ["BUY", "SELL"]:
open_df = self._get_sub_df(trade_df, symbol, dir=direction, offset='OPEN')
close_df = self._get_sub_df(trade_df, symbol, dir=("SELL" if direction == "BUY" else "BUY"), offset='CLOSE')
close_df['profit'] = (close_df['price'] - open_df['price']) * (1 if direction == "BUY" else -1)
profit_volumes += close_df.loc[close_df['profit'] >= 0].shape[0] # 盈利手数
loss_volumes += close_df.loc[close_df['profit'] < 0].shape[0] # 亏损手数
profit_value += close_df.loc[close_df['profit'] >= 0, 'profit'].sum() * self.quotes[symbol]['volume_multiple']
loss_value += close_df.loc[close_df['profit'] < 0, 'profit'].sum() * self.quotes[symbol]['volume_multiple']
winning_rate = profit_volumes / (profit_volumes + loss_volumes) if profit_volumes + loss_volumes else 0
profit_pre_volume = profit_value / profit_volumes if profit_volumes else 0
loss_pre_volume = loss_value / loss_volumes if loss_volumes else 0
profit_loss_ratio = abs(profit_pre_volume / loss_pre_volume) if loss_pre_volume else float("inf")
return {
"profit_volumes": profit_volumes,
"loss_volumes": loss_volumes,
"profit_value": profit_value,
"loss_value": loss_value,
"winning_rate": winning_rate,
"profit_loss_ratio": profit_loss_ratio
}
def _get_tqsdk_punchlines(self, ror):
tqsdk_punchlines = [
'幸好是模拟账户，不然你就亏完啦',
'触底反弹,与其执迷修改参数，不如改变策略思路去天勤官网策略库进修',
'越挫越勇，不如去天勤量化官网策略库进修',
'不要灰心，少侠重新来过',
'策略看来小有所成',
'策略看来的得心应手',
'策略看来春风得意，堪比当代索罗斯',
'策略看来独孤求败，小心过拟合噢'
]
ror_level = [i for i, k in enumerate([-1, -0.5, -0.2, 0, 0.2, 0.5, 1]) if ror < k]
if len(ror_level) > 0:
return tqsdk_punchlines[ror_level[0]]
else:
return tqsdk_punchlines[-1]
def _get_sub_df(self, origin_df, symbol, dir, offset):
df = origin_df.where(
(origin_df['symbol'] == symbol) & (origin_df['offset'] == offset) & (origin_df['direction'] == dir))
df.dropna(inplace=True)
df.reset_index(drop=True, inplace=True)
return df
def metrics(self, **kwargs):
self.default_metrics.update(kwargs)
return [{
self.report_id: {"metrics": self.default_metrics.copy()}
}]
def full(self):
data = self.metrics()
data += self.daily_balance()
data += self.daily_profit()
data += self.drawdown()
data += self.sharp_rolling()
data += self.sortino_rolling()
# data += self.calmar_rolling()
return data
def daily_balance(self):
"""每日资金曲线"""
return [{
self.report_id: {
"charts": {
"daily_balance": {
"title": {
"left": 'center',
"text": "每日账户资金"
},
"xAxis": {
"type": 'category',
"data": self.account_df['date'].to_dict()
},
"yAxis": {
"type": 'value',
"min": 'dataMin',
"max": 'dataMax',
},
"series": {
"0": {
"data": self.account_df['balance'].map(lambda x: '%.2f' % x).to_dict(),
"type": 'line'
}
}
}
}
}
}]
def daily_profit(self):
"""每日盈亏"""
profit = Series(np.where(self.account_df['profit'] >= 0, self.account_df['profit'], float('nan'))) # 收益
loss = Series(np.where(self.account_df['profit'] < 0, self.account_df['profit'], float('nan'))) # 亏损
return [{
self.report_id: {
"charts": {
"daily_profit": {
"title": {
"left": 'center',
"text": "每日盈亏"
},
"xAxis": {
"type": 'category',
"data": self.account_df['date'].to_dict()
},
"yAxis": {
"type": 'value'
},
"series": {
"0": {
"data": profit.to_dict(),
"type": 'bar',
"itemStyle": {
"color": "#ee6666"
},
"stack": 'one',
},
"1": {
"data": loss.to_dict(),
"type": 'bar',
"itemStyle": {
"color": "#91cc75"
},
"stack": 'one',
}
}
}
}
}
}]
def drawdown(self):
"""回撤"""
return [{
self.report_id: {
"charts": {
"drawdown": {
"title": {
"left": 'center',
"text": "回撤"
},
"xAxis": {
"type": 'category',
"data": self.account_df['date'].to_dict()
},
"yAxis": {
"type": 'value'
},
"series": {
"0": {
"data": self.account_df['drawdown'].to_dict(),
"type": 'line'
}
}
}
}
}
}]
def sharp_rolling(self):
"""滚动夏普比率图表"""
rolling_sharp = self.account_df['daily_yield'].rolling(TRADING_DAYS_OF_MONTH).apply(get_sharp)
return [{
self.report_id: {
"charts": {
"sharp_rolling": {
"title": {
"left": 'center',
"text": "滚动夏普比率图表"
},
"xAxis": {
"type": 'category',
"data": self.account_df['date'].to_dict()
},
"yAxis": {
"type": 'value'
},
"series": {
"0": {
"data": rolling_sharp.to_dict(),
"type": 'line'
}
}
}
}
}
}]
def sortino_rolling(self):
"""滚动索提诺比率图表"""
rolling_sortino = self.account_df['daily_yield'].rolling(TRADING_DAYS_OF_MONTH).apply(get_sortino)
return [{
self.report_id: {
"charts": {
"sortino_rolling": {
"title": {
"left": 'center',
"text": "滚动索提诺比率图表"
},
"xAxis": {
"type": 'category',
"data": self.account_df['date'].to_dict()
},
"yAxis": {
"type": 'value'
},
"series": {
"0": {
"name": "滚动索提诺比率图表",
"data": rolling_sortino.to_dict(),
"type": 'line'
}
}
}
}
}
}]
def calmar_rolling(self):
"""滚动卡玛比率图表"""
rolling_calmar = self.account_df['daily_yield'].rolling(TRADING_DAYS_OF_MONTH).apply(
lambda x: get_calmar(x, self.account_df.loc[x.index]['drawdown'].max()))
return [{
self.report_id: {
"charts": {
"calmar_rolling": {
"title": {
"left": 'center',
"text": "滚动卡玛比率图表"
},
"xAxis": {
"type": 'category',
"data": self.account_df['date'].to_dict()
},
"yAxis": {
"type": 'value'
},
"series": {
"0": {
"data": rolling_calmar.to_dict(),
"type": 'line'
},
}
}
}
}
}]
```Shown in full with attribution under the source's licence. Licence: Apache-2.0
This summary was written by Stratmill's research agent from the original; it is not a copy of the source.