In stoch.py lines 47 -> 50
fastk.name = f"STOCHF_{fast_k}" fastd.name = f"STOCHF_{slow_d}" slowk.name = f"STOCH_{slow_k}" slowd.name = f"STOCH_{slow_d}"
If slow_d = slow_k (or fast_k=slow_d) slow (or fast) stochastic names become identical. in this case the returned dataframe lose a column.
Thanks a lot for your work on this project ;)

Hi Kevin,

Apologies in advance for a very stupid question. I updated the code base today. But I'm not sure what happened since then as using the 'above' function is not appending anymore.

Below line is not appending now but was working in a previous version
df.ta.above(df['MACD_12_26_9'], df['MACDs_12_26_9'], append=True)

This is appending just fine
df.ta.macd(fast=12, slow=26, signal=9, min_periods=None, append=True)

Can you take a look and tell me what I am doing wrong. Thanks in advance.

hi. When I studied the technical indicators, I found that Stochastic RSI is useful。
so，I modified the RSI code appropriately，But it's not perfect.
I want to add it to the code with your expertise。
Thanks!

calc in tradingview URL：
https://www.tradingview.com/support/solutions/43000502333-stochastic-rsi-stoch-rsi/

my code：

def stochrsi(close, length=None, scalar=None, drift=None,**kwargs):
    # Validate arguments
    length = int(length) if length and length > 0 else 14
    scalar = 100
    drift = int(drift) if drift and drift != 0 else 1

    # Calculate Result
    negative = close.diff(drift)
    positive = negative.copy()

    positive[positive < 0] = 0  # Make negatives 0 for the postive series
    negative[negative > 0] = 0  # Make postives 0 for the negative series
    
    alpha = (1.0 / length) if length > 0 else 0.5
    positive_avg = positive.ewm(alpha=alpha, adjust=False).mean()
    negative_avg = negative.ewm(alpha=alpha, adjust=False).mean().abs()

    rsi = scalar * positive_avg / (positive_avg + negative_avg)
    rsi_low   =  rsi.rolling(length).min()
    rsi_high =  rsi.rolling(length).max()

    fastk = 100 * (rsi - rsi_low) / (rsi_high-rsi_low)

    slowk = fastk.rolling(3).mean()
    slowd = slowk.rolling(3).mean()

    stochdf = pd.DataFrame(list(zip(rsi, slowk,slowd)))
    stochdf.columns=['rsi', 'stochrsi', 'stochrsi_3']

    return stochdf

Hi,

I came across your library, it's awesome ! Especially the strategy() method, which allows to add all the indicators to a dataframe. Is there a way to add all the indicators for a specific category (momentum, statistics, trend...) only ? If not it would make a great feature !

Hi,
thanks a lot for your very useful library TA.
today I pip install it,and test it.
I found 2 problems.
1, in jupyter notebook and pycharm, I can import ha, but can not import cdl_doji.
2, df.ta.strategy() does not work.I tried df.ta.strategy(high), error exist yet.
just look at the photos.

@twopirllc , I've just updated pandas-ta package to the latest version and it has updated my installed package to pandas_ta-0.1.63b.

I've run my one of the strategy, where I'm using CMO (Chande Momentum Oscillator) and fond that calculation are different than previous stable version of pandas_ta-0.1.39b0.

I didn't look into the recent commits, but checked README and observed that you have corrected CMO indicator too. Perhaps, the current change is not producing accurate results. I've validated it through some of the charting software.

Not sure, what's the correct value the current one (with pandas_ta-0.1.63b) or the previous version(pandas_ta-0.1.39b0)

It would be super convenient for end users ( especially people trying things in their jupyter environment) if the resulting dataframe, after calculating an indicator, had default plot settings already setup so that
myta_df.plot() plots a sensible graph...

What do you think ?

hi。
when i use the RSI calc，it different with Tradingview。
the tradingview use RMA calc

url：https://www.tradingview.com/support/solutions/43000502338-relative-strength-index-rsi/

Calculation
RSI = 100 – 100/ (1 + RS)
RS = Average Gain of n days UP  / Average Loss of n days DOWN
For a practical example, the built-in Pine Script function rsi(), could be replicated in long form as follows.

change = change(close)
gain = change >= 0 ? change : 0.0
loss = change < 0 ? (-1) * change : 0.0
avgGain = rma(gain, 14)
avgLoss = rma(loss, 14)
rs = avgGain / avgLoss
rsi = 100 - (100 / (1 + rs))
"rsi", above, is exactly equal to rsi(close, 14).

so, Maybe you can view or update the calc code. ^_^

I am trying to test project files under tests directory
it shows the following error:

Collecting talib
Using cached talib-0.1.1.tar.gz (1.3 kB)
Using legacy setup.py install for talib, since package 'wheel' is not installed.
Installing collected packages: talib
Running setup.py install for talib: started
Running setup.py install for talib: finished with status 'error'

