Example7_7.py

# Example 7-7
# # Backtesting a Year-on-Year Seasonal Trending Strategy

import numpy as np
import pandas as pd

df = pd.read_table("miscFiles/SPX_20071123.txt")
df["Date"] = df["Date"].round().astype("int")
df["Date"] = pd.to_datetime(df["Date"], format="%Y%m%d")
df.set_index("Date", inplace=True)

eomPrice = df.resample("M").last()[
    :-1
]  # End of month prices.  Need to remove last date because it isn't really end of January.
monthlyRet = eomPrice.pct_change(1, fill_method=None)

positions = np.zeros(monthlyRet.shape)

for m in range(13, monthlyRet.shape[0]):
    hasData = np.where(np.isfinite(monthlyRet.iloc[m - 12, :]))[0]
    sortidx = np.argsort(monthlyRet.iloc[m - 12, hasData])
    badData = np.where(
        np.logical_not(np.isfinite(monthlyRet.iloc[m - 1, hasData[sortidx]]))
    )[
        0
    ]  # these are indices
    sortidx.drop(sortidx.index[badData], inplace=True)
    topN = np.floor(len(sortidx) / 10).astype("int")
    positions[m - 1, hasData[sortidx.values[np.arange(0, topN)]]] = -1
    positions[m - 1, hasData[sortidx.values[np.arange(-topN, 0)]]] = 1

capital = np.nansum(np.array(pd.DataFrame(abs(positions)).shift()), axis=1)
capital[capital == 0] = 1
ret = (
    np.nansum(np.array(pd.DataFrame(positions).shift()) * np.array(monthlyRet), axis=1)
    / capital
)
ret = np.delete(ret, np.arange(13))
avgret = np.nanmean(ret) * 12
sharpe = np.sqrt(12) * np.nanmean(ret) / np.nanstd(ret)

print("Avg ann return=%f Sharpe ratio=%f" % (avgret, sharpe))
# Avg ann return=-0.012679 Sharpe ratio=-0.122247