Building Strategy Tear Sheets for Returns, Positions, and Attribution
Code pyfolio
Summary
The document describes a reporting workflow for analyzing a trading strategy from return data and, when available, holdings, transactions, benchmark returns, market data, and factor information. Its full report brings together return and event analysis, then adds position and transaction analysis when those inputs are supplied. Optional sections cover round trips, capacity, and factor-based performance attribution. It also supports applying a slippage adjustment to returns and estimating intraday positions when daily holdings do not represent an intraday strategy well.
The input definitions make clear that the analysis depends on correctly structured, dated data, including trade amounts and prices, dollar positions, and factor returns or loadings. Several report components are conditional on optional inputs, so the output’s scope varies with data availability. The document describes reporting and diagnostics rather than a trading strategy or a guarantee of performance assessment quality; it does not establish that its charts or statistics resolve data-quality issues, model risk, or limitations in the chosen benchmarks and assumptions.
Key ideas
- The full report combines returns analysis with position and transaction analysis when those data are available.
- Benchmark returns, slippage, intraday estimates, and bootstrap analysis are optional inputs or settings.
- Round-trip, capacity, and factor-attribution sections require additional trading or market data.
- The report’s usefulness depends on correctly aligned returns, holdings, transactions, and factor data.
Tags
Full text
# tears.py
```py
#
# Copyright 2019 Quantopian, Inc.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
from __future__ import division
import warnings
from time import time
import empyrical as ep
from IPython.display import display, Markdown
import matplotlib.gridspec as gridspec
import matplotlib.pyplot as plt
import pandas as pd
from . import _seaborn as sns
from . import capacity
from . import perf_attrib
from . import plotting
from . import pos
from . import round_trips
from . import timeseries
from . import txn
from . import utils
FACTOR_PARTITIONS = {
'style': ['momentum', 'size', 'value', 'reversal_short_term',
'volatility'],
'sector': ['basic_materials', 'consumer_cyclical', 'financial_services',
'real_estate', 'consumer_defensive', 'health_care',
'utilities', 'communication_services', 'energy', 'industrials',
'technology']
}
def timer(msg_body, previous_time):
current_time = time()
run_time = current_time - previous_time
message = "\nFinished " + msg_body + " (required {:.2f} seconds)."
print(message.format(run_time))
return current_time
def create_full_tear_sheet(returns,
positions=None,
transactions=None,
market_data=None,
benchmark_rets=None,
slippage=None,
live_start_date=None,
sector_mappings=None,
round_trips=False,
estimate_intraday='infer',
hide_positions=False,
cone_std=(1.0, 1.5, 2.0),
bootstrap=False,
unadjusted_returns=None,
turnover_denom='AGB',
set_context=True,
factor_returns=None,
factor_loadings=None,
pos_in_dollars=True,
header_rows=None,
factor_partitions=FACTOR_PARTITIONS):
"""
Generate a number of tear sheets that are useful
for analyzing a strategy's performance.
- Fetches benchmarks if needed.
- Creates tear sheets for returns, and significant events.
If possible, also creates tear sheets for position analysis
and transaction analysis.
Parameters
----------
returns : pd.Series
Daily returns of the strategy, noncumulative.
- Time series with decimal returns.
- Example:
2015-07-16 -0.012143
2015-07-17 0.045350
2015-07-20 0.030957
2015-07-21 0.004902
positions : pd.DataFrame, optional
Daily net position values.
- Time series of dollar amount invested in each position and cash.
- Days where stocks are not held can be represented by 0 or NaN.
- Non-working capital is labelled 'cash'
- Example:
index 'AAPL' 'MSFT' cash
2004-01-09 13939.3800 -14012.9930 711.5585
2004-01-12 14492.6300 -14624.8700 27.1821
2004-01-13 -13853.2800 13653.6400 -43.6375
transactions : pd.DataFrame, optional
Executed trade volumes and fill prices.
- One row per trade.
- Trades on different names that occur at the
same time will have identical indicies.
- Example:
index amount price symbol
2004-01-09 12:18:01 483 324.12 'AAPL'
2004-01-09 12:18:01 122 83.10 'MSFT'
2004-01-13 14:12:23 -75 340.43 'AAPL'
market_data : pd.DataFrame, optional
Daily market_data
- DataFrame has a multi-index index, one level is dates and another is
market_data contains volume & price, equities as columns
slippage : int/float, optional
Basis points of slippage to apply to returns before generating
tearsheet stats and plots.
If a value is provided, slippage parameter sweep
plots will be generated from the unadjusted returns.
Transactions and positions must also be passed.
- See txn.adjust_returns_for_slippage for more details.
live_start_date : datetime, optional
The point in time when the strategy began live trading,
after its backtest period. This datetime should be normalized.
hide_positions : bool, optional
If True, will not output any symbol names.
round_trips: boolean, optional
If True, causes the generation of a round trip tear sheet.
sector_mappings : dict or pd.Series, optional
Security identifier to sector mapping.
