tony lam

Technical Indicator - Relative Strength Index (RSI)

Quantitative Model

What is Relative Strength Index?

The relative strength index (RSI) is a technical indicator for financial market analysis. By measuring the magnitude of recent price changes, it evaluates the over-bought or over-sold conditions of a stock or other financial instruments.

RSI is classified as a momentum oscillator, where.

  • Momentum refers to the rate of the rise or fall in price
  • Oscillator means its value always moves between two extremes (i.e 0 to 100 for RSI)

The relative strength index was originally developed by J. Welles Wilder and published in a 1978 book, "New Concepts in Technical Trading Systems". Nowadays, it has become one of the most popular oscillator indicators in technical analysis.



Calculation

First, we defined 2 series {Ut} and {Dt} representing the upward and downward price changes respectively. For t = 1,2,3,...

If Closet - Closet-1 >= 0,

  • Ut = Closet - Closet-1
  • Dt = 0

If Closet - Closet-1 < 0,

  • Ut = 0
  • Dt = Closet-1 - Closet

Now, an averaging function AVG(.) is applied to {Ut} and {Dt} to derive the n-period average. There are 2 commonly used averaging functions:

  • Simple Moving Average (SMA) - a rolling mean on the past n observations.

    • SMAt = (Xt + Xt-1 + ... + Xt-n+1) / n

  • Exponential Moving Average (EMA) - all historical observations are exponentially weighted

    • EMAt = Xt * α + EMAt-1 * (1-α), where α = 1 / n

The ratio of these averages is the relative strength:

RS = AVG(U,n) / AVG(D,n)

The relative strength index is then computed as:

RSI = 100 - 100/(1 + RS)


Interpretation

An RSI reading of 70 or above is usually interpreted as a financial instrument is becoming overbought or overvalued, which may indicate a trend reversal or corrective pullback in price in the near future. An RSI reading of 30 or below could be interpreted as an oversold or undervalued condition.

As seen from above formula, if a stock keep rising, we will get a zero sequence of {Dt}. Hence AVG(D,n) will be zero and the RS value will approach infinity, so that the resulting RSI will approach 100. As a result, this stock is said to be overbought.



RSI Implementation

There is a technical analysis python package 'ta-lib' that provides the RSI computation.

1
2
3
import talib
_in = [...]
_out = talib.RSI(_in, _n)

On the other hand, as the RSI computation is not too complicated, it is also a good exercise to implement the code on our own.

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
def RSI(data, n, isEMA=True):
    m = len(data)-1
    k = 1/float(n)
    RSI = [None]*m
    # error handling for insufficient data
    if m<n:
        return RSI
    # compute 1 step price difference
    delta = [data[i+1]-data[i] for i in range(0,m)]
    # create the positive gains (up) and negative gains (down) series
    gain = [abs(i) if i>0 else 0 for i in delta]
    loss = [abs(i) if i<0 else 0 for i in delta]
    avg_gain = sum(gain[0:n])/float(n)
    avg_loss = sum(loss[0:n])/float(n)
    
    for i in range(n,m):
        # Exponential Moving Average
        if isEMA:
            avg_gain = gain[i]*k + avg_gain*(1-k)
            avg_loss = loss[i]*k + avg_loss*(1-k)
        # Simple Moving Average
        else:
            avg_gain = (avg_gain*n - gain[i-n] + gain[i])*k
            avg_loss = (avg_loss*n - loss[i-n] + loss[i])*k
        # calculate RSI value
        if avg_loss==0:
            RSI[i] = 100.0
        else:
            RSI[i] = 100.0 - (100.0 / (1.0 + avg_gain/avg_loss))
    return RSI


Example

Suppose we have 100 observations of daily closing prices, and we are interested in calculating a 14-day RSI.


10, 10.1, 9.92, 9.89, 9.19, 8.5, 8.96, 8.04, 7.67, 8.6, 8.72, 9.14, 10.03, 10.79, 11.39, 11.71, 11.1, 12.05, 11.21, 11.5, 12, 12.53, 12.59, 13.53, 12.87, 13.27, 13.72, 13.3, 13.88, 14.17, 14.77, 14.15, 13.75, 14.51, 13.57, 12.57, 13.02, 13.21, 12.41, 12.28, 11.41, 12, 11, 10.81, 10.07, 9.9, 9.77, 10.48, 11.38, 10.46, 10.28, 10.3, 10.78, 11.49, 12.27, 12.59, 12.74, 12.08, 11.19, 12.07, 13, 13.87, 13.15, 12.91, 13.19, 13.17, 12.24, 12.35, 12.43, 12.75, 11.97, 12.93, 13.73, 13.33, 13.74, 13.54, 14.23, 13.41, 12.52, 13.34, 14.06, 14, 13.83, 14.32, 15.18, 15.08, 14.86, 13.99, 13.48, 12.89, 13.17, 14.16, 13.5, 13.27, 13.62, 13.69, 13.4, 12.74, 13.09, 13.96 


