Portfolio analytics for quants, written in Python

test.xlsx Below two functions are not working in attached data.

1. qs.plots.snapshot()
2. qs.reports.full()

Function returns: OverflowError: cannot convert float infinity to integer

Hi, I'm experiencing problems simply using the example from the README to make a report, ie.:

%matplotlib inline
import quantstats as qs

# extend pandas functionality with metrics, etc.
qs.extend_pandas()

# fetch the daily returns for a stock
stock = qs.utils.download_returns('FB')

qs.reports.html(stock, "SPY)

This results in

---------------------------------------------------------------------------
ValueError                                Traceback (most recent call last)
<ipython-input-112-5ce6901da6d9> in <module>
----> 1 qs.reports.html(stock, benchmark)

~/WORKSPACE/venv/lib/python3.8/site-packages/quantstats/reports.py in html(returns, benchmark, rf, grayscale, title, output, compounded)
     56     tpl = tpl.replace('{{v}}', __version__)
     57 
---> 58     mtrx = metrics(returns=returns, benchmark=benchmark,
     59                    rf=rf, display=False, mode='full',
     60                    sep=True, internal="True",

~/WORKSPACE/venv/lib/python3.8/site-packages/quantstats/reports.py in metrics(returns, benchmark, rf, display, mode, sep, compounded, **kwargs)
    297 
    298     # return df
--> 299     dd = _calc_dd(df, display=(display or "internal" in kwargs))
    300 
    301     metrics = _pd.DataFrame()

~/WORKSPACE/venv/lib/python3.8/site-packages/quantstats/reports.py in _calc_dd(df, display)
    573                 by='max drawdown', ascending=True
    574             )['max drawdown'].values[0] / pct,
--> 575             'Longest DD Days': str(round(ret_dd.sort_values(
    576                 by='days', ascending=False)['days'].values[0])),
    577             'Avg. Drawdown %': ret_dd['max drawdown'].mean() / pct,

ValueError: cannot convert float NaN to integer

Any help would be much appreciated.

The 6 months rolling do not work when there is not enough data.

Here is the stack trace it generate:

...
  File "C:\Users\cacer\Desktop\datacrunch-trading-cli\backtest\export\quants.py", line 84, in finalize
    quantstats.reports.html(merged.daily_profit_pct, merged.close, output=True, download_filename=self.html_output_file)
  File "C:\Users\cacer\AppData\Local\Programs\Python\Python39\lib\site-packages\quantstats\reports.py", line 192, in html
    _plots.rolling_beta(returns, benchmark, grayscale=grayscale,
  File "C:\Users\cacer\AppData\Local\Programs\Python\Python39\lib\site-packages\quantstats\_plotting\wrappers.py", line 514, in rolling_beta
    fig = _core.plot_rolling_beta(returns, benchmark,
  File "C:\Users\cacer\AppData\Local\Programs\Python\Python39\lib\site-packages\quantstats\_plotting\core.py", line 515, in plot_rolling_beta
    mmin = min([-100, int(beta.min()*100)])
ValueError: cannot convert float NaN to integer

A simple fix can be to do fillna() on the beta after this line: https://github.com/ranaroussi/quantstats/blob/bfc247071fa1b80772cb101f2b31b72b8227cabf/quantstats/_plotting/core.py#L506

I can do a pull request if you want. Please tell me if you found anything that would make this fix incorrect.