Security ids as keys, sectors as values.
estimate_intraday: boolean or str, optional
Instead of using the end-of-day positions, use the point in the day
where we have the most $ invested. This will adjust positions to
better approximate and represent how an intraday strategy behaves.
By default, this is 'infer', and an attempt will be made to detect
an intraday strategy. Specifying this value will prevent detection.
cone_std : float, or tuple, optional
If float, The standard deviation to use for the cone plots.
If tuple, Tuple of standard deviation values to use for the cone plots
- The cone is a normal distribution with this standard deviation
centered around a linear regression.
bootstrap : boolean (optional)
Whether to perform bootstrap analysis for the performance
metrics. Takes a few minutes longer.
turnover_denom : str
Either AGB or portfolio_value, default AGB.
- See full explanation in txn.get_turnover.
factor_returns : pd.Dataframe, optional
Returns by factor, with date as index and factors as columns
factor_loadings : pd.Dataframe, optional
Factor loadings for all days in the date range, with date and
ticker as index, and factors as columns.
pos_in_dollars : boolean, optional
indicates whether positions is in dollars
header_rows : dict or OrderedDict, optional
Extra rows to display at the top of the perf stats table.
set_context : boolean, optional
If True, set default plotting style context.
- See plotting.context().
factor_partitions : dict, optional
dict specifying how factors should be separated in perf attrib
factor returns and risk exposures plots
- See create_perf_attrib_tear_sheet().
"""
if (unadjusted_returns is None) and (slippage is not None) and\
(transactions is not None):
unadjusted_returns = returns.copy()
returns = txn.adjust_returns_for_slippage(returns, positions,
transactions, slippage)
positions = utils.check_intraday(estimate_intraday, returns,
positions, transactions)
create_returns_tear_sheet(
returns,
positions=positions,
transactions=transactions,
live_start_date=live_start_date,
cone_std=cone_std,
benchmark_rets=benchmark_rets,
bootstrap=bootstrap,
turnover_denom=turnover_denom,
header_rows=header_rows,
set_context=set_context)
create_interesting_times_tear_sheet(returns,
benchmark_rets=benchmark_rets,
set_context=set_context)
if positions is not None:
create_position_tear_sheet(returns, positions,
hide_positions=hide_positions,
set_context=set_context,
sector_mappings=sector_mappings,
estimate_intraday=False)
if transactions is not None:
create_txn_tear_sheet(returns, positions, transactions,
unadjusted_returns=unadjusted_returns,
estimate_intraday=False,
set_context=set_context)
if round_trips:
create_round_trip_tear_sheet(
returns=returns,
positions=positions,
transactions=transactions,
sector_mappings=sector_mappings,
estimate_intraday=False)
if market_data is not None:
create_capacity_tear_sheet(returns, positions, transactions,
market_data,
liquidation_daily_vol_limit=0.2,
last_n_days=125,
estimate_intraday=False)
if factor_returns is not None and factor_loadings is not None:
create_perf_attrib_tear_sheet(returns, positions, factor_returns,
factor_loadings, transactions,
pos_in_dollars=pos_in_dollars,
factor_partitions=factor_partitions)
@plotting.customize
def create_simple_tear_sheet(returns,
positions=None,
transactions=None,
benchmark_rets=None,
slippage=None,
estimate_intraday='infer',
live_start_date=None,
turnover_denom='AGB',
header_rows=None):
"""
Simpler version of create_full_tear_sheet; generates summary performance
statistics and important plots as a single image.
- Plots: cumulative returns, rolling beta, rolling Sharpe, underwater,
exposure, top 10 holdings, total holdings, long/short holdings,
daily turnover, transaction time distribution.
- Never accept market_data input (market_data = None)
- Never accept sector_mappings input (sector_mappings = None)
- Never perform bootstrap analysis (bootstrap = False)
- Never hide posistions on top 10 holdings plot (hide_positions = False)
- Always use default cone_std (cone_std = (1.0, 1.5, 2.0))
Parameters
----------
returns : pd.Series
Daily returns of the strategy, noncumulative.
- Time series with decimal returns.
- Example:
2015-07-16 -0.012143
2015-07-17 0.045350
2015-07-20 0.030957
2015-07-21 0.004902
positions : pd.DataFrame, optional
Daily net position values.
- Time series of dollar amount invested in each position and cash.
- Days where stocks are not held can be represented by 0 or NaN.
- Non-working capital is labelled 'cash'
- Example:
index 'AAPL' 'MSFT' cash
2004-01-09 13939.3800 -14012.9930 711.5585
2004-01-12 14492.6300 -14624.8700 27.1821
2004-01-13 -13853.2800 13653.6400 -43.6375
transactions : pd.DataFrame, optional
Executed trade volumes and fill prices.
- One row per trade.
- Trades on different names that occur at the
same time will have identical indicies.
- Example:
index amount price symbol
2004-01-09 12:18:01 483 324.12 'AAPL'
2004-01-09 12:18:01 122 83.10 'MSFT'
2004-01-13 14:12:23 -75 340.43 'AAPL'
benchmark_rets : pd.Series, optional
Daily returns of the benchmark, noncumulative.
slippage : int/float, optional
Basis points of slippage to apply to returns before generating
tearsheet stats and plots.