ERROR: Command errored out with exit status 1:
 command: 'D:\dev\projects\pandas-ta\venv\Scripts\python.exe' -u -c 'import sys, setuptools, tokenize; sys.argv[0] = '"'"'C:\\Users\\TAS01\\AppData\\Local\\Temp\\pycharm-packaging\\talib\\setup.py'"'"'; __file__='"'"'C:\\Users\\TAS01\\AppData\\Local\\Temp\\pycharm-packaging\\talib\\setup.py'"'"';f=getattr(tokenize, '"'"'open'"'"', open)(__file__);code=f.read().replace('"'"'\r\n'"'"', '"'"'\n'"'"');f.close();exec(compile(code, __file__, '"'"'exec'"'"'))' install --record 'C:\Users\TAS01\AppData\Local\Temp\pip-record-4jfhoad9\install-record.txt' --single-version-externally-managed --compile --install-headers 'D:\dev\projects\pandas-ta\venv\include\site\python3.8\talib'
     cwd: C:\Users\TAS01\AppData\Local\Temp\pycharm-packaging\talib\
Complete output (14 lines):
running install
Traceback (most recent call last):
  File "<string>", line 1, in <module>
  File "C:\Users\TAS01\AppData\Local\Temp\pycharm-packaging\talib\setup.py", line 32, in <module>
    setup(
  File "C:\Users\TAS01\AppData\Local\Programs\Python\Python38\lib\distutils\core.py", line 148, in setup
    dist.run_commands()
  File "C:\Users\TAS01\AppData\Local\Programs\Python\Python38\lib\distutils\dist.py", line 966, in run_commands
    self.run_command(cmd)
  File "C:\Users\TAS01\AppData\Local\Programs\Python\Python38\lib\distutils\dist.py", line 985, in run_command
    cmd_obj.run()
  File "C:\Users\TAS01\AppData\Local\Temp\pycharm-packaging\talib\setup.py", line 20, in run
    raise Exception("You probably meant to install and run ta-lib")
Exception: You probably meant to install and run ta-lib
----------------------------------------

ERROR: Command errored out with exit status 1: 'D:\dev\projects\pandas-ta\venv\Scripts\python.exe' -u -c 'import sys, setuptools, tokenize; sys.argv[0] = '"'"'C:\Users\TAS01\AppData\Local\Temp\pycharm-packaging\talib\setup.py'"'"'; file='"'"'C:\Users\TAS01\AppData\Local\Temp\pycharm-packaging\talib\setup.py'"'"';f=getattr(tokenize, '"'"'open'"'"', open)(file);code=f.read().replace('"'"'\r\n'"'"', '"'"'\n'"'"');f.close();exec(compile(code, file, '"'"'exec'"'"'))' install --record 'C:\Users\TAS01\AppData\Local\Temp\pip-record-4jfhoad9\install-record.txt' --single-version-externally-managed --compile --install-headers 'D:\dev\projects\pandas-ta\venv\include\site\python3.8\talib' Check the logs for full command output.

Hi, thanks for the amazing library!

Is there a way to perform parallel processing of the technical indicators, especially if you are trying to calculate all of them, such as

df.ta.strategy(name='all')

apdict = [
        mpf.make_addplot(stck.VP,color='y',type = 'line', panel = 1)
        ]

#columns first uper like('Open,High,Low....')
#df.columns = map(str.title, df.columns)

#mpf.plot
mpf.plot(stck,style='yahoo',volume=False,type='candle',addplot=apdict,figratio=(16,8),title=ticker)

RETURNS:

Traceback (most recent call last):
  File "C:/Users/Srika/PycharmProjects/latest project/s&r.py", line 88, in <module>
    mpf.plot(stck,style='yahoo',volume=False,type='candle',addplot=apdict,figratio=(16,8),title=ticker)
  File "C:\Users\Srika\PycharmProjects\latest project\venv\lib\site-packages\mplfinance\plotting.py", line 501, in plot
    ax = _addplot_columns(panid,panels,ydata,apdict,xdates,config)
  File "C:\Users\Srika\PycharmProjects\latest project\venv\lib\site-packages\mplfinance\plotting.py", line 673, in _addplot_columns
    yd = [y for y in ydata if not math.isnan(y)]
  File "C:\Users\Srika\PycharmProjects\latest project\venv\lib\site-packages\mplfinance\plotting.py", line 673, in <listcomp>
    yd = [y for y in ydata if not math.isnan(y)]
TypeError: must be real number, not method

version: pandas-ta-0.1.65b0

$ python3.7
Python 3.7.7 (default, Apr 1 2020, 13:48:52)
[GCC 9.3.0] on linux
Type "help", "copyright", "credits" or "license" for more information.
import pandas_ta as ta
Traceback (most recent call last):
File "", line 1, in
File "/lib/python3.7/site-packages/pandas_ta/init.py", line 21, in
from pandas_ta.core import *
File "/lib/python3.7/site-packages/pandas_ta/core.py", line 8, in
from pandas_ta.candles import *
ModuleNotFoundError: No module named 'pandas_ta.candles'

Would it be possible to add the ST indicator? Definition and sample implementation is bellow.

https://www.tradingfuel.com/supertrend-indicator-formula-and-calculation/

https://www.quantopian.com/posts/error-computing-super-trend-indicator#5a204aeb9160496397ed534c

Would you kindly provide an example of how your library work guys on multindex ( row major )
i.e.

looking forward to your response

best regards

on line 28 of kama.py you have:
result = [npNaN for _ in range(0, length - 1)] + [0]

I think it's supposed to be:
result = [np.NaN for _ in range(0, length - 1)] + [0]

I am running python 3.5

here is an error I got

File "C:/StockServer/TWSAPI/SuperTrendTest3.py", line 3, in
import pandas_ta as ta

File "C:\Users\googlecloud\Anaconda3\envs\py35\lib\site-packages\pandas_ta_init_.py", line 22, in
from .momentum.ao import ao

File "C:\Users\googlecloud\Anaconda3\envs\py35\lib\site-packages\pandas_ta\momentum\ao.py", line 33
ao.name = f"AO_{fast}_{slow}"
^
SyntaxError: invalid syntax

Currently the charts when displayed, start late, meaning ema15 starts after 15dats, etc..

typically we need full charts, where by we need to load previous data and display from the date we needed..

