Candlestick Patterns Using Python

In this tutorial, I will build a candlestick chart pattern scanner using Python and pandas ta lib.

I have the stocks data stored in mongoDB. It is very easy to retrieve and insert data in mongo.

Let us first import the necessary packages for pandas, talib and pymongo. Make sure you have numpy, pandas_ta, talib,pymongo and mongoDB installed.

In [1]:

import pymongo
from pymongo import MongoClient
import datetime
import pandas as pd
from talib.abstract import *
import pandas_ta as ta

Let us initiate the mongo DB Connection.

In [2]:

client = MongoClient('mongodb://149.56.26.138:27017')
db = client['stocksdata']

Let us look at our stock document. Document is similar to a row in MySQL.

In [3]:

db.eod_stock_data.find_one()

Out[3]:

{'_id': ObjectId('6202f99a80d4ed2d9c536156'),
 'date': datetime.datetime(2020, 1, 2, 0, 0),
 'open': 21.86,
 'high': 21.86,
 'low': 21.315,
 'close': 21.42,
 'adjusted_close': 21.3733,
 'volume': 3062541,
 'ticker': 'AA',
 'perchange': None}

For this tutorial, let us try to find the candlestick patterns for APPLE (ticker aapl) stock.

Below command will extract the data from mongo for the last 300 days.

In [4]:

db.eod_stock_data.find({'ticker':'AAPL'}).sort([('date',pymongo.DESCENDING)]).limit(300)

Out[4]:

<pymongo.cursor.Cursor at 0x7fd02223bfa0>

Note above command will return a mongo Cursor, we will have iterate over it to see the data for each day.

Let us store this data in a variable.

In [39]:

docs = db.eod_stock_data.find({'ticker':'AAPL'}).sort([('date',pymongo.DESCENDING)]).limit(300)

In [40]:

for doc in docs:
    print(doc)
    break

{'_id': ObjectId('6378006ca528ace221e1a5ac'), 'date': datetime.datetime(2022, 11, 18, 0, 0), 'open': 152.305, 'high': 152.57, 'low': 149.97, 'close': 151.29, 'adjusted_close': 151.29, 'volume': 70799117, 'ticker': 'AAPL', 'perchange': 0.38}

Let us convert the above data in to pandas Dataframe. since pandas_ta requires pandas dataframe as an input.

In [44]:

docs = db.eod_stock_data.find({'ticker':'AAPL'}).sort([('date',pymongo.DESCENDING)]).limit(300)

In [45]:

df = pd.DataFrame(docs)

Let us check the first row in our data.

In [46]:

df.head(1)

Out[46]:

	_id	date	open	high	low	close	adjusted_close	volume	ticker	perchange
0	6378006ca528ace221e1a5ac	2022-11-18	152.305	152.57	149.97	151.29	151.29	70799117	AAPL	0.38

the first row in our dataframe shows the data for latest date. However the pandas_ta require the data in ascending order. Let us change the order of our data in ascending form.

In [47]:

df = df[::-1]

Also we need to change the "date" name to "datetime".

In [48]:

df.rename(columns={'date':'datetime'},inplace=True)

In [49]:

df.head(1)

Out[49]:

	_id	datetime	open	high	low	close	adjusted_close	volume	ticker	perchange
299	6202f9d580d4ed2d9c5379f9	2021-09-10	155.0	155.48	148.7	148.97	148.5636	140893203	AAPL	-3.31

Let us re-print our first row.

In [50]:

df.head(1)

Out[50]:

	_id	datetime	open	high	low	close	adjusted_close	volume	ticker	perchange
299	6202f9d580d4ed2d9c5379f9	2021-09-10	155.0	155.48	148.7	148.97	148.5636	140893203	AAPL	-3.31

We would need to set the datetime as index which is also required by pandas_ta.

In [51]:

df.set_index(pd.DatetimeIndex(df["datetime"]), inplace=True)

In [52]:

df.head(1)

Out[52]:

	_id	datetime	open	high	low	close	adjusted_close	volume	ticker	perchange
datetime
2021-09-10	6202f9d580d4ed2d9c5379f9	2021-09-10	155.0	155.48	148.7	148.97	148.5636	140893203	AAPL	-3.31

pandas_ta only requires datetime, open, high, low, close and volume. We can remove the other columns.