Hi there,

Running the simple code below:

import quantstats as qs

# fetch the daily returns for a stock
stock = qs.utils.download_returns('FB')

qs.reports.plots(mode="full",returns=stock)

returns the following error:

  File "C:\Users\rapha\OneDrive\Dokumente\GitHub\algorithms\ml4t\tmp.py", line 16, in <module>
    qs.reports.plots(mode="full",returns=stock)

  File "C:\Users\rapha\anaconda3\envs\py36-backtrader\lib\site-packages\quantstats\reports.py", line 520, in plots
    show=True, ylabel=False)

  File "C:\Users\rapha\anaconda3\envs\py36-backtrader\lib\site-packages\quantstats\_plotting\wrappers.py", line 381, in yearly_returns
    savefig=savefig, show=show)

  File "C:\Users\rapha\anaconda3\envs\py36-backtrader\lib\site-packages\quantstats\_plotting\core.py", line 111, in plot_returns_bars
    df.index.date[:1][0].strftime('%Y'),

AttributeError: 'Int64Index' object has no attribute 'date'

If I extend pandas functionnality as suggest in the Quick Start example (qs.extend_pandas), I get a different error at the same place in the code:

  File "C:\Users\rapha\OneDrive\Dokumente\GitHub\algorithms\ml4t\tmp.py", line 16, in <module>
    qs.reports.plots(mode="full",returns=stock)

  File "C:\Users\rapha\anaconda3\envs\py36-backtrader\lib\site-packages\quantstats\reports.py", line 520, in plots
    show=True, ylabel=False)

  File "C:\Users\rapha\anaconda3\envs\py36-backtrader\lib\site-packages\quantstats\_plotting\wrappers.py", line 381, in yearly_returns
    savefig=savefig, show=show)

  File "C:\Users\rapha\anaconda3\envs\py36-backtrader\lib\site-packages\quantstats\_plotting\core.py", line 111, in plot_returns_bars
    df.index.date[:1][0].strftime('%Y'),

AttributeError: 'Int64Index' object has no attribute 'date'

Any idea what the issue is? My environment seems to fit the requirements:

• Python = 3.6.12
• pandas = 1.1.3
• numpy = 1.19.2
• scipy = 1.5.2
• matplotlib = 3.2.2
• seaborn = 0.11.1
• tabulate = 0.8.7
• yfinance = 0.1.55
• plotly = 4.14.3

Best, Raphaël

I noticed a couple of calculation issues with the rolling_sharpe() and rolling_sortino() functions used in the tearsheet graphs:

1. In both rolling_sharpe() and rolling_sortino(): The period variable is used to specify the number of days in the rolling window (by default 126). However, period is also used to annualize the daily returns data for returns and benchmarks, which is incorrect. Irrespective of the number of days in the rolling window, the returns are still daily returns, so the factor should be 252. [For #74, a new variable would need to be created -- period can't do double duty as both the number of datapoints in the rolling window, AND the timespan that each datapoint represents :-) ]
2. In rolling_sortino() in wrappers.py, the target downside deviation is not calculated as per the Red Rock Capital whitepaper cited in the sortino() function in stats.py. rolling_sortino() throws away all the positive returns and only takes the standard deviation of the negative returns, which is the opposite of the whitepaper's recommendation. I'm currently using this in rolling_sortino() as a workaround:

   from empyrical import roll_sortino_ratio
   returns = roll_sortino_ratio(returns, period)
   

I'd be happy to provide a pull request for these items, just let me know.

qs.reports.full(stock,benchmark='^NSEI',rf=0.065)

Performance Metrics Strategy Benchmark

Start Period 2014-11-14 2014-11-14 End Period 2019-11-14 2019-11-14 Risk-Free Rate 0.06% 0.06% Time in Market 100.0% 100.0%

Cumulative Return -100.0% -100.0% CAGR% -100.0% -100.0% Sharpe -62.45 -118.52 Sortino -15.39 -15.73

Easily build an index for a given dictionary of weights and lookback period.