 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
%matplotlib inline
import numpy as np
from matplotlib import pyplot as plt
import talib

period = 14
obs = [10, 10.1, 9.92, 9.89, 9.19, 8.5, 8.96, 8.04, 7.67, 8.6, \
        8.72, 9.14, 10.03, 10.79, 11.39, 11.71, 11.1, 12.05, 11.21, 11.5, \
        12, 12.53, 12.59, 13.53, 12.87, 13.27, 13.72, 13.3, 13.88, 14.17, \
        14.77, 14.15, 13.75, 14.51, 13.57, 12.57, 13.02, 13.21, 12.41, 12.28, \
        11.41, 12, 11, 10.81, 10.07, 9.9, 9.77, 10.48, 11.38, 10.46, \
        10.28, 10.3, 10.78, 11.49, 12.27, 12.59, 12.74, 12.08, 11.19, 12.07, \
        13, 13.87, 13.15, 12.91, 13.19, 13.17, 12.24, 12.35, 12.43, 12.75, \
        11.97, 12.93, 13.73, 13.33, 13.74, 13.54, 14.23, 13.41, 12.52, 13.34, \
        14.06, 14, 13.83, 14.32, 15.18, 15.08, 14.86, 13.99, 13.48, 12.89, \
        13.17, 14.16, 13.5, 13.27, 13.62, 13.69, 13.4, 12.74, 13.09, 13.96]

def RSI(data, n, isEMA=True):
    m = len(data)-1
    k = 1/float(n)
    RSI = [None]*m

    # error handling for insufficient data
    if m<n:
        return RSI

    # compute 1 step price difference
    delta = [data[i+1]-data[i] for i in range(0,m)]

    # create the positive gains (up) and negative gains (down) series
    gain = [abs(i) if i>0 else 0 for i in delta]
    loss = [abs(i) if i<0 else 0 for i in delta]
    avg_gain = sum(gain[0:n])/float(n)
    avg_loss = sum(loss[0:n])/float(n)
    
    for i in range(n,m):
        # Exponential Moving Average
        if isEMA:
            avg_gain = gain[i]*k + avg_gain*(1-k)
            avg_loss = loss[i]*k + avg_loss*(1-k)
        # Simple Moving Average
        else:
            avg_gain = (avg_gain*n - gain[i-n] + gain[i])*k
            avg_loss = (avg_loss*n - loss[i-n] + loss[i])*k
        # calculate RSI value
        if avg_loss==0:
            RSI[i] = 100.0
        else:
            RSI[i] = 100.0 - (100.0 / (1.0 + avg_gain/avg_loss))
    return RSI

# calculate RSI series 
RSI1 = talib.RSI(np.array(obs), period)
RSI2 = RSI(obs, period)
print("RSI1=",RSI1)
print("RSI2=",RSI2)
# plot 
x = np.array([i for i in range(len(obs))])
y = np.array([i for i in obs])
y2 = RSI1
plt.figure()
plt.style.use(['dark_background']) 
fig, axs = plt.subplots(2)
axs[0].set_title('Closing Price')
axs[0].plot(x, y, linestyle='-', color='b')
axs[1].set_title('RSI')
axs[1].plot(x, y2, linestyle='-', color='g')
plt.show()

Result

result

Now, we have a better understanding about RSI! In the coming post, let's create an RSI strategy and see how useful it is.




Like and follow me

If you find my articles inspiring, like this post and follow me here to receive my latest updates.

Enter my promote code "AjpQDMOSmzG2" for any purchase on ALGOGENE, you will automatically get 5% discount.


3 0
Posted on : 2021-05-18 13:45:59.776000
 
Bee Bee
Very clear explanation. Thanks for sharing! 
0 0
Posted on : 2021-05-27 11:46:59.411000
 
Jeremy
I compared the results from talib and the provided RSI function. 
I found the first few generated RSI values are different but the rest are the same. Any reasons?
0 0
Posted on : 2021-06-19 01:10:42.463000
 
tony lam
Original Posted by - b'Jeremy':
I compared the results from talib and the provided RSI function. 
I found the first few generated RSI values are different but the rest are the same. Any reasons?
The above RSI function applies a simple average of the first n observations as initial value. 
"ta-lib" may have a different initialization method. 
For EMA version of RSI, as historical observation carries a decay factor k^(n-1), the results will converge even if the initial values are different. 
0 0
Posted on : 2021-06-29 13:22:51.734000



Categories