Stochastic RSI computation with Python¶
This is not investment advice.
Stochastic RSI indicator is derived from the values of regular RSI indicator (Relative Strength Index). It means it indicates changes in another indicator. This can cause to create lots of false trading signals.
In this article we will try to find ways how to supress these false trading signals.
First we will need to compute ragular RSI, detailed proces is described here:
https://tcoil.info/compute-rsi-for-stocks-with-python-relative-strength-index/
Once we have RSI, we can proceed to compute the Stochastic RSI.
Equation:
$$ StochRSI(n)_{i} = \frac{RSI(n)_{i} - minRSI(n)_{i}}{maxRSI(n)_{i} - minRSI(n)_{i}} \times 100 $$$ StochRSI(n)_{i} $ ... Stochastic RSI of window n at step i
$ RSI(n)_{i} $ ... RSI of window n at step i
$ minRSI(n)_{i} $ ... lowest RSI value looking back over n datapoint window at step i
$ maxRSI(n)_{i} $ ... highest RSI value looking back over n datapoint window at step i
We took the equation from Fidelity, this gives us the value range between 0 and 100: https://www.fidelity.com/learning-center/trading-investing/technical-analysis/technical-indicator-guide/stochrsi
But we are not finished yet. This equation will give us only one datapoint line. We would like to have the same graphs as on tradingview.com for Stochastic RSI.
For that we will check definition of regular Stochastic oscillator and compare it to the equation above.
$$ \%K(n)_{i} = \frac{Close_{i} - minClose(n)_{i}}{maxClose(n)_{i} - minClose(n)_{i}} \times 100 $$$ \%K(n)_{i} $ ... so called Slow stochastic indicator at step i
$ Close(n)_{i} $ ... RSI of window n at step i
$ minClose(n)_{i} $ ... lowest closing price looking back over n datapoint window at step i
$ maxClose(n)_{i} $ ... highest closing looking back over n datapoint window at step i
$ \%D(n)_{i} $ ... so called Fast stochastic indicator at step i, it is typically 3 datapoint moving average of $ \%K $
From that we see that if we feed RSI values into Stochastic equation, we get ...
... wait for it ...
Yes, we get Stochastic RSI.
We will construct the graph using $ \%K $ and $ \%D$.
Buy/Sell signals are generated when $\%K$ and $\%D$ lines cross (and ideally when they cross when stochastic is in oversold or overbought levels)
Oversold level: values below or equal 20 Overbought level: above or equal 80
Regular Stochastic definition: https://www.investopedia.com/terms/s/stochasticoscillator.asp
Tradingview is showing parameters for the stiochastic RSI function like this:
StochRSI(3,3,14,14)
which suggests that we take RSI 14 window, then from that we take Stochastic 14 window and then both %K and %D are smoothed with 3 window. Smoothing for %K was not defined in above equations but I will implement it in the code, because I take tradingview as a benchmark.
Details about trading strategies with Stoch RSI:
https://www.investopedia.com/terms/s/stochrsi.asp
https://www.tradingview.com/support/solutions/43000502333-stochastic-rsi-stoch-rsi/
https://academy.binance.com/economics/stochastic-rsi-explained
https://tradingsim.com/blog/stochastic-rsi/
https://www.incrediblecharts.com/indicators/stochastic-rsi.php
https://www.fidelity.com/learning-center/trading-investing/technical-analysis/technical-indicator-guide/stochrsi
Importing libraries¶
#optional installations:
#!pip install yfinance --upgrade --no-cache-dir
#!pip3 install pandas_datareader
from IPython.core.display import display, HTML
display(HTML("<style>.container { width:100% !important; }</style>"))
# ___library_import_statements___
import pandas as pd
# for pandas_datareader, otherwise it might have issues, sometimes there is some version mismatch
pd.core.common.is_list_like = pd.api.types.is_list_like
# make pandas to print dataframes nicely
pd.set_option('expand_frame_repr', False)
import pandas_datareader.data as web
import numpy as np
import matplotlib.pyplot as plt
import datetime
import time
#newest yahoo API
import yfinance as yahoo_finance
#optional
#yahoo_finance.pdr_override()
%matplotlib inline
# ___variables___
ticker = 'TSLA'
start_time = datetime.datetime(2017, 1, 1)
#end_time = datetime.datetime(2019, 1, 20)
end_time = datetime.datetime.now().date().isoformat() # today
Get stock data¶
# yahoo gives only daily historical data
connected = False
while not connected:
try:
ticker_df = web.get_data_yahoo(ticker, start=start_time, end=end_time)
connected = True
print('connected to yahoo')
except Exception as e:
print("type error: " + str(e))
time.sleep( 5 )
pass
# use numerical integer index instead of date
#ticker_df = ticker_df.reset_index()
print(ticker_df.head(5))
df = ticker_df
Compute RSI¶
def computeRSI (data, time_window):
diff = data.diff(1).dropna() # diff in one field(one day)
#this preservers dimensions off diff values
up_chg = 0 * diff
down_chg = 0 * diff
# up change is equal to the positive difference, otherwise equal to zero
up_chg[diff > 0] = diff[ diff>0 ]
# down change is equal to negative deifference, otherwise equal to zero
down_chg[diff < 0] = diff[ diff < 0 ]