Example

FAANG = qs.utils.make_index(
    {
        'FB': 0.2,
        'AAPL': 0.2,
        'AMZN': 0.2,
        'NFLX': 0.2,
        'GOOG': 0.2
    },
)

qs.plots.snapshot(FAANG.dropna(), "SPY")

Output:

Hi team, I have tried to run reports functions on jupyter notebook. However, there's no outpute for qs.reports.html(stock, "SPY"), could any one advise what's missing? Thanks `

• qs.reports.metrics(mode='basic|full", ...) - shows basic/full metrics
• qs.reports.plots(mode='basic|full", ...) - shows basic/full plots
• qs.reports.basic(...) - shows basic metrics and plots
• qs.reports.full(...) - shows full metrics and plots
• qs.reports.html(...) - generates a complete report as html`

Is there any reason why you calculate profit factor as

return abs(returns.sum() / returns[returns < 0].sum())

as opposed to:

return abs(returns[returns > 0].sum() / returns[returns < 0].sum())

I thought normally profit factor is total profit / total loss

Hello,

First of all, thank you very much for releasing this lib publicly. I've been using it since it's so much simpler to use than many other ones.

I did come across a problem when there aren't any drawdowns or if there are fewer than 5 drawdowns. The problem is in at least two places that I've seen.

In stats._drawdown_details(drawdown), when returning None here, some code later expects a DataFrame and therefore it fails.

# no drawdown :)
if len(starts) == 0:
    return None

Even if above we return an empty DataFrame with the right column names, the code that uses this function still fails because range(1,6) is assigned to an index: E.g. in reports.full()

dd_info = _stats.drawdown_details(dd).sort_values(
        by='max drawdown', ascending=True)[:5]
dd_info.index = range(1, 6)   # <-- Fails here

Hi I've added to the metrics the Serenity index from this papar: https://www.keyquant.com/Download/GetFile?Filename=%5CPublications%5CKeyQuant_WhitePaper_APT_Part1.pdf also in attachment. KeyQuant_WhitePaper_APT_Part1.pdf

I've also modified the UPI calculation, invoking the ulcer index function, removing copied code and calculating the drawdown on prices and not on daily returns

Ciao, Giuseppe

import quantstats quantstats.reports.full(navData, benchmark= indexData)# both have NAV data respectively for Strategy and Index

Versions: quantstats- '0.0.59' matplotlib- '3.6.2' Python- '3.9'

ValueError Traceback (most recent call last) ~\AppData\Local\Temp\ipykernel_13940\3446853376.py in ----> 1 quantstats.reports.full(navData, benchmark= indexData)

~\AppData\Roaming\Python\Python39\site-packages\quantstats\reports.py in full(returns, benchmark, rf, grayscale, figsize, display, compounded, periods_per_year, match_dates) 314 print('\n\n') 315 print('[Strategy Visualization]\nvia Matplotlib') --> 316 plots(returns=returns, benchmark=benchmark, 317 grayscale=grayscale, figsize=figsize, mode='full', 318 periods_per_year=periods_per_year, prepare_returns=False)

~\AppData\Roaming\Python\Python39\site-packages\quantstats\reports.py in plots(returns, benchmark, grayscale, figsize, mode, compounded, periods_per_year, prepare_returns, match_dates) 680 show=True, ylabel=False, 681 prepare_returns=False) --> 682 _plots.yearly_returns(returns, benchmark, 683 grayscale=grayscale, 684 figsize=(figsize[0], figsize[0]*.5),

~\AppData\Roaming\Python\Python39\site-packages\quantstats_plotting\wrappers.py in yearly_returns(returns, benchmark, fontname, grayscale, hlw, hlcolor, hllabel, match_volatility, log_scale, figsize, ylabel, subtitle, compounded, savefig, show, prepare_returns) 386 returns = returns.resample('A').last() 387 --> 388 fig = _core.plot_returns_bars(returns, benchmark, 389 fontname=fontname, 390 hline=returns.mean(),

~\AppData\Roaming\Python\Python39\site-packages\quantstats_plotting\core.py in plot_returns_bars(returns, benchmark, returns_label, hline, hlw, hlcolor, hllabel, resample, title, match_volatility, log_scale, figsize, grayscale, fontname, ylabel, subtitle, savefig, show) 84 # --------------- 85 colors, _, _ = _get_colors(grayscale) ---> 86 df = _pd.DataFrame(index=returns.index, data={returns_label: returns}) 87 if isinstance(benchmark, _pd.Series): 88 df['Benchmark'] = benchmark[benchmark.index.isin(returns.index)]