How can we get this thing done

when i run this, i get error

import pandas as pd
import pandas_ta as ta

# Help about this, 'ta', extension
help(pd.DataFrame().ta)

# List of all indicators
pd.DataFrame().ta.indicators()

# Help about the log_return indicator
help(ta.log_return)

# Help about the log_return indicator as a DataFrame Extension
help(pd.DataFrame().ta.log_return)

File "C:\Users\Rakesh\AppData\Local\Continuum\anaconda3\envs\zip355\lib\site-packages\pandas_ta\momentum\ao.py", line 33
ao.name = f"AO_{fast}_{slow}"
^
SyntaxError: invalid syntax

I am running the following
(zip355) C:\Users\Rakesh>pip install -U git+https://github.com/twopirllc/pandas-ta
Collecting git+https://github.com/twopirllc/pandas-ta
Cloning https://github.com/twopirllc/pandas-ta to c:\users\Rakesh\appdata\local\temp\pip-req-build-1l4j_eh8
Requirement not upgraded as not directly required: pandas in c:\users\Rakesh\appdata\local\continuum\anaconda3\envs\zip355\lib\site-packages (from pandas-ta==0.1.38b0) (0.22.0)
Requirement not upgraded as not directly required: python-dateutil>=2 in c:\users\Rakesh\appdata\local\continuum\anaconda3\envs\zip355\lib\site-packages (from pandas->pandas-ta==0.1.38b0) (2.8.1)
Requirement not upgraded as not directly required: pytz>=2011k in c:\users\Rakesh\appdata\local\continuum\anaconda3\envs\zip355\lib\site-packages (from pandas->pandas-ta==0.1.38b0) (2019.3)
Requirement not upgraded as not directly required: numpy>=1.9.0 in c:\users\Rakesh\appdata\local\continuum\anaconda3\envs\zip355\lib\site-packages (from pandas->pandas-ta==0.1.38b0) (1.14.2)
Requirement not upgraded as not directly required: six>=1.5 in c:\users\Rakesh\appdata\local\continuum\anaconda3\envs\zip355\lib\site-packages (from python-dateutil>=2->pandas->pandas-ta==0.1.38b0) (1.11.0)
Building wheels for collected packages: pandas-ta
Running setup.py bdist_wheel for pandas-ta ... done
Stored in directory: C:\Users\Rakesh\AppData\Local\Temp\pip-ephem-wheel-cache-bdcsthu0\wheels\64\67\96\15e918c3b53b4a323b5bd037c7f08be5ef6908141c50f07c76
Successfully built pandas-ta
jupyterlab-server 1.0.0 has requirement jsonschema>=3.0.1, but you'll have jsonschema 2.6.0 which is incompatible.
Installing collected packages: pandas-ta
Successfully installed pandas-ta-0.1.38b0

Hi,

I am discovering pandas_ta, and I was wondering why 'append' is an option in the API, but not join ?
It seems to me, naively, that join would be a useful option, especially to link an indicator with the time of the current dataframe...

hi mate,
i try to use your indicator but i have a problem when i use the parameter lenght:
sma=df.ta.sma(close=df['close'],lenght=10) print(df['close'].tail(3)) print(sma.tail(3))

and the output:

3922 8079.54
3923 8073.16
3924 8076.36
Name: close, dtype: float64
3922 8064.433
3923 8062.750
3924 8061.351
Name: SMA_10, dtype: float64

and if i put
sma=df.ta.sma(close=df['close'],lenght=20)

the output is
3922 8079.54
3923 8073.16
3924 8076.36
Name: close, dtype: float64
3922 8064.433
3923 8062.750
3924 8061.351
Name: SMA_10, dtype: float64

so not change anything..
where is my mistake ? because every time use the else condition

Validate Arguments

close = verify_series(close)
length = int(length) if length and length > 0 else 10
offset = get_offset(offset)

Hi , is it possible for you to add relative strength? The only thing i found was relative strength and not relative strength. Here is a few links about relative strength.

[Relative Strength]
https://www.investopedia.com/terms/r/relativestrength.asp

[Relative Strength vs Relative Strength Index]
https://advisoranalyst.com/2016/11/01/relative-strength-vs-relative-strength-index-whats-the-difference.html/

When running locally, this runs perfect. But when publishing to Azure Serverless (App function), I always get "ModuleNotFoundError: No module named 'pandas'". However the pandas is installed before.

The example notebook shows

e.ta.indicators()

but I think this needs to be

e().ta.indicators()

Or you need to assign e = pd.DataFrame() ... coders choice. :)

Hello,

First, thank you for this awesome package! It's nicely done!

I have the following issue - The data is in reverse order on my end, i.e. most recent candle is index 0.

