compare cagr result with another python package

[sugizo]

env google colab

code

pip install -U QuantStats openseries okama

from openseries import OpenTimeSeries
import yfinance as yf
import quantstats as qs
import okama as ok

ticker = 'AAPL' 
period = '5y' # 1d, 5d, 1mo, 3mo, 6mo, 1y, 2y, 5y, 10y, ytd, max, etc
asset_list = ['AAPL.US']
ccy = 'USD'
first_date = '2020-05'

# openseries
stock = yf.Ticker(ticker = ticker)
history = stock.history(period = period)
series = OpenTimeSeries.from_df(dframe = history.loc[:, "Close"] )
print('Compound Annual Growth Rate (CAGR) :', series.geo_ret) 

# quantstats
qs.extend_pandas()
stock = qs.utils.download_returns(ticker = ticker, period = period)
print('Compound Annual Growth Rate (CAGR) :', stock.cagr() )

# okama
x = ok.AssetList(asset_list, ccy = ccy, first_date = first_date)
#x.describe(years = [1, 2, 3, 4, 5] )
x.get_cagr()

result

question why the cagr result value differs is so far with another python module ?

thanks and best regards

[tomdlc]

Found the same problem, I have the followig values: starting value: 171978000 ending value: 414839000 period = 5 CAGR = 0.193

the .cagr() function calculates 0.173

Can't find what doing this at first sight.

[tomdlc]

Found the same problem, I have the followig values: starting value: 171978000 ending value: 414839000 period = 5 CAGR = 0.193

the .cagr() function calculates 0.173

Can't find what doing this at first sight.

So the fix is in the "periods" variable for the cagr() function, it defaults to 252 days assuming you have daily data for the input, so in the case you are giving whole year values use cagr(periods=365).

[sugizo]

you right but when you use another function it won't affect the cagr value result even you already set the periods value e.g.

qs.reports.metrics(stock, periods_per_year = 252)
qs.reports.basic(stock, periods_per_year = 252)

detail explained here : https://github.com/CaptorAB/openseries/issues/291#issuecomment-2888422892

best regards

[tomdlc]

Yeah I think you would have to work around that, the focus isn't on stock valuation rather than risk and performance

[xaviergilbert]

it defaults to 252 days assuming you have daily data for the input

The problem is here:

years = (returns.index[-1] - returns.index[0]).days / periods

If you get the number of days you should divide by 365 fullstop.

Either:

• periods should be replaced by 365 here
• it should count the number of rows instead of days and keep dividing by periods

From 2024-01-01 to 2024-12-31 you have 365 days. It makes no sense to divide by any other number than 365

[ranaroussi]

The bug where CAGR did not accept the periods dynamically was fixed in version 0.0.64. Thanks for bringing this to my attention.