C:\ProgramData\Anaconda3\lib\site-packages\pandas\core\frame.py in init(self, data, index, columns, dtype, copy) 634 elif isinstance(data, dict): 635 # GH#38939 de facto copy defaults to False only in non-dict cases --> 636 mgr = dict_to_mgr(data, index, columns, dtype=dtype, copy=copy, typ=manager) 637 elif isinstance(data, ma.MaskedArray): 638 import numpy.ma.mrecords as mrecords

C:\ProgramData\Anaconda3\lib\site-packages\pandas\core\internals\construction.py in dict_to_mgr(data, index, columns, dtype, typ, copy) 500 # TODO: can we get rid of the dt64tz special case above? 501 --> 502 return arrays_to_mgr(arrays, columns, index, dtype=dtype, typ=typ, consolidate=copy) 503 504

C:\ProgramData\Anaconda3\lib\site-packages\pandas\core\internals\construction.py in arrays_to_mgr(arrays, columns, index, dtype, verify_integrity, typ, consolidate) 123 124 # don't force copy because getting jammed in an ndarray anyway --> 125 arrays = _homogenize(arrays, index, dtype) 126 # _homogenize ensures 127 # - all(len(x) == len(index) for x in arrays)

C:\ProgramData\Anaconda3\lib\site-packages\pandas\core\internals\construction.py in _homogenize(data, index, dtype) 623 val = lib.fast_multiget(val, oindex._values, default=np.nan) 624 --> 625 val = sanitize_array( 626 val, index, dtype=dtype, copy=False, raise_cast_failure=False 627 )

C:\ProgramData\Anaconda3\lib\site-packages\pandas\core\construction.py in sanitize_array(data, index, dtype, copy, raise_cast_failure, allow_2d) 599 subarr = maybe_infer_to_datetimelike(subarr) 600 --> 601 subarr = _sanitize_ndim(subarr, data, dtype, index, allow_2d=allow_2d) 602 603 if isinstance(subarr, np.ndarray):

C:\ProgramData\Anaconda3\lib\site-packages\pandas\core\construction.py in _sanitize_ndim(result, data, dtype, index, allow_2d) 650 if allow_2d: 651 return result --> 652 raise ValueError("Data must be 1-dimensional") 653 if is_object_dtype(dtype) and isinstance(dtype, ExtensionDtype): 654 # i.e. PandasDtype("O")

ValueError: Data must be 1-dimensional

I have a problem using the download.returns function with the non default option for the period argument.

It is also not very clear from the code what the period argument is supposed to do. I take it that is there to start collecting data after the specified date.

For example, I doing this and I get nothing.

start_date = '2020-11-21'
qs.utils.download_returns('SPY', period=start_date)

I am running quantstats with a custom pandas dataframe of daily returns values from an algo that I am developing. The returns data is formatted like the download from the y-finance with the first column as daily time stamps and the second column with the daily returns. When I run (below) I get the following error.

qs.reports.full(returns)

returns is a pandas data frame from June 2022 to end of December 2022.

        Strategy_Returns

time
2022-06-01 0.000000 2022-06-02 0.000000 2022-06-03 0.000000 2022-06-04 0.000000 2022-06-05 0.000000 ... ... 2022-12-23 0.000000 2022-12-24 0.000000 2022-12-25 -0.000307 2022-12-26 0.015963 2022-12-27 0.004344

[210 rows x 1 columns]

Any insights as to what is the underlying cause of the error would be appreciated or does my data set perhaps expose a bug in the library?