# check pandas documentation for ewm
# https://pandas.pydata.org/pandas-docs/stable/reference/api/pandas.DataFrame.ewm.html
# values are related to exponential decay
# we set com=time_window-1 so we get decay alpha=1/time_window
up_chg_avg = up_chg.ewm(com=time_window-1 , min_periods=time_window).mean()
down_chg_avg = down_chg.ewm(com=time_window-1 , min_periods=time_window).mean()
rs = abs(up_chg_avg/down_chg_avg)
rsi = 100 - 100/(1+rs)
return rsi
df['RSI'] = computeRSI(df['Adj Close'], 14)
print(df.head())
print(df.tail())
Stochastic¶
def stochastic(data, k_window, d_window, window):
# input to function is one column from df
# containing closing price or whatever value we want to extract K and D from
min_val = data.rolling(window=window, center=False).min()
max_val = data.rolling(window=window, center=False).max()
stoch = ( (data - min_val) / (max_val - min_val) ) * 100
K = stoch.rolling(window=k_window, center=False).mean()
#K = stoch
D = K.rolling(window=d_window, center=False).mean()
return K, D
df['K'], df['D'] = stochastic(df['RSI'], 3, 3, 14)
df.head()
df.tail()
Plotting¶
def plot_price(df):
# plot price
plt.figure(figsize=(15,5))
plt.plot(df['Adj Close'])
plt.title('Price chart (Adj Close)')
plt.show()
return None
def plot_RSI(df):
# plot correspondingRSI values and significant levels
plt.figure(figsize=(15,5))
plt.title('RSI chart')
plt.plot(df['RSI'])
plt.axhline(0, linestyle='--', alpha=0.1)
plt.axhline(20, linestyle='--', alpha=0.5)
plt.axhline(30, linestyle='--')
plt.axhline(70, linestyle='--')
plt.axhline(80, linestyle='--', alpha=0.5)
plt.axhline(100, linestyle='--', alpha=0.1)
plt.show()
return None
def plot_stoch_RSI(df):
# plot corresponding Stoch RSI values and significant levels
plt.figure(figsize=(15,5))
plt.title('stochRSI chart')
plt.plot(df['K'])
plt.plot(df['D'])
plt.axhline(0, linestyle='--', alpha=0.1)
plt.axhline(20, linestyle='--', alpha=0.5)
#plt.axhline(30, linestyle='--')
#plt.axhline(70, linestyle='--')
plt.axhline(80, linestyle='--', alpha=0.5)
plt.axhline(100, linestyle='--', alpha=0.1)
plt.show()
return None
def plot_all(df):
plot_price(df)
plot_RSI(df)
plot_stoch_RSI(df)
return None
plot_all(df)
As we can see, the stochastic RSI is quite noisy. Either we can trade on shorter timeframes or if we prefer longer term swing trading we can analyze weekly Stochastic RSI that should give much stronger signals.
Supressing false signals in stoch RSI¶
We can supress the noise by using weekly timeframe. Note that by using weekly timeframe, you will lose lots of trailing data, since we are using 14 week window RSI and from that we have 14 week window stochastic.
# weekly timeframe aggregation
agg_dict = {'Open': 'first',
'High': 'max',
'Low': 'min',
'Close': 'last',
'Adj Close': 'last',
'Volume': 'mean'}
# resampled dataframe
# 'W' means weekly aggregation
df_res = df.resample('W').agg(agg_dict)
df_res['RSI'] = computeRSI(df_res['Adj Close'], 14)
df_res['K'], df_res['D'] = stochastic(df_res['RSI'], 3, 3, 14)
def plot_weekly_stoch_RSI(df_res):
# WEEKLY plot corresponding Stoch RSI values and significant levels
plt.figure(figsize=(15,5))
plt.title('WEEKLY stochRSI chart')
df_res = df_res.reset_index()
plt.plot(df_res['Date'], df_res['K'].fillna(0)) # NaN to zeros so plot is in scale
plt.plot(df_res['Date'], df_res['D'].fillna(0)) # NaN to zeros so plot is in scale
#df_res.reset_index().plot(x='Date', y='K', figsize=(15,5))
#df_res.reset_index().plot(x='Date', y='D', figsize=(15,5))
plt.axhline(0, linestyle='--', alpha=0.1)
plt.axhline(20, linestyle='--', alpha=0.5)
#plt.axhline(30, linestyle='--')
#plt.axhline(70, linestyle='--')
plt.axhline(80, linestyle='--', alpha=0.5)
plt.axhline(100, linestyle='--', alpha=0.1)
plt.show()
return None
def plot_mixed(df, df_res):
plot_price(df)
plot_RSI(df)
plot_weekly_stoch_RSI(df_res)
return None
plot_mixed(df, df_res)
The weekly stochastic signals look pretty actionable.
While comparing daily closing price, daily RSI and weekly Stochastic RSI, we can see that the weekly Stochastic %K and %D lines are crossing over during overbought/oversold conditions around the same time when daily RSI is signalling more or less the same.
This gives us actionable confluence of signals to take positions.
Sources:¶
https://www.fidelity.com/learning-center/trading-investing/technical-analysis/technical-indicator-guide/stochrsi https://tradingsim.com/blog/stochastic-rsi/ https://www.investopedia.com/terms/s/stochrsi.asp
Details about trading strategies with Stoch RSI:
https://www.investopedia.com/terms/s/stochrsi.asp
https://www.tradingview.com/support/solutions/43000502333-stochastic-rsi-stoch-rsi/
https://academy.binance.com/economics/stochastic-rsi-explained
https://tradingsim.com/blog/stochastic-rsi/
https://www.incrediblecharts.com/indicators/stochastic-rsi.php
https://www.fidelity.com/learning-center/trading-investing/technical-analysis/technical-indicator-guide/stochrsi
Good algo for stochastic function:
https://github.com/tristcoil/Binance-StochasticRSI/blob/master/main.py