If a value is provided, slippage parameter sweep
plots will be generated from the unadjusted returns.
Transactions and positions must also be passed.
- See txn.adjust_returns_for_slippage for more details.
live_start_date : datetime, optional
The point in time when the strategy began live trading,
after its backtest period. This datetime should be normalized.
turnover_denom : str, optional
Either AGB or portfolio_value, default AGB.
- See full explanation in txn.get_turnover.
header_rows : dict or OrderedDict, optional
Extra rows to display at the top of the perf stats table.
set_context : boolean, optional
If True, set default plotting style context.
"""
positions = utils.check_intraday(estimate_intraday, returns,
positions, transactions)
if (slippage is not None) and (transactions is not None):
returns = txn.adjust_returns_for_slippage(returns, positions,
transactions, slippage)
always_sections = 4
positions_sections = 4 if positions is not None else 0
transactions_sections = 2 if transactions is not None else 0
live_sections = 1 if live_start_date is not None else 0
benchmark_sections = 1 if benchmark_rets is not None else 0
vertical_sections = sum([
always_sections,
positions_sections,
transactions_sections,
live_sections,
benchmark_sections,
])
if live_start_date is not None:
live_start_date = ep.utils.get_utc_timestamp(live_start_date)
plotting.show_perf_stats(returns,
benchmark_rets,
positions=positions,
transactions=transactions,
turnover_denom=turnover_denom,
live_start_date=live_start_date,
header_rows=header_rows)
fig = plt.figure(figsize=(14, vertical_sections * 6))
gs = gridspec.GridSpec(vertical_sections, 3, wspace=0.5, hspace=0.5)
ax_rolling_returns = plt.subplot(gs[:2, :])
i = 2
if benchmark_rets is not None:
ax_rolling_beta = plt.subplot(gs[i, :], sharex=ax_rolling_returns)
i += 1
ax_rolling_sharpe = plt.subplot(gs[i, :], sharex=ax_rolling_returns)
i += 1
ax_underwater = plt.subplot(gs[i, :], sharex=ax_rolling_returns)
i += 1
plotting.plot_rolling_returns(returns,
factor_returns=benchmark_rets,
live_start_date=live_start_date,
cone_std=(1.0, 1.5, 2.0),
ax=ax_rolling_returns)
ax_rolling_returns.set_title('Cumulative returns')
if benchmark_rets is not None:
plotting.plot_rolling_beta(returns, benchmark_rets, ax=ax_rolling_beta)
plotting.plot_rolling_sharpe(returns, ax=ax_rolling_sharpe)
plotting.plot_drawdown_underwater(returns, ax=ax_underwater)
if positions is not None:
# Plot simple positions tear sheet
ax_exposures = plt.subplot(gs[i, :])
i += 1
ax_top_positions = plt.subplot(gs[i, :], sharex=ax_exposures)
i += 1
ax_holdings = plt.subplot(gs[i, :], sharex=ax_exposures)
i += 1
ax_long_short_holdings = plt.subplot(gs[i, :])
i += 1
positions_alloc = pos.get_percent_alloc(positions)
plotting.plot_exposures(returns, positions, ax=ax_exposures)
plotting.show_and_plot_top_positions(returns,
positions_alloc,
show_and_plot=0,
hide_positions=False,
ax=ax_top_positions)
plotting.plot_holdings(returns, positions_alloc, ax=ax_holdings)
plotting.plot_long_short_holdings(returns, positions_alloc,
ax=ax_long_short_holdings)
if transactions is not None:
# Plot simple transactions tear sheet
ax_turnover = plt.subplot(gs[i, :])
i += 1
ax_txn_timings = plt.subplot(gs[i, :])
i += 1
plotting.plot_turnover(returns,
transactions,
positions,
turnover_denom=turnover_denom,
ax=ax_turnover)
plotting.plot_txn_time_hist(transactions, ax=ax_txn_timings)
for ax in fig.axes:
plt.setp(ax.get_xticklabels(), visible=True)
@plotting.customize
def create_returns_tear_sheet(returns, positions=None,
transactions=None,
live_start_date=None,
cone_std=(1.0, 1.5, 2.0),
benchmark_rets=None,
bootstrap=False,
turnover_denom='AGB',
header_rows=None,
return_fig=False):
"""
Generate a number of plots for analyzing a strategy's returns.
- Fetches benchmarks, then creates the plots on a single figure.
- Plots: rolling returns (with cone), rolling beta, rolling sharpe,
rolling Fama-French risk factors, drawdowns, underwater plot, monthly
and annual return plots, daily similarity plots,
and return quantile box plot.
- Will also print the start and end dates of the strategy,
performance statistics, drawdown periods, and the return range.
Parameters
----------
returns : pd.Series
Daily returns of the strategy, noncumulative.
- See full explanation in create_full_tear_sheet.
positions : pd.DataFrame, optional
Daily net position values.
- See full explanation in create_full_tear_sheet.
transactions : pd.DataFrame, optional
Executed trade volumes and fill prices.