ValueError Traceback (most recent call last) Input In [274], in <cell line: 1>() ----> 1 qs.reports.full(returns)

File ~/opt/anaconda3/lib/python3.9/site-packages/quantstats/reports.py:317, in full(returns, benchmark, rf, grayscale, figsize, display, compounded, periods_per_year, match_dates) 314 print('\n\n') 315 print('[Strategy Visualization]\nvia Matplotlib') --> 317 plots(returns=returns, benchmark=benchmark, 318 grayscale=grayscale, figsize=figsize, mode='full', 319 periods_per_year=periods_per_year, prepare_returns=False)

File ~/opt/anaconda3/lib/python3.9/site-packages/quantstats/reports.py:684, in plots(returns, benchmark, grayscale, figsize, mode, compounded, periods_per_year, prepare_returns, match_dates) 677 if benchmark is not None: 678 _plots.returns(returns, benchmark, match_volatility=True, 679 grayscale=grayscale, 680 figsize=(figsize[0], figsize[0].5), 681 show=True, ylabel=False, 682 prepare_returns=False) --> 684 _plots.yearly_returns(returns, benchmark, 685 grayscale=grayscale, 686 figsize=(figsize[0], figsize[0].5), 687 show=True, ylabel=False, 688 prepare_returns=False) 690 _plots.histogram(returns, grayscale=grayscale, 691 figsize=(figsize[0], figsize[0].5), 692 show=True, ylabel=False, 693 prepare_returns=False) 695 _plots.daily_returns(returns, grayscale=grayscale, 696 figsize=(figsize[0], figsize[0].3), 697 show=True, ylabel=False, 698 prepare_returns=False)

File ~/opt/anaconda3/lib/python3.9/site-packages/quantstats/_plotting/wrappers.py:388, in yearly_returns(returns, benchmark, fontname, grayscale, hlw, hlcolor, hllabel, match_volatility, log_scale, figsize, ylabel, subtitle, compounded, savefig, show, prepare_returns) 385 returns = returns.resample('A').apply(_df.sum) 386 returns = returns.resample('A').last() --> 388 fig = _core.plot_returns_bars(returns, benchmark, 389 fontname=fontname, 390 hline=returns.mean(), 391 hlw=hlw, 392 hllabel=hllabel, 393 hlcolor=hlcolor, 394 match_volatility=match_volatility, 395 log_scale=log_scale, 396 resample=None, 397 title=title, 398 figsize=figsize, 399 grayscale=grayscale, 400 ylabel=ylabel, 401 subtitle=subtitle, 402 savefig=savefig, show=show) 403 if not show: 404 return fig

File ~/opt/anaconda3/lib/python3.9/site-packages/quantstats/_plotting/core.py:86, in plot_returns_bars(returns, benchmark, returns_label, hline, hlw, hlcolor, hllabel, resample, title, match_volatility, log_scale, figsize, grayscale, fontname, ylabel, subtitle, savefig, show) 84 # --------------- 85 colors, _, _ = _get_colors(grayscale) ---> 86 df = _pd.DataFrame(index=returns.index, data={returns_label: returns}) 87 if isinstance(benchmark, _pd.Series): 88 df['Benchmark'] = benchmark[benchmark.index.isin(returns.index)]

File ~/opt/anaconda3/lib/python3.9/site-packages/pandas/core/frame.py:636, in DataFrame.init(self, data, index, columns, dtype, copy) 630 mgr = self._init_mgr( 631 data, axes={"index": index, "columns": columns}, dtype=dtype, copy=copy 632 ) 634 elif isinstance(data, dict): 635 # GH#38939 de facto copy defaults to False only in non-dict cases --> 636 mgr = dict_to_mgr(data, index, columns, dtype=dtype, copy=copy, typ=manager) 637 elif isinstance(data, ma.MaskedArray): 638 import numpy.ma.mrecords as mrecords

File ~/opt/anaconda3/lib/python3.9/site-packages/pandas/core/internals/construction.py:502, in dict_to_mgr(data, index, columns, dtype, typ, copy) 494 arrays = [ 495 x 496 if not hasattr(x, "dtype") or not isinstance(x.dtype, ExtensionDtype) 497 else x.copy() 498 for x in arrays 499 ] 500 # TODO: can we get rid of the dt64tz special case above? --> 502 return arrays_to_mgr(arrays, columns, index, dtype=dtype, typ=typ, consolidate=copy)