In [53]:

df.drop(columns=['datetime','adjusted_close','perchange','_id','ticker'],inplace=True)

In [54]:

df.head(1)

Out[54]:

	open	high	low	close	volume
datetime
2021-09-10	155.0	155.48	148.7	148.97	140893203

Ok, now we have our dataframe in correct (pandas_ta) format. Let us run df.ta.strategy('All') step to calculate everything.
Note you can run it selectively too for specific patterns, for that checkout the pandas_ta documentation page.

In [55]:

df.ta.strategy("All")

0it [00:00, ?it/s]/home/anaconda3/envs/condapy38/lib/python3.8/multiprocessing/pool.py:48: FutureWarning: The series.append method is deprecated and will be removed from pandas in a future version. Use pandas.concat instead.
  return list(map(*args))
131it [00:00, 477.00it/s]

The above step will add columns for the candlestick patterns. Let us check our dataframe. Last row will contain the data for the most recent date.

In [56]:

df.tail(1)

Out[56]:

	open	high	low	close	volume	ABER_ZG_5_15	ABER_SG_5_15	ABER_XG_5_15	ABER_ATR_5_15	ACCBL_20	...	VIDYA_14	VTXP_14	VTXM_14	VWAP_D	VWMA_10	WCP	WILLR_14	WMA_10	ZL_EMA_10	ZS_30
datetime
2022-11-18	152.305	152.57	149.97	151.29	70799117	149.762153	154.894866	144.62944	5.132713	134.810048	...	149.822045	0.924025	0.938888	151.276667	145.953573	151.28	-19.743711	148.330364	150.943228	1.123647

1 rows × 283 columns

The above row contains 283 columns. Let us print the name of these columns.

In [57]:

df.columns

Out[57]:

Index(['open', 'high', 'low', 'close', 'volume', 'ABER_ZG_5_15',
       'ABER_SG_5_15', 'ABER_XG_5_15', 'ABER_ATR_5_15', 'ACCBL_20',
       ...
       'VIDYA_14', 'VTXP_14', 'VTXM_14', 'VWAP_D', 'VWMA_10', 'WCP',
       'WILLR_14', 'WMA_10', 'ZL_EMA_10', 'ZS_30'],
      dtype='object', length=283)

Let us print the value of some of these candlestick patterns.

In [58]:

df['CDL_3LINESTRIKE']

Out[58]:

datetime
2021-09-10    0.0
2021-09-13    0.0
2021-09-14    0.0
2021-09-15    0.0
2021-09-16    0.0
             ... 
2022-11-14    0.0
2022-11-15    0.0
2022-11-16    0.0
2022-11-17    0.0
2022-11-18    0.0
Name: CDL_3LINESTRIKE, Length: 300, dtype: float64

In [59]:

df['CDL_HAMMER']

Out[59]:

datetime
2021-09-10    0.0
2021-09-13    0.0
2021-09-14    0.0
2021-09-15    0.0
2021-09-16    0.0
             ... 
2022-11-14    0.0
2022-11-15    0.0
2022-11-16    0.0
2022-11-17    0.0
2022-11-18    0.0
Name: CDL_HAMMER, Length: 300, dtype: float64

Note if the pattern is not true, the value is 0.0, so to find out the days when there is a candlestick pattern, check when the value is not zero as shown below.

In [60]:

for index,row in df.iterrows():
    if row['CDL_HAMMER'] > 0 or row['CDL_HAMMER'] < 0:
        print(index,row['CDL_HAMMER'])

2021-10-13 00:00:00 100.0
2022-03-10 00:00:00 100.0
2022-04-25 00:00:00 100.0
2022-05-02 00:00:00 100.0
2022-09-16 00:00:00 100.0
2022-09-23 00:00:00 100.0

The actual implementation of above scanner is here...
https://tradestie.com/apps/indicators/stocks-candlestick-patterns/