- See full explanation in create_full_tear_sheet.
live_start_date : datetime, optional
The point in time when the strategy began live trading,
after its backtest period.
cone_std : float, or tuple, optional
If float, The standard deviation to use for the cone plots.
If tuple, Tuple of standard deviation values to use for the cone plots
- The cone is a normal distribution with this standard deviation
centered around a linear regression.
benchmark_rets : pd.Series, optional
Daily noncumulative returns of the benchmark.
- This is in the same style as returns.
bootstrap : boolean, optional
Whether to perform bootstrap analysis for the performance
metrics. Takes a few minutes longer.
turnover_denom : str, optional
Either AGB or portfolio_value, default AGB.
- See full explanation in txn.get_turnover.
header_rows : dict or OrderedDict, optional
Extra rows to display at the top of the perf stats table.
return_fig : boolean, optional
If True, returns the figure that was plotted on.
"""
if benchmark_rets is not None:
returns = utils.clip_returns_to_benchmark(returns, benchmark_rets)
plotting.show_perf_stats(returns, benchmark_rets,
positions=positions,
transactions=transactions,
turnover_denom=turnover_denom,
bootstrap=bootstrap,
live_start_date=live_start_date,
header_rows=header_rows)
plotting.show_worst_drawdown_periods(returns)
vertical_sections = 11
if live_start_date is not None:
vertical_sections += 1
live_start_date = ep.utils.get_utc_timestamp(live_start_date)
if benchmark_rets is not None:
vertical_sections += 1
if bootstrap:
vertical_sections += 1
fig = plt.figure(figsize=(14, vertical_sections * 6))
gs = gridspec.GridSpec(vertical_sections, 3, wspace=0.5, hspace=0.5)
ax_rolling_returns = plt.subplot(gs[:2, :])
i = 2
ax_rolling_returns_vol_match = plt.subplot(gs[i, :],
sharex=ax_rolling_returns)
i += 1
ax_rolling_returns_log = plt.subplot(gs[i, :],
sharex=ax_rolling_returns)
i += 1
ax_returns = plt.subplot(gs[i, :],
sharex=ax_rolling_returns)
i += 1
if benchmark_rets is not None:
ax_rolling_beta = plt.subplot(gs[i, :], sharex=ax_rolling_returns)
i += 1
ax_rolling_volatility = plt.subplot(gs[i, :], sharex=ax_rolling_returns)
i += 1
ax_rolling_sharpe = plt.subplot(gs[i, :], sharex=ax_rolling_returns)
i += 1
ax_drawdown = plt.subplot(gs[i, :], sharex=ax_rolling_returns)
i += 1
ax_underwater = plt.subplot(gs[i, :], sharex=ax_rolling_returns)
i += 1
ax_monthly_heatmap = plt.subplot(gs[i, 0])
ax_annual_returns = plt.subplot(gs[i, 1])
ax_monthly_dist = plt.subplot(gs[i, 2])
i += 1
ax_return_quantiles = plt.subplot(gs[i, :])
i += 1
plotting.plot_rolling_returns(
returns,
factor_returns=benchmark_rets,
live_start_date=live_start_date,
cone_std=cone_std,
ax=ax_rolling_returns)
ax_rolling_returns.set_title(
'Cumulative returns')
plotting.plot_rolling_returns(
returns,
factor_returns=benchmark_rets,
live_start_date=live_start_date,
cone_std=None,
volatility_match=(benchmark_rets is not None),
legend_loc=None,
ax=ax_rolling_returns_vol_match)
ax_rolling_returns_vol_match.set_title(
'Cumulative returns volatility matched to benchmark')
plotting.plot_rolling_returns(
returns,
factor_returns=benchmark_rets,
logy=True,
live_start_date=live_start_date,
cone_std=cone_std,
ax=ax_rolling_returns_log)
ax_rolling_returns_log.set_title(
'Cumulative returns on logarithmic scale')
plotting.plot_returns(
returns,
live_start_date=live_start_date,
ax=ax_returns,
)
ax_returns.set_title(
'Returns')
if benchmark_rets is not None:
plotting.plot_rolling_beta(
returns, benchmark_rets, ax=ax_rolling_beta)
plotting.plot_rolling_volatility(
returns, factor_returns=benchmark_rets, ax=ax_rolling_volatility)
plotting.plot_rolling_sharpe(
returns, ax=ax_rolling_sharpe)
# Drawdowns
plotting.plot_drawdown_periods(
returns, top=5, ax=ax_drawdown)
plotting.plot_drawdown_underwater(
returns=returns, ax=ax_underwater)
plotting.plot_monthly_returns_heatmap(returns, ax=ax_monthly_heatmap)
plotting.plot_annual_returns(returns, ax=ax_annual_returns)
plotting.plot_monthly_returns_dist(returns, ax=ax_monthly_dist)
plotting.plot_return_quantiles(
returns,
live_start_date=live_start_date,
ax=ax_return_quantiles)
if bootstrap and (benchmark_rets is not None):
ax_bootstrap = plt.subplot(gs[i, :])
plotting.plot_perf_stats(returns, benchmark_rets,
ax=ax_bootstrap)
elif bootstrap:
raise ValueError('bootstrap requires passing of benchmark_rets.')
for ax in fig.axes:
plt.setp(ax.get_xticklabels(), visible=True)
if return_fig:
return fig
@plotting.customize
def create_position_tear_sheet(returns, positions,
show_and_plot_top_pos=2, hide_positions=False,
sector_mappings=None, transactions=None,
estimate_intraday='infer', return_fig=False):
"""
Generate a number of plots for analyzing a
strategy's positions and holdings.