For example:

>>> data.SPY
           date    open    high     low   close       volume
0    2020-05-06  288.04  288.45  283.87  284.34   67758728.0
1    2020-05-05  286.64  289.25  283.71  286.19   79330400.0
2    2020-05-04  280.74  283.90  279.13  283.57   80873200.0
3    2020-05-01  285.31  290.66  281.52  282.79  125063900.0
4    2020-04-30  291.71  293.32  288.59  290.48  122901700.0
...         ...     ...     ...     ...     ...          ...
6862 1993-02-04   44.97   45.09   44.47   45.00     531500.0
6863 1993-02-03   44.41   44.84   44.38   44.81     529400.0
6864 1993-02-02   44.22   44.38   44.13   44.34     201300.0
6865 1993-02-01   43.97   44.25   43.97   44.25     480500.0
6866 1993-01-29   43.97   43.97   43.75   43.94    1003200.0

[6867 rows x 6 columns]

And e.g. when I do atr, it seems that it goes in reverse order, filling most recent with NaN:

>>> data.SPY.ta.atr()
0            NaN
1            NaN
2            NaN
3            NaN
4            NaN
          ...   
6862    0.556601
6863    0.565054
6864    0.580380
6865    0.552330
6866    0.545352
Name: ATR_14, Length: 6867, dtype: float64

Now I know that the easiest thing in the world is to reverse the data with [::-1], but still... would be great if we could specify the order somehow, or even better - detect it form the date field.

What's the best way to create all relevant features? I tried to use the df.ta.strategy(name='all')

Error:

`Index(['dt', 'open', 'high', 'low', 'close', 'volumn', 'trading_dt', 'stock'], dtype='object')

KeyError                                  Traceback (most recent call last)
D:\Anaconda\envs\py36\lib\site-packages\pandas\core\indexes\base.py in get_loc(self, key, method, tolerance)
   2645             try:
-> 2646                 return self._engine.get_loc(key)
   2647             except KeyError:

pandas\_libs\index.pyx in pandas._libs.index.IndexEngine.get_loc()

pandas\_libs\index.pyx in pandas._libs.index.IndexEngine.get_loc()

pandas\_libs\hashtable_class_helper.pxi in pandas._libs.hashtable.PyObjectHashTable.get_item()

pandas\_libs\hashtable_class_helper.pxi in pandas._libs.hashtable.PyObjectHashTable.get_item()

KeyError: 'high'

During handling of the above exception, another exception occurred:

KeyError                                  Traceback (most recent call last)
<ipython-input-98-ce3fa983de61> in <module>
----> 1 ch_3 = ta_pd_v2(ch)

<ipython-input-77-d2409a027e53> in ta_pd_v2(df)
      8     df1["close-1"] = df1["close"].shift(1)
      9 
---> 10     df.ta.strategy(name='all')
     11     return df1

D:\Anaconda\envs\py36\lib\site-packages\pandas_ta\core.py in strategy(self, **kwargs)
    367         if name is None or name == "" or not isinstance(name, str): # Extra check
    368             name = "all"
--> 369         self._all(**kwargs) if name == "all" else None
    370 
    371 

D:\Anaconda\envs\py36\lib\site-packages\pandas_ta\core.py in _all(self, **kwargs)
    343         for kind in indicators:
    344             fn = getattr(self, kind)
--> 345             fn(append=append, **kwargs)
    346             print(f"[+] {kind}") if verbose else None
    347 

D:\Anaconda\envs\py36\lib\site-packages\pandas_ta\core.py in _wrapper(*class_methods, **method_kwargs)
     22     def _wrapper(*class_methods, **method_kwargs):
     23         cm = class_methods[0]
---> 24         result = method(cm, **method_kwargs)
     25 
     26         cm._add_prefix_suffix(result, **method_kwargs)

D:\Anaconda\envs\py36\lib\site-packages\pandas_ta\core.py in aberration(self, high, low, close, length, atr_length, offset, **kwargs)
   1067     @finalize
   1068     def aberration(self, high=None, low=None, close=None, length=None, atr_length=None, offset=None, **kwargs):
-> 1069         high = self._get_column(high, 'high')
   1070         low = self._get_column(low, 'low')
   1071         close = self._get_column(close, 'close')

D:\Anaconda\envs\py36\lib\site-packages\pandas_ta\core.py in _get_column(self, series, default)
    229         # Apply default if no series nor a default.
    230         elif series is None or default is None:
--> 231             return df[self.adjusted] if self.adjusted is not None else df[default]
    232         # Ok.  So it's a str.
    233         elif isinstance(series, str):

D:\Anaconda\envs\py36\lib\site-packages\pandas\core\frame.py in __getitem__(self, key)
   2798             if self.columns.nlevels > 1:
   2799                 return self._getitem_multilevel(key)
-> 2800             indexer = self.columns.get_loc(key)
   2801             if is_integer(indexer):
   2802                 indexer = [indexer]

D:\Anaconda\envs\py36\lib\site-packages\pandas\core\indexes\base.py in get_loc(self, key, method, tolerance)
   2646                 return self._engine.get_loc(key)
   2647             except KeyError:
-> 2648                 return self._engine.get_loc(self._maybe_cast_indexer(key))
   2649         indexer = self.get_indexer([key], method=method, tolerance=tolerance)
   2650         if indexer.ndim > 1 or indexer.size > 1:

pandas\_libs\index.pyx in pandas._libs.index.IndexEngine.get_loc()

pandas\_libs\index.pyx in pandas._libs.index.IndexEngine.get_loc()

pandas\_libs\hashtable_class_helper.pxi in pandas._libs.hashtable.PyObjectHashTable.get_item()

pandas\_libs\hashtable_class_helper.pxi in pandas._libs.hashtable.PyObjectHashTable.get_item()