File ~/opt/anaconda3/lib/python3.9/site-packages/pandas/core/internals/construction.py:125, in arrays_to_mgr(arrays, columns, index, dtype, verify_integrity, typ, consolidate) 122 index = ensure_index(index) 124 # don't force copy because getting jammed in an ndarray anyway --> 125 arrays = _homogenize(arrays, index, dtype) 126 # _homogenize ensures 127 # - all(len(x) == len(index) for x in arrays) 128 # - all(x.ndim == 1 for x in arrays) (...) 131 132 else: 133 index = ensure_index(index)

File ~/opt/anaconda3/lib/python3.9/site-packages/pandas/core/internals/construction.py:625, in _homogenize(data, index, dtype) 622 val = dict(val) 623 val = lib.fast_multiget(val, oindex._values, default=np.nan) --> 625 val = sanitize_array( 626 val, index, dtype=dtype, copy=False, raise_cast_failure=False 627 ) 628 com.require_length_match(val, index) 630 homogenized.append(val)

File ~/opt/anaconda3/lib/python3.9/site-packages/pandas/core/construction.py:601, in sanitize_array(data, index, dtype, copy, raise_cast_failure, allow_2d) 598 subarr = cast(np.ndarray, subarr) 599 subarr = maybe_infer_to_datetimelike(subarr) --> 601 subarr = _sanitize_ndim(subarr, data, dtype, index, allow_2d=allow_2d) 603 if isinstance(subarr, np.ndarray): 604 # at this point we should have dtype be None or subarr.dtype == dtype 605 dtype = cast(np.dtype, dtype)

File ~/opt/anaconda3/lib/python3.9/site-packages/pandas/core/construction.py:652, in _sanitize_ndim(result, data, dtype, index, allow_2d) 650 if allow_2d: 651 return result --> 652 raise ValueError("Data must be 1-dimensional") 653 if is_object_dtype(dtype) and isinstance(dtype, ExtensionDtype): 654 # i.e. PandasDtype("O") 656 result = com.asarray_tuplesafe(data, dtype=np.dtype("object"))

ValueError: Data must be 1-dimensional

​

As it said in readme:

%matplotlib inline import quantstats as qs

qs.extend_pandas()

stock = qs.utils.download_returns('FB')

qs.stats.sharpe(stock)

stock.sharpe()

the execution result:

Got error from yahoo api for ticker FB, Error: {'code': 'Not Found', 'description': 'No data found, symbol may be delisted'}

• FB: No timezone found, symbol may be delisted nan

consider the code similar to #233

import pandas as pd
import quantstats as qs

pv_raw = [100, 101, 99, 98, 97, 101, 99]
dates = pd.date_range("2022-12-01", periods=7)
pv = pd.DataFrame(pv_raw, index=dates, columns=["pv"])
vol = qs.stats.volatility(pv)
sharpe = qs.stats.sharpe(pv)

# calculate daily returns, no first day return as since data is daily (assumed EOD) 
rt = pv.pct_change()
rt = rt[1:]
vol2 = qs.stats.volatility(rt)
sharpe2 = qs.stats.sharpe(rt)

printouts

>>> vol[0], vol2[0]
(0.3417106479175939, 0.3742046996695246)
>>> sharpe[0], sharpe2[0]
(-0.9136086220829638, -0.9733213582901914)

volatility, sharpe and almost every other metric differs b/c when calculated from pv quantstats calculates statistics for the one extra first day as well (where returns are NaN). The results differ only a little bit, but it is quite annoying especially when trying to debug and compare results vs external stats valuation

on the other hand, if correct, trimmed returns (like rt above) are supplied to qs.reports.html(), then plots do not start at 0/100% point, which is also confusing