- Plots: gross leverage, exposures, top positions, and holdings.
- Will also print the top positions held.
Parameters
----------
returns : pd.Series
Daily returns of the strategy, noncumulative.
- See full explanation in create_full_tear_sheet.
positions : pd.DataFrame
Daily net position values.
- See full explanation in create_full_tear_sheet.
show_and_plot_top_pos : int, optional
By default, this is 2, and both prints and plots the
top 10 positions.
If this is 0, it will only plot; if 1, it will only print.
hide_positions : bool, optional
If True, will not output any symbol names.
Overrides show_and_plot_top_pos to 0 to suppress text output.
sector_mappings : dict or pd.Series, optional
Security identifier to sector mapping.
Security ids as keys, sectors as values.
transactions : pd.DataFrame, optional
Prices and amounts of executed trades. One row per trade.
- See full explanation in create_full_tear_sheet.
estimate_intraday: boolean or str, optional
Approximate returns for intraday strategies.
See description in create_full_tear_sheet.
return_fig : boolean, optional
If True, returns the figure that was plotted on.
"""
positions = utils.check_intraday(estimate_intraday, returns,
positions, transactions)
if hide_positions:
show_and_plot_top_pos = 0
vertical_sections = 7 if sector_mappings is not None else 6
fig = plt.figure(figsize=(14, vertical_sections * 6))
gs = gridspec.GridSpec(vertical_sections, 3, wspace=0.5, hspace=0.5)
ax_exposures = plt.subplot(gs[0, :])
ax_top_positions = plt.subplot(gs[1, :], sharex=ax_exposures)
ax_max_median_pos = plt.subplot(gs[2, :], sharex=ax_exposures)
ax_holdings = plt.subplot(gs[3, :], sharex=ax_exposures)
ax_long_short_holdings = plt.subplot(gs[4, :])
ax_gross_leverage = plt.subplot(gs[5, :], sharex=ax_exposures)
positions_alloc = pos.get_percent_alloc(positions)
plotting.plot_exposures(returns, positions, ax=ax_exposures)
plotting.show_and_plot_top_positions(
returns,
positions_alloc,
show_and_plot=show_and_plot_top_pos,
hide_positions=hide_positions,
ax=ax_top_positions)
plotting.plot_max_median_position_concentration(positions,
ax=ax_max_median_pos)
plotting.plot_holdings(returns, positions_alloc, ax=ax_holdings)
plotting.plot_long_short_holdings(returns, positions_alloc,
ax=ax_long_short_holdings)
plotting.plot_gross_leverage(returns, positions,
ax=ax_gross_leverage)
if sector_mappings is not None:
sector_exposures = pos.get_sector_exposures(positions,
sector_mappings)
if len(sector_exposures.columns) > 1:
sector_alloc = pos.get_percent_alloc(sector_exposures)
sector_alloc = sector_alloc.drop('cash', axis='columns')
ax_sector_alloc = plt.subplot(gs[6, :], sharex=ax_exposures)
plotting.plot_sector_allocations(returns, sector_alloc,
ax=ax_sector_alloc)
for ax in fig.axes:
plt.setp(ax.get_xticklabels(), visible=True)
if return_fig:
return fig
@plotting.customize
def create_txn_tear_sheet(returns, positions, transactions,
turnover_denom='AGB', unadjusted_returns=None,
estimate_intraday='infer', return_fig=False):
"""
Generate a number of plots for analyzing a strategy's transactions.
Plots: turnover, daily volume, and a histogram of daily volume.
Parameters
----------
returns : pd.Series
Daily returns of the strategy, noncumulative.
- See full explanation in create_full_tear_sheet.
positions : pd.DataFrame
Daily net position values.
- See full explanation in create_full_tear_sheet.
transactions : pd.DataFrame
Prices and amounts of executed trades. One row per trade.
- See full explanation in create_full_tear_sheet.
turnover_denom : str, optional
Either AGB or portfolio_value, default AGB.
- See full explanation in txn.get_turnover.
unadjusted_returns : pd.Series, optional
Daily unadjusted returns of the strategy, noncumulative.
Will plot additional swippage sweep analysis.
- See pyfolio.plotting.plot_swippage_sleep and
pyfolio.plotting.plot_slippage_sensitivity
estimate_intraday: boolean or str, optional
Approximate returns for intraday strategies.
See description in create_full_tear_sheet.
return_fig : boolean, optional
If True, returns the figure that was plotted on.