KeyError: 'high'`

So, I used the customised feature generation script but run into error for certain features:

`def ta_pd_v2(df):
    df1 = df.copy()
    df1.columns = map(str.lower, df1.columns) 
    df1["open-1"] = df1["open"].shift(1)
    df1["high-1"] = df1["high"].shift(1)
    df1["low-1"] = df1["low"].shift(1)
    df1["close-1"] = df1["close"].shift(1)
    
    df1 = df1.ta.strategy(name='all')
    return df1

Example of error:

---------------------------------------------------------------------------
KeyError                                  Traceback (most recent call last)
D:\Anaconda\envs\py36\lib\site-packages\pandas\core\indexes\base.py in get_loc(self, key, method, tolerance)
   2645             try:
-> 2646                 return self._engine.get_loc(key)
   2647             except KeyError:

pandas\_libs\index.pyx in pandas._libs.index.IndexEngine.get_loc()

pandas\_libs\index.pyx in pandas._libs.index.IndexEngine.get_loc()

pandas\_libs\hashtable_class_helper.pxi in pandas._libs.hashtable.PyObjectHashTable.get_item()

pandas\_libs\hashtable_class_helper.pxi in pandas._libs.hashtable.PyObjectHashTable.get_item()

KeyError: 'volume'

During handling of the above exception, another exception occurred:

KeyError                                  Traceback (most recent call last)
<ipython-input-28-5d947df895f7> in <module>
----> 1 ch.ta.strategy(name='all')

D:\Anaconda\envs\py36\lib\site-packages\pandas_ta\core.py in strategy(self, **kwargs)
    367         if name is None or name == "" or not isinstance(name, str): # Extra check
    368             name = "all"
--> 369         self._all(**kwargs) if name == "all" else None
    370 
    371 

D:\Anaconda\envs\py36\lib\site-packages\pandas_ta\core.py in _all(self, **kwargs)
    343         for kind in indicators:
    344             fn = getattr(self, kind)
--> 345             fn(append=append, **kwargs)
    346             print(f"[+] {kind}") if verbose else None
    347 

D:\Anaconda\envs\py36\lib\site-packages\pandas_ta\core.py in _wrapper(*class_methods, **method_kwargs)
     22     def _wrapper(*class_methods, **method_kwargs):
     23         cm = class_methods[0]
---> 24         result = method(cm, **method_kwargs)
     25 
     26         cm._add_prefix_suffix(result, **method_kwargs)

D:\Anaconda\envs\py36\lib\site-packages\pandas_ta\core.py in ad(self, high, low, close, volume, open_, signed, offset, **kwargs)
   1161         low = self._get_column(low, 'low')
   1162         close = self._get_column(close, 'close')
-> 1163         volume = self._get_column(volume, 'volume')
   1164 
   1165         result = ad(high=high, low=low, close=close, volume=volume, open_=open_, signed=signed, offset=offset, **kwargs)

D:\Anaconda\envs\py36\lib\site-packages\pandas_ta\core.py in _get_column(self, series, default)
    229         # Apply default if no series nor a default.
    230         elif series is None or default is None:
--> 231             return df[self.adjusted] if self.adjusted is not None else df[default]
    232         # Ok.  So it's a str.
    233         elif isinstance(series, str):

D:\Anaconda\envs\py36\lib\site-packages\pandas\core\frame.py in __getitem__(self, key)
   2798             if self.columns.nlevels > 1:
   2799                 return self._getitem_multilevel(key)
-> 2800             indexer = self.columns.get_loc(key)
   2801             if is_integer(indexer):
   2802                 indexer = [indexer]

D:\Anaconda\envs\py36\lib\site-packages\pandas\core\indexes\base.py in get_loc(self, key, method, tolerance)
   2646                 return self._engine.get_loc(key)
   2647             except KeyError:
-> 2648                 return self._engine.get_loc(self._maybe_cast_indexer(key))
   2649         indexer = self.get_indexer([key], method=method, tolerance=tolerance)
   2650         if indexer.ndim > 1 or indexer.size > 1:

pandas\_libs\index.pyx in pandas._libs.index.IndexEngine.get_loc()

pandas\_libs\index.pyx in pandas._libs.index.IndexEngine.get_loc()

pandas\_libs\hashtable_class_helper.pxi in pandas._libs.hashtable.PyObjectHashTable.get_item()

pandas\_libs\hashtable_class_helper.pxi in pandas._libs.hashtable.PyObjectHashTable.get_item()

KeyError: 'volume'

My another attempt:

`
import traceback
def ta_pd(df):
    df1 = df.copy()
    indicators = pd.DataFrame().ta.indicators(as_list=True)
    append = True
    df1.columns = map(str.lower, df1.columns)
    
    df1["open-1"] = df1["open"].shift(1)
    df1["high-1"] = df1["high"].shift(1)
    df1["low-1"] = df1["low"].shift(1)
    df1["close-1"] = df1["close"].shift(1)
    
