#!/usr/bin/env python3
# coding: utf-8
""" Functions to plot backtest. """
# Built-in packages
# External packages
import numpy as np
import seaborn as sns
from matplotlib import pyplot as plt
# Internal packages
from fynance.backtest.print_stats import set_text_stats
from fynance.features.metrics import drawdown, roll_sharpe
from fynance.features.money_management import iso_vol
# Set plot style
plt.style.use('seaborn-v0_8')
__all__ = ['display_perf']
# =========================================================================== #
# Printer Tools #
# =========================================================================== #
def compute_perf(logret, signal, fee):
fees = np.zeros(logret.shape)
fees[1:] = (signal[1:] - signal[:-1]) * fee
pctret = np.exp(logret) - 1 - fees
return np.cumprod(pctret * signal + 1)
[docs]
def display_perf(
y_idx, y_est, period=252, title='', params_iv={},
plot_drawdown=True, plot_roll_sharpe=True, x_axis=None,
underlying='Underlying', win=252, fees=0,
):
""" Plot performance and print KPI of backtest.
Print dynamic plot of performance indicators (perf, rolling sharpe
and draw down) of a strategy (raw and iso-volatility) versus its
underlying.
Parameters
----------
y_idx : np.ndarray[np.float64, ndim=1]
Time series of log-returns of the underlying.
y_est : np.ndarray[np.float64, ndim=1]
Time series of the signal's strategy.
period : int, optional
Number of period per year. Default is 252.
title : str or list of str, optional
Title of performance strategy, default is empty.
plot_drawdown : bool, optional
If true plot drawdowns, default is True.
plot_roll_sharpe : bool, optional
If true plot rolling sharpe ratios, default is True.
x_axis : list or np.asarray, optional
x-axis to plot (e.g. list of dates).
underlying : str, optional
Name of the underlying, default is 'Underlying'.
win : int, optional
Size of the window of rolling sharpe, default is 252.
fees : float, optional
Fees to apply at the strategy performance.
Returns
-------
perf_idx : np.ndarray[np.float64, ndim=1]
Time series of underlying performance.
perf_est : np.ndarray[np.float64, ndim=1]
Time series of raw strategy performance.
perf_ivo : np.ndarray[np.float64, ndim=1]
Time series of iso-vol strategy performance.
See Also
--------
PlotBackTest, set_text_stats
"""
if x_axis is None:
x_axis = range(y_idx.size)
# Compute perf.
perf_idx = np.exp(np.cumsum(y_idx))
perf_est = compute_perf(y_idx, y_est, fees)
iv = iso_vol(np.exp(np.cumsum(y_idx)), **params_iv)
perf_ivo = compute_perf(y_idx, y_est * iv, fees)
# Print stats. table
txt = set_text_stats(
y_idx, period=period,
Strategy=y_est,
Strat_IsoVol=y_est * iv,
underlying=underlying,
fees=fees
)
print(txt)
# Plot results
n = 1 + plot_roll_sharpe + plot_drawdown
f, ax = plt.subplots(n, 1, figsize=(9, 6), sharex=True)
if n == 1:
ax_perf, ax_dd, ax_dd = ax, None, None
ax_perf.set_xlabel('Date')
elif n == 2:
ax_perf = ax[0]
(ax_dd, ax_roll) = (ax[1], None) if plot_drawdown else (None, ax[1])
ax[-1].set_xlabel('Date')
else:
ax_perf, ax_dd, ax_roll = ax[0], ax[1], ax[2]
# Plot performances
ax_perf.plot(
x_axis,
100 * perf_est,
color=sns.xkcd_rgb["pale red"],
LineWidth=2.
)
ax_perf.plot(
x_axis,
100 * perf_ivo,
color=sns.xkcd_rgb["medium green"],
LineWidth=1.8
)
ax_perf.plot(
x_axis,
100 * perf_idx,
color=sns.xkcd_rgb["denim blue"],
LineWidth=1.5
)
# Set notify motion function
def motion(event):
N = len(ax_perf.lines[0].get_ydata())
w, h = f.get_size_inches() * f.dpi - 200
x = max(event.x - 100, 0)
j = int(x / w * N)
ax_perf.legend([
'Strategy: {:.0f} %'.format(ax_perf.lines[0].get_ydata()[j] - 100),
'Strat Iso-Vol: {:.0f} %'.format(ax_perf.lines[1].get_ydata()[j] - 100),
'{}: {:.0f} %'.format(underlying, ax_perf.lines[2].get_ydata()[j] - 100),
], loc='upper left', frameon=True, fontsize=10)
if plot_drawdown:
ax_dd.legend([
'Strategy: {:.2f} %'.format(ax_dd.lines[0].get_ydata()[j]),
'Strat Iso-Vol: {:.2f} %'.format(ax_dd.lines[1].get_ydata()[j]),
'{}: {:.2f} %'.format(underlying, ax_dd.lines[2].get_ydata()[j]),
], loc='upper left', frameon=True, fontsize=10)
if plot_roll_sharpe:
ax_roll.legend([
'Strategy: {:.2f}'.format(ax_roll.lines[0].get_ydata()[j]),
'Strat Iso-Vol: {:.2f}'.format(ax_roll.lines[1].get_ydata()[j]),
'{}: {:.2f}'.format(underlying, ax_roll.lines[2].get_ydata()[j]),
], loc='upper left', frameon=True, fontsize=10)
ax_perf.legend(
['Strategy', 'Strat Iso-Vol', underlying],
loc='upper left', frameon=True, fontsize=10
)
ax_perf.set_ylabel('Perf.')
ax_perf.set_yscale('log')
ax_perf.set_title(title)
ax_perf.tick_params(axis='x', rotation=30, labelsize=10)
# Plot DrawDowns
if plot_drawdown:
ax_dd.plot(
x_axis,
100 * drawdown(perf_est),
color=sns.xkcd_rgb["pale red"],
LineWidth=1.4
)
ax_dd.plot(
x_axis,
100 * drawdown(perf_ivo),
color=sns.xkcd_rgb["medium green"],
LineWidth=1.2
)
ax_dd.plot(
x_axis,
100 * drawdown(perf_idx),
color=sns.xkcd_rgb["denim blue"],
LineWidth=1.
)
ax_dd.set_ylabel('% DrawDown')
ax_dd.set_title('DrawDown in percentage')
ax_dd.tick_params(axis='x', rotation=30, labelsize=10)
# Plot rolling Sharpe ratio
if plot_roll_sharpe:
ax_roll.plot(
x_axis,
roll_sharpe(perf_est, period=period, w=win),
color=sns.xkcd_rgb["pale red"],
LineWidth=1.4
)
ax_roll.plot(
x_axis,
roll_sharpe(perf_ivo, period=period, w=win),
color=sns.xkcd_rgb["medium green"],
LineWidth=1.2
)
ax_roll.plot(
x_axis,
roll_sharpe(perf_idx, period=period, w=win),
color=sns.xkcd_rgb["denim blue"],
LineWidth=1.
)
ax_roll.set_ylabel('Sharpe ratio')
ax_roll.set_yscale('log')
ax_roll.set_xlabel('Date')
ax_roll.set_title('Rolling Sharpe ratio')
ax_roll.tick_params(axis='x', rotation=30, labelsize=10)
f.canvas.mpl_connect('motion_notify_event', motion)
plt.show()
return perf_idx, perf_est, perf_ivo
