对冲基金绩效评估

1、对冲基金绩效评估

（a）年化算术平均收益
（b）年化几何平均收益
（c）超额收益年化波动率
（d）年化夏普比率
（e）市场beta
（f ）年化市场阿尔法
（g）年度信息比率
（h）最大回撤
（i）每月超额收益的偏度
（j）每月超额收益的超额峰度

a)

import pandas as pd
import numpy as np
import statsmodels.api as sm
file = "PS_2_Performance_Measurement.xlsx"
# 读取数据
factors = pd.read_excel(file, sheet_name="factors", parse_dates=["date"], index_col="date")
factors.head()

rets = pd.read_excel(file, sheet_name="total_returns", parse_dates=["date"], index_col="date")
rets.head()

# 计算超额收益
xrets = rets.sub(factors.loc[:,"RF"], axis=0) # axis = 0 row wise
# 年化算术平均回报
avg_rets = rets.mean() * 12 
avg_rets.map("{:.2%}".format)

b)

# 年化几何平均收益
T = len(rets) / 12 # number of years
geo_avg_rets = (1 + rets).prod().pow(1 / T) - 1
geo_avg_rets.map("{:.2%}".format)

c)

# 超额收益年化波动率
vol = xrets.std() 
vol.map("{:.2%}".format)

d)

# 年化 Sharpe Ratio
sr = (xrets.mean() * 12) / vol
round(sr, 2)

e) f) g)

index_names = xrets.columns # 不同的基金策略的名字
# 建立空的 dictionaries 储存线性回归结果
alpha = {}
beta = {}
ssr = {}

for index in index_names:
    # 每个基金的超额收益
    y = xrets.loc[:,index]
    # 市场收益
    x = factors.loc[:,"Mkt-RF"]
    # 回归
    reg = sm.OLS(y, sm.add_constant(x)).fit()
    # 更新参数
    alpha.update({index: reg.params.loc["const"] * 12})
    beta.update({index: reg.params.loc["Mkt-RF"]})
    ssr.update({index: reg.ssr})
# transform dictionaries to pandas Series
alpha = pd.Series(alpha)
beta = pd.Series(beta)
ssr = pd.Series(ssr)
round(beta, 2)
alpha.map("{:.2%}".format)
ir = alpha / np.sqrt(ssr / len(xrets) * 12)
round(ir, 2)    

h)

cum_rets = (1 + rets).cumprod()
#找到t之前的最大值 最高水位线
hwm = cum_rets.expanding().max()
#回撤
dd = (hwm - cum_rets) / hwm
max_dd = dd.max()
max_dd.map("{:.2%}".format)

i)

# 偏度
skewness = xrets.skew()
round(skewness, 2)

j)

# 峰度
excess_kurtosis = xrets.kurt()
round(excess_kurtosis, 2)

# 收集所有统计数据
names = ["Ann. arith. ave. ret", "Ann. geo. ave. ret", "Ann. volatility", 
         "Ann. Sharpe ratio", "Market beta", "Ann. alpha to MKT", "Ann. IR", 
         "Max drawdown", "Skewness", "Excess kurtosis"]

# a list of pandas Series
stats = [avg_rets, geo_avg_rets, vol, sr, beta, 
         alpha, ir, max_dd, skewness, excess_kurtosis]

# create a pandas DataFrame
stats = pd.DataFrame(stats,index=names)#不能换顺序

for col in [0, 1, 2, 5, 7]: stats.iloc[col] = stats.iloc[col].map("{:.2%}".format)

for col in [3, 4, 6, 8, 9]: stats.iloc[col] = stats.iloc[col].map("{:.2f}".format)

stats

2.画图

在一张图里画出累计收益，最高水位线和回撤

import matplotlib.pyplot as plt

fund_name = "Global Mac Hedge Fund USD"

fig, ax1 = plt.subplots(figsize=(8,4), dpi=100)
# ax1 画累计收益和最高水位线
ax1.plot(cum_rets.loc[:,fund_name], color="blue", label="Cumulative return")
ax1.plot(cum_rets.loc[:,fund_name].expanding().max(), color="grey", ls="--", label="High water mark")
# ax2 画回撤
ax2 = ax1.twinx()
ax2.plot(dd.loc[:,fund_name], color="green", label="Drawdown")
# y轴标签
ax1.set_ylabel("Cumulative return")
ax2.set_ylabel("Drawdown")

ax1.legend(loc="upper left", frameon=False)
ax2.legend(loc="upper center", frameon=False)

plt.title(fund_name)
plt.show()

3. 因子模型

对于Equity Long/Short，运行两个回归：

   （i）对冲基金指数的超额收益对市场超额收益的单变量回归； 
   （ii）市场，规模，价值和momentum因素的多元回归。

i）单因子模型

fund_name = "Ln/Sh Eq Hedge Fund USD"
y = xrets.loc[:,fund_name]
x = factors.loc[:,"Mkt-RF"]
uv_regression = sm.OLS(y, sm.add_constant(x)).fit()
uv_alpha = uv_regression.params.loc["const"] * 12 * 100
uv_alpha = round(uv_alpha, 2)

ii）多因子模型

x = factors.loc[:,["Mkt-RF", "SMB", "HML", "UMD"]]
mv_regression = sm.OLS(y, sm.add_constant(x)).fit()
mv_alpha = mv_regression.params.loc["const"] * 12 * 100
round(mv_alpha, 2)

展示回归结果

from stargazer.stargazer import Stargazer
table = Stargazer([uv_regression, mv_regression])# include reg results
table.add_line("Annualized Alpha (in %)", [uv_alpha, mv_alpha])
table.covariate_order(["const", "Mkt-RF", "SMB", "HML", "UMD"])
table