"""
positions = utils.check_intraday(estimate_intraday, returns,
positions, transactions)
vertical_sections = 6 if unadjusted_returns is not None else 4
fig = plt.figure(figsize=(14, vertical_sections * 6))
gs = gridspec.GridSpec(vertical_sections, 3, wspace=0.5, hspace=0.5)
ax_turnover = plt.subplot(gs[0, :])
ax_daily_volume = plt.subplot(gs[1, :], sharex=ax_turnover)
ax_turnover_hist = plt.subplot(gs[2, :])
ax_txn_timings = plt.subplot(gs[3, :])
plotting.plot_turnover(
returns,
transactions,
positions,
turnover_denom=turnover_denom,
ax=ax_turnover)
plotting.plot_daily_volume(returns, transactions, ax=ax_daily_volume)
try:
plotting.plot_daily_turnover_hist(transactions,
positions,
turnover_denom=turnover_denom,
ax=ax_turnover_hist)
except ValueError:
warnings.warn('Unable to generate turnover plot.', UserWarning)
plotting.plot_txn_time_hist(transactions, ax=ax_txn_timings)
if unadjusted_returns is not None:
ax_slippage_sweep = plt.subplot(gs[4, :])
plotting.plot_slippage_sweep(unadjusted_returns,
positions,
transactions,
ax=ax_slippage_sweep
)
ax_slippage_sensitivity = plt.subplot(gs[5, :])
plotting.plot_slippage_sensitivity(unadjusted_returns,
positions,
transactions,
ax=ax_slippage_sensitivity
)
for ax in fig.axes:
plt.setp(ax.get_xticklabels(), visible=True)
if return_fig:
return fig
@plotting.customize
def create_round_trip_tear_sheet(returns, positions, transactions,
sector_mappings=None,
estimate_intraday='infer', return_fig=False):
"""
Generate a number of figures and plots describing the duration,
frequency, and profitability of trade "round trips."
A round trip is started when a new long or short position is
opened and is only completed when the number of shares in that
position returns to or crosses zero.
Parameters
----------
returns : pd.Series
Daily returns of the strategy, noncumulative.
- See full explanation in create_full_tear_sheet.
positions : pd.DataFrame
Daily net position values.
- See full explanation in create_full_tear_sheet.
transactions : pd.DataFrame
Prices and amounts of executed trades. One row per trade.
- See full explanation in create_full_tear_sheet.
sector_mappings : dict or pd.Series, optional
Security identifier to sector mapping.
Security ids as keys, sectors as values.
estimate_intraday: boolean or str, optional
Approximate returns for intraday strategies.
See description in create_full_tear_sheet.
return_fig : boolean, optional
If True, returns the figure that was plotted on.
"""
positions = utils.check_intraday(estimate_intraday, returns,
positions, transactions)
transactions_closed = round_trips.add_closing_transactions(positions,
transactions)
# extract_round_trips requires BoD portfolio_value
trades = round_trips.extract_round_trips(
transactions_closed,
portfolio_value=positions.sum(axis='columns') / (1 + returns)
)
if len(trades) < 5:
warnings.warn(
"""Fewer than 5 round-trip trades made.
Skipping round trip tearsheet.""", UserWarning)
return
round_trips.print_round_trip_stats(trades)
plotting.show_profit_attribution(trades)
if sector_mappings is not None:
sector_trades = round_trips.apply_sector_mappings_to_round_trips(
trades, sector_mappings)
plotting.show_profit_attribution(sector_trades)
fig = plt.figure(figsize=(14, 3 * 6))
gs = gridspec.GridSpec(3, 2, wspace=0.5, hspace=0.5)
ax_trade_lifetimes = plt.subplot(gs[0, :])
ax_prob_profit_trade = plt.subplot(gs[1, 0])
ax_holding_time = plt.subplot(gs[1, 1])
ax_pnl_per_round_trip_dollars = plt.subplot(gs[2, 0])
ax_pnl_per_round_trip_pct = plt.subplot(gs[2, 1])
plotting.plot_round_trip_lifetimes(trades, ax=ax_trade_lifetimes)
plotting.plot_prob_profit_trade(trades, ax=ax_prob_profit_trade)
trade_holding_times = [x.days for x in trades['duration']]
sns.distplot(trade_holding_times, kde=False, ax=ax_holding_time)
ax_holding_time.set(xlabel='Holding time in days')
sns.distplot(trades.pnl, kde=False, ax=ax_pnl_per_round_trip_dollars)
ax_pnl_per_round_trip_dollars.set(xlabel='PnL per round-trip trade in $')
sns.distplot(trades.returns.dropna() * 100, kde=False,
ax=ax_pnl_per_round_trip_pct)
ax_pnl_per_round_trip_pct.set(
xlabel='Round-trip returns in %')
gs.tight_layout(fig)
if return_fig:
return fig
@plotting.customize
def create_interesting_times_tear_sheet(returns, benchmark_rets=None,
periods=None, legend_loc='best',
return_fig=False):
"""
Generate a number of returns plots around interesting points in time,
like the flash crash and 9/11.