    #print(len(indicators), "will be added.")
    for kind in indicators:
        try:
            print(kind)
            df1.ta(kind=kind, append=append)
        except Exception as e:
            print("\nError with indicator:", kind)
            print("") # print(".", end="")

    indicators2 = ["log_return", "percent_return", "trend_return"]
    for kind in indicators2:
        try:
            print(kind)
            df1.ta(kind=kind, append=append, cumulative=True)
        except Exception as e:
            print("\nError with indicator:", kind)
            print("") # print(".", end="")
    
    df1.drop(columns = ['dt', 'open', 'high', 'low', 'close', 'volumn', 'trading_dt', 'stock', 'open-1', 'high-1', 'low-1', 'close-1',
                           
    return df1
` 

and the error:

ad

Error with indicator: ad

adosc

Error with indicator: adosc

I recently stumbled upon your project (literally about 2 hours ago) and I've only had a chance to go through a few of your tools but I noticed a few of them that are ripe for vectorization. For example, looking at the WMA implementation you have (I've removed some fluff to conserve space since this is already a long post):

from pandas import Series
from numpy import arange

def wma(close, length):
    total_weight = 0.5 * length * (length + 1)
    weights_ = Series(arange(1, length + 1))
    def linear(w):
        def _compute(x):
            return (w * x).sum() / total_weight
        return _compute
    close_ = close.rolling(length, min_periods=length)
    wma = close_.apply(linear(weights), raw=True)
    return wma

Using convolution and Numpy's ability to operate on an entire array at the same time we can get a massive speedup over you method, albeit at the expense of ease of understanding. Here's what I came up with:

from numpy import arange
from scipy.ndimage import convolve1d as conv

def wma(close, length):
    weights = arange(1, length) / arange(1, length).sum()
    return conv(close, weights=weights, axis=0, mode='nearest')

Just as a quick test I pulled in the last 1000 5 minute time-frame tick marks from an exchange, then created a length 10, 100, and 1000 DataFrame on the closes. Here is the code:

import numpy as np, timeit
setup = '''
import pandas as pd
from numpy import arange
from pandas import Series
from scipy.ndimage import convolve1d as conv
def my_wma(dr, length):
    wts = arange(1, length) / arange(1, length).sum()
    return conv(dr, weights=wts, axis=0, mode='nearest')
def your_wma(close, length):
    total_weight = 0.5 * length * (length + 1)
    weights = Series(arange(1, length + 1))
    def linear(w):
        def _compute(x):
            return (w * x).sum() / total_weight
        return _compute
    close_ = close.rolling(length, min_periods=length)
    wma = close_.apply(linear(weights), raw=True)
    return wma
    
df = READ IN FROM CSV
df1 = df['close'].tail(10)
df2 = df['close'].tail(100)
df3 = df['close'].tail(1000)
'''
myt1 = timeit.repeat('my_wma(df1.values, 9)', setup=setup, number=100, repeat=10)
myt2 = timeit.repeat('my_wma(df2.values, 9)', setup=setup, number=100, repeat=10)
myt3 = timeit.repeat('my_wma(df3.values, 9)', setup=setup, number=100, repeat=10)
yourt1 = timeit.repeat('your_wma(df1, 9)', setup=setup, number=100, repeat=10)
yourt2 = timeit.repeat('your_wma(df2, 9)', setup=setup, number=100, repeat=10)
yourt3 = timeit.repeat('your_wma(df3, 9)', setup=setup, number=100, repeat=10)
print('Average Speed difference: {:0.01f}x'.format(np.mean(np.divide(yourt1, myt1))))
print('Average Speed difference: {:0.01f}x'.format(np.mean(np.divide(yourt2, myt2))))
print('Average Speed difference: {:0.01f}x'.format(np.mean(np.divide(yourt3, myt3))))

and the result was...

Average Speed difference: 29.2x
Average Speed difference: 613.0x
Average Speed difference: 5258.9x

I also ran

%timeit my_wma(df1.values, 9)
%timeit my_wma(df2.values, 9)
%timeit my_wma(df3.values, 9)
%timeit your_wma(df1, 9)
%timeit your_wma(df2, 9)
%timeit your_wma(df3, 9)

And got the results:

22.1 µs ± 58 ns per loop (mean ± std. dev. of 7 runs, 10000 loops each)
22.7 µs ± 177 ns per loop (mean ± std. dev. of 7 runs, 10000 loops each)
27.9 µs ± 42.5 ns per loop (mean ± std. dev. of 7 runs, 10000 loops each)

664 µs ± 3.02 µs per loop (mean ± std. dev. of 7 runs, 1000 loops each)
14.2 ms ± 20.4 µs per loop (mean ± std. dev. of 7 runs, 100 loops each)
150 ms ± 575 µs per loop (mean ± std. dev. of 7 runs, 10 loops each)

I realize that this is speedup would really be beneficial to someone who is calculating the averages for all historical points and that real time trading that the speed would probably not make a huge impact. However, it's clear to me that this method is much more efficient and IMO every bit of speedup helps, especially if one is back-testing.

The indicator in its current form does not return the signal line. Could this be updated to return a DataFrame with the FT and FT signal values?

https://library.tradingtechnologies.com/trade/chrt-ti-ehler-fisher-transformation.html

Hi, I think the Donchian Channels indicator (DC, a volatility indicator) isn't computing the right thing : the computation should be on the highest highs and the lowest lows, instead of on the closes. So the code therefore becomes :

Definition :

def donchian(high, low, lower_length=None, upper_length=None, offset=None, **kwargs):

line 15 :

# Calculate Result
lower = low.rolling(lower_length, min_periods=lower_min_periods).min()
upper = high.rolling(upper_length, min_periods=upper_min_periods).max()
mid = 0.5 * (lower + upper)

Hi,

I have been using your library to understand TA. First off, thank you for the amazing work. It has been really useful to have!