Plots: returns around the dotcom bubble burst, Lehmann Brothers' failure,
9/11, US downgrade and EU debt crisis, Fukushima meltdown, US housing
bubble burst, EZB IR, Great Recession (August 2007, March and September
of 2008, Q1 & Q2 2009), flash crash, April and October 2014.
benchmark_rets must be passed, as it is meaningless to analyze performance
during interesting times without some benchmark to refer to.
Parameters
----------
returns : pd.Series
Daily returns of the strategy, noncumulative.
- See full explanation in create_full_tear_sheet.
benchmark_rets : pd.Series
Daily noncumulative returns of the benchmark.
- This is in the same style as returns.
periods: dict or OrderedDict, optional
historical event dates that may have had significant
impact on markets
legend_loc : plt.legend_loc, optional
The legend's location.
return_fig : boolean, optional
If True, returns the figure that was plotted on.
"""
rets_interesting = timeseries.extract_interesting_date_ranges(
returns, periods)
if not rets_interesting:
warnings.warn('Passed returns do not overlap with any'
'interesting times.', UserWarning)
return
utils.print_table(pd.DataFrame(rets_interesting)
.describe().transpose()
.loc[:, ['mean', 'min', 'max']] * 100,
name='Stress Events',
float_format='{0:.2f}%'.format)
if benchmark_rets is not None:
returns = utils.clip_returns_to_benchmark(returns, benchmark_rets)
bmark_interesting = timeseries.extract_interesting_date_ranges(
benchmark_rets, periods)
num_plots = len(rets_interesting)
# 2 plots, 1 row; 3 plots, 2 rows; 4 plots, 2 rows; etc.
num_rows = int((num_plots + 1) / 2.0)
fig = plt.figure(figsize=(14, num_rows * 6.0))
gs = gridspec.GridSpec(num_rows, 2, wspace=0.5, hspace=0.5)
for i, (name, rets_period) in enumerate(rets_interesting.items()):
# i=0 -> 0, i=1 -> 0, i=2 -> 1 ;; i=0 -> 0, i=1 -> 1, i=2 -> 0
ax = plt.subplot(gs[int(i / 2.0), i % 2])
ep.cum_returns(rets_period).plot(
ax=ax, color='forestgreen', label='algo', alpha=0.7, lw=2)
if benchmark_rets is not None:
ep.cum_returns(bmark_interesting[name]).plot(
ax=ax, color='gray', label='benchmark', alpha=0.6)
ax.legend(['Algo',
'benchmark'],
loc=legend_loc, frameon=True, framealpha=0.5)
else:
ax.legend(['Algo'],
loc=legend_loc, frameon=True, framealpha=0.5)
ax.set_title(name)
ax.set_ylabel('Returns')
ax.set_xlabel('')
if return_fig:
return fig
@plotting.customize
def create_capacity_tear_sheet(returns, positions, transactions,
market_data,
liquidation_daily_vol_limit=0.2,
trade_daily_vol_limit=0.05,
last_n_days=utils.APPROX_BDAYS_PER_MONTH * 6,
days_to_liquidate_limit=1,
estimate_intraday='infer',
return_fig=False):
"""
Generates a report detailing portfolio size constraints set by
least liquid tickers. Plots a "capacity sweep," a curve describing
projected sharpe ratio given the slippage penalties that are
applied at various capital bases.
Parameters
----------
returns : pd.Series
Daily returns of the strategy, noncumulative.
- See full explanation in create_full_tear_sheet.
positions : pd.DataFrame
Daily net position values.
- See full explanation in create_full_tear_sheet.
transactions : pd.DataFrame
Prices and amounts of executed trades. One row per trade.
- See full explanation in create_full_tear_sheet.
market_data : pd.DataFrame
Daily market_data
- DataFrame has a multi-index index, one level is dates and another is
market_data contains volume & price, equities as columns
liquidation_daily_vol_limit : float
Max proportion of a daily bar that can be consumed in the
process of liquidating a position in the
"days to liquidation" analysis.
trade_daily_vol_limit : float
Flag daily transaction totals that exceed proportion of
daily bar.
last_n_days : integer
Compute max position allocation and dollar volume for only
the last N days of the backtest
days_to_liquidate_limit : integer
Display all tickers with greater max days to liquidation.
estimate_intraday: boolean or str, optional
Approximate returns for intraday strategies.
See description in create_full_tear_sheet.
return_fig : boolean, optional
If True, returns the figure that was plotted on.