Wherever possible, I have been double checking the indicator values calculated either on Yahoo Finance (YF) or StockCharts (SC). As such, I wanted to ask for some clarifications on couple of things:

1. Using df.ta.rsi(length=14) gives me RSI results that are little bit off from YF or SC or manual calculation in excel. However, with df.ta.rsi(length=13), results pretty match with YF (Period 14) or manual work (again using 14 periods). I am not sure why there is a mismatch with ta.rsi default value of 14. Can you please check?

2. In df.ta.stoch(), I noticed calling with default values gave different results from YF/SC. Digging deeper, in stoch.py, lines 18 & 19, I'm thinking shouldn't the value used be fast_k instead of slow_k. If I update the below to fast_k, I do get matching results with both sources mentioned. Maybe you had something else in mind while implementing?

Calculate Result

lowest_low   =  low.rolling(slow_k).min()
highest_high = high.rolling(slow_k).max()

3. Using df.ta.adx(), I do get the ADX_14 value that matches with YF/SC but DMP_14 & DMN_14 are not matching. I was trying to understand your methodology of ADX calculation. I think the difference is due to smoothing techniques used. In adx.py, atr & rma calculations use ewm techniques. SC uses a different way of smoothing TR, +DM & -DM. For example:

First TR14 = Sum of first 14 periods of TR1
Second TR14 = First TR14 - (First TR14/14) + Current TR1
Subsequent Values = Prior TR14 - (Prior TR14/14) + Current TR1

as explained here:
https://school.stockcharts.com/doku.php?id=technical_indicators:average_directional_index_adx

I'm not sure what the above smoothing methodology is called but can you please comment on the difference? Also, is there way to use your existing functions to implement this smoothing method?

Sorry for the long post. I greatly appreciate your work and am trying to understand as much as possible. Any further help is greatly appreciated too!

The pip install works perfectly fine but the import pandas_ta as ta on the jupyter notebook version of ibm watson studio has this error below

AttributeError Traceback (most recent call last)
in
5 import numpy as np
6 import pandas as pd
----> 7 import pandas_ta as ta
8 import pandas_datareader.data as web
9 import datetime

C:\ProgramData\WatsonStudioDesktop\miniconda3\envs\desktop\lib\site-packages\pandas_ta_init_.py in
122
123 # DataFrame Extension
--> 124 from .core import *
125

C:\ProgramData\WatsonStudioDesktop\miniconda3\envs\desktop\lib\site-packages\pandas_ta\core.py in
25
26
---> 27 @pd.api.extensions.register_dataframe_accessor('ta')
28 class AnalysisIndicators(BasePandasObject):
29 """AnalysisIndicators is class that extends the Pandas DataFrame via

AttributeError: module 'pandas.api' has no attribute 'extensions'

Hi, can you also add force index?

https://www.investopedia.com/terms/f/force-index.asp#:~:text=The%20force%20index%20is%20a,book%20Trading%20for%20a%20Living.

https://www.daytrading.com/force-index

https://tlc.thinkorswim.com/center/reference/Tech-Indicators/studies-library/E-F/ForceIndex#:~:text=The%20Force%20Index%20study%20combines,negative%20values%20indicate%20the%20downtrend.

Originally posted by @SoftDevDanial in #88 (comment)

I've tried the ADX indicator like below
adx = ta.adx(df[ticker]['high'], df[ticker]['low'], df[ticker]['close'], length=14)
and returns close but not exactly the same results as trading view. The difference is about 1% up/down.

Hi all , Hi Kevin

I have run in to the below error while applying a window calc to one of the features implemented within the lib

it goes as
df['roll_entp_10']= df_test.ENTP_10.rolling(10) which is a series , after explicitly turning in to a series still getting the same error

just for reference

pd.version 1.0.3

I have looked though in to the following as suggested
pandas-dev/pandas#11704
is there any additional guidance you can provide on that issue ?
best regards

Hi KJ,

Can you look into kst.py and maybe update the naming for kst_signal? Reason is that if I call KST multiple times (short, medium and long term parameters), KSTS_9 gets overridden. Its not a big deal to handle but wondering if this was intentional or not on your part.
For example, following returns only one KSTS_9
df.ta.kst(append=True) #short-term
df.ta.kst(roc1=10, roc2=13, roc3=15, roc4=20, sma1=10, sma2=13, sma3=15, sma4=20, signal=9, append=True) #Medium term weekly
df.ta.kst(roc1=9, roc2=12, roc3=18, roc4=24, sma1=6, sma2=6, sma3=6, sma4=9, signal=9, append=True) #Long term monthly

Another enhancement request is adding a signal logic to TRIX .py (similar to KST)?

Thanks again for all of your work. This package is amazing!

Hello my friend

Thank you very much for creating such a very convenient library.

I often use some indicators
KDJ, Brar and Aberration(Keltner channel) related libraries.

I implemented its algorithm using pandas.

so, I hope can be add to Panda-TA,
can you help me? thank you!