"""
positions = utils.check_intraday(estimate_intraday, returns,
positions, transactions)
print("Max days to liquidation is computed for each traded name "
"assuming a 20% limit on daily bar consumption \n"
"and trailing 5 day mean volume as the available bar volume.\n\n"
"Tickers with >1 day liquidation time at a"
" constant $1m capital base:")
max_days_by_ticker = capacity.get_max_days_to_liquidate_by_ticker(
positions, market_data,
max_bar_consumption=liquidation_daily_vol_limit,
capital_base=1e6,
mean_volume_window=5)
max_days_by_ticker.index = (
max_days_by_ticker.index.map(utils.format_asset))
print("Whole backtest:")
utils.print_table(
max_days_by_ticker[max_days_by_ticker.days_to_liquidate >
days_to_liquidate_limit])
max_days_by_ticker_lnd = capacity.get_max_days_to_liquidate_by_ticker(
positions, market_data,
max_bar_consumption=liquidation_daily_vol_limit,
capital_base=1e6,
mean_volume_window=5,
last_n_days=last_n_days)
max_days_by_ticker_lnd.index = (
max_days_by_ticker_lnd.index.map(utils.format_asset))
print("Last {} trading days:".format(last_n_days))
utils.print_table(
max_days_by_ticker_lnd[max_days_by_ticker_lnd.days_to_liquidate > 1])
llt = capacity.get_low_liquidity_transactions(transactions, market_data)
llt.index = llt.index.map(utils.format_asset)
print('Tickers with daily transactions consuming >{}% of daily bar \n'
'all backtest:'.format(trade_daily_vol_limit * 100))
utils.print_table(
llt[llt['max_pct_bar_consumed'] > trade_daily_vol_limit * 100])
llt = capacity.get_low_liquidity_transactions(
transactions, market_data, last_n_days=last_n_days)
print("Last {} trading days:".format(last_n_days))
utils.print_table(
llt[llt['max_pct_bar_consumed'] > trade_daily_vol_limit * 100])
bt_starting_capital = positions.iloc[0].sum() / (1 + returns.iloc[0])
fig, ax_capacity_sweep = plt.subplots(figsize=(14, 6))
plotting.plot_capacity_sweep(returns, transactions, market_data,
bt_starting_capital,
min_pv=100000,
max_pv=300000000,
step_size=1000000,
ax=ax_capacity_sweep)
if return_fig:
return fig
@plotting.customize
def create_perf_attrib_tear_sheet(returns,
positions,
factor_returns,
factor_loadings,
transactions=None,
pos_in_dollars=True,
factor_partitions=FACTOR_PARTITIONS,
return_fig=False):
"""
Generate plots and tables for analyzing a strategy's performance.
Parameters
----------
returns : pd.Series
Returns for each day in the date range.
positions: pd.DataFrame
Daily holdings (in dollars or percentages), indexed by date.
Will be converted to percentages if positions are in dollars.
Short positions show up as cash in the 'cash' column.
factor_returns : pd.DataFrame
Returns by factor, with date as index and factors as columns
factor_loadings : pd.DataFrame
Factor loadings for all days in the date range, with date
and ticker as index, and factors as columns.
transactions : pd.DataFrame, optional
Prices and amounts of executed trades. One row per trade.
- See full explanation in create_full_tear_sheet.
- Default is None.
pos_in_dollars : boolean, optional
Flag indicating whether `positions` are in dollars or percentages
If True, positions are in dollars.
factor_partitions : dict
dict specifying how factors should be separated in factor returns
and risk exposures plots
- Example:
{'style': ['momentum', 'size', 'value', ...],
'sector': ['technology', 'materials', ... ]}
return_fig : boolean, optional
If True, returns the figure that was plotted on.
"""
portfolio_exposures, perf_attrib_data = perf_attrib.perf_attrib(
returns, positions, factor_returns, factor_loadings, transactions,
pos_in_dollars=pos_in_dollars
)
display(Markdown("## Performance Relative to Common Risk Factors"))
# aggregate perf attrib stats and show summary table
perf_attrib.show_perf_attrib_stats(returns, positions, factor_returns,
factor_loadings, transactions,
pos_in_dollars)
# one section for the returns plot, and for each factor grouping
# one section for factor returns, and one for risk exposures
if factor_partitions is not None:
vertical_sections = 1 + 2 * max(len(factor_partitions), 1)
else:
vertical_sections = 1 + 2
current_section = 0
fig = plt.figure(figsize=[14, vertical_sections * 6])
gs = gridspec.GridSpec(vertical_sections, 1,
wspace=0.5, hspace=0.5)
perf_attrib.plot_returns(perf_attrib_data,
ax=plt.subplot(gs[current_section]))
current_section += 1
if factor_partitions is not None:
for factor_type, partitions in factor_partitions.items():
columns_to_select = perf_attrib_data.columns.intersection(
partitions
)
perf_attrib.plot_factor_contribution_to_perf(
perf_attrib_data[columns_to_select],
ax=plt.subplot(gs[current_section]),
title=(
'Cumulative common {} returns attribution'
).format(factor_type)
)
current_section += 1
for factor_type, partitions in factor_partitions.items():
columns_to_select = portfolio_exposures.columns.intersection(
partitions
)
perf_attrib.plot_risk_exposures(
portfolio_exposures[columns_to_select],
ax=plt.subplot(gs[current_section]),
title='Daily {} factor exposures'.format(factor_type)
)
current_section += 1
else:
perf_attrib.plot_factor_contribution_to_perf(
perf_attrib_data,
ax=plt.subplot(gs[current_section])
)
current_section += 1
perf_attrib.plot_risk_exposures(
portfolio_exposures,
ax=plt.subplot(gs[current_section])
)
# gs.tight_layout(fig)
if return_fig:
return fig
```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.