BRAR

    #BRAR

    df.columns = ['time','open','high','low','close','volume']
    df['HO']=df['high']-df['open']
    df['OL']=df['open']-df['low']
    df['HCY']=df['high']-df['close'].shift(1)
    df['CYL']=df['close'].shift(1)-df['low']
    df.loc[df['HCY']<0,'HCY']=0
    df.loc[df['CYL']<0,'CYL']=0
    df['ar']=df['HO'].rolling(26).sum()/df['OL'].rolling(26).sum()*100
    df['br']=df['HCY'].rolling(26).sum()/df['CYL'].rolling(26).sum()*100

BIAS

    df.columns = ['time','open','high','low','close','volume']
    result_MA=df.ta(kind='sma',length=MA_lengh)
    C=df['close']
    BIAS=(C-result_MA)/result_MA

KDJ

    #KDJ
    N=44
    M1=9
    H6['llv_low']=H6['low'].rolling(N).min()
    H6['hhv_high']=H6['high'].rolling(N).max()
    H6['rsv']=(H6['close']-H6['llv_low'])/(H6['hhv_high']-H6['llv_low'])
    H6['k']=H6['rsv'].ewm(adjust=False,alpha=1/M1).mean()
    H6['d']=H6['k'].ewm(adjust=False,alpha=1/M1).mean()
    H6['j']=3*H6['k']-2*H6['d']

Aberration

def Aberration(RAW,MA_lengh,ATR_lengh):
    df = pd.DataFrame(RAW)
    df.columns = ['time','open','high','low','close','volume']
    JG=(df['high']+df['low']+df['close'])/3
    ZG=df.ta(kind='sma',close=JG,length=MA_lengh)
    ATR=df.ta(kind='ATR',length=ATR_lengh)
    SG=ZG+ATR
    XG=ZG-ATR
   
    result = pd.concat([ZG,SG,XG,ATR], axis=1)

More Notebooks with different strategies and visualizations (matplotlib, bokeh, mplfinance)

@twopirllc Thanks for creating this wonderful python module. I'm extensively using this module for my algos.

I found an issue with VWAP indicator when I ran it with my backtesting data. As per definition, VWAP should be calculated on daily data(Intraday).

Since, we pass series data, which contains past dates data as well. It calculates the cumulative sum incorrectly in that case. Each day's opening volume and hlc price will be different for sure.

Thus, the calculation for vwap should start with fresh data of each day. eg cumsum()

Note: Calculation is absolutely correct in a case of series data contains only 1-day data.

Maybe we could give a try to group a series data according to date and performing a calculation. It's just my thought. But, I would be happy to hear from you. As it's important for cross-checking strategies with backtesting data.

Hi Kevin,

While experimenting with the many indicators that you have implemented (big thanks for that!), I noticed a mistake in the calculation for the True Range.
I guess pandas_ta/volatility/true_range.py line 61 should be
ranges = [high - low, high - prev_close, prev_close - low]
instead of
ranges = [high - low, high - prev_close, low - prev_close]

Regards,
Wout

Hi, I'm trying to calculate indicators for Live tick data,

You have done a great work!, its very much helpful and makes life easy ;)

Thanks & Regards,
Vignesh Murugan

Hi,

Using your (amazing) library, I have found that CMF is producing a lot of NaNs on my data. Checking which was the reason I have figured out that in my data there are a lot of cases where 'high' and 'low' price are exactly the same, hence it produces NaNs due to this part of the code:

Hence my proposal is to add an epsilon component
from sys import float_info as sflt

ad *= volume / (hl_range + sflt.epsilon)

What is your opinion on that?

Thank you for all your work,
Lluis

BUG: NOT ENOUGH WATER TO PUT OUT THE FIRE HERE

Hi @twopirllc,

I'm playing with the new improved strategy() method, it's quite an improvement nice work ! You should push it to pip ;)

In my workflow, I'm computing a bunch of data on the indicators and then select the relevant ones. However, I don't know a priori which ones nor their parameters, and they are not always the same.

Since all the indictors don't have the same number of parameters, they would be passed as a list of floats, instead of their explicit names. For example :

CustomStrategy = ta.Strategy(name="MyStrat",
	                     ta=[{'kind':'sma',    'params':[30]}, 
                                 {'kind':'bbands', 'params':[20,2]}},
                                 {'kind':'macd',   'params':[8,21]}, 
                                ],
	                     )

This way, the code would be more modular. Is it possible to implement that ?

let's say i have 100 rows in data frame, applied all strategies. Now, i want to add 1 more row, i don't want it to recalculate all the whole 101 rows instead only the last one.

I see on Tradingview script calc SWMA very simple, how to convert it to python use pandas ? : https://www.tradingview.com/script/6MWFvnPO-Sine-Weighted-Moving-Average/

PI = 2 * asin(1)
sum = 0.0
weightSum = 0.0

for i = 0 to length - 1
weight = sin((i + 1) * PI / (length + 1))
sum := sum + nz(src[i]) * weight
weightSum := weightSum + weight

swma = sum / weightSum

_Originally posted by @cbogithub in https://github.com/twopirllc/pandas-ta/issues/22#issuecomment-508996474_

@twopirllc Thanks for creating this wonderful python module.
I'm extensively using this module for my algos.

when i use the ATR func,the return different with tradingvew.
it mabe use rma to calc.
so the ATR=RMA(TA,lengh)

can you update it? Thanks !

Hello,

Thank you for your great work.

Regarding the Accumulation/Distribution indicator, it's possible to have two versions of it: using close and open OR using close, high and low. However, if I'm not mistaken, as they would have the same name, it is impossible to have the two versions inside the same dataframe (without renaming).

Could you please dynamically name the series regarding the version chosen?

Thanks

Thank you for this great library.

I've used the strategy method to add all indicators to a data frame. DPO_1 got a massive feature importance score and any models trained with that indicator had accuracy of 99 to 100% 😄.

It seems that when center=True (the default), this line leaks future data into the past.

Open to documentation ideas.