Blockchain API only gives me data for every 4 days

Trying to download daily bitcoin miner fees using the blockchain.com API. However, even though the graph on the website (https://www.blockchain.com/explorer/charts/transaction-fees) shows daily data, the API only gives data for every 4 days.

import requests
import pandas as pd

url = "https://api.blockchain.info/charts/transaction-fees?timespan=all&format=json"

response = requests.get(url)
data = response.json()
df = pd.DataFrame(data['values'])

# Convert the timestamp to a readable format
df['x'] = pd.to_datetime(df['x'], unit='s')  # 'x' is the Unix timestamp
df.set_index('x', inplace=True)  # Set the datetime as the index
df.rename(columns={'y': 'fees'}, inplace=True)  # Rename 'y' to 'hashrate'

print(df.head())

I tried adapting the code to rolling 4d window and interpolating, but there is a quite large error when I do this, as in this case I don't get the correct data for each day, but estimate it based on the surrounding days:

import requests
import pandas as pd

url = "https://api.blockchain.info/charts/transaction-fees?timespan=all&rollingAverage=4days&format=json"

response = requests.get(url)
data = response.json()

df = pd.DataFrame(data['values'])

# Convert the timestamp to a readable format
df['x'] = pd.to_datetime(df['x'], unit='s')  
df.set_index('x', inplace=True)  
df.rename(columns={'y': 'fees'}, inplace=True)  # Rename 'y' to 'hashrate'
df_daily = df.resample('D').interpolate(method='linear')

print(df_daily.head())

-- Update

I assume this is an api limitation. As the raw data also misses 3 days each time: https://api.blockchain.info/charts/transaction-fees

-- Update 2 I have added 'sampled=false' to the api request, and now I get data for every 15min, which is too much. I am just looking for daily data, but the API doc is not so good:

https://www.blockchain.com/explorer/api/charts_api

As stated in the docs:

$sampled- Boolean set to 'true' or 'false' (default 'true'). If true, limits the number of datapoints returned to ~1.5k for performance reasons. (Optional)

Apparently, the API automatically starts adjusting (part of) the timedelta to accommodate the (implied) request for no more than ~1.5k datapoints. It's not immediately clear where it draws the line, but here are 2 workarounds:

Option 1

Use 'sampled': 'false':

import requests
import pandas as pd

url = 'https://api.blockchain.info/charts/transaction-fees'

params = {'start': '2009-01-01',
          'timespan': 'all',
          'format': 'json',
          'sampled': 'false'}

response = requests.get(url, params=params)
data = response.json()

df = pd.DataFrame(data['values'])
df['x'] = pd.to_datetime(df['x'], unit='s')  
df.set_index('x', inplace=True)  
df.rename(columns={'y': 'fees'}, inplace=True)

Option 2

Use a for loop and request smaller chunks:

import requests
import pandas as pd

url = 'https://api.blockchain.info/charts/transaction-fees'

timespan = 6 # up to 6 (years) still gets you daily results, 
# from 7 it starts using '2d' timedelta prior to 2023, and then worse from there

# set params for get request
params = {'start': '2009-01-01',
          'timespan': f'{timespan}years',
          'format': 'json'}

# initialize list to collect values per request
values = []

for year in range(2009, 2024, timespan):
       
    request_params = params.copy()
    request_params['start'] = f'{year}-01-01'
    
    response = requests.get(url, params=request_params)
    
    data = response.json()

    # append values to `values`
    values.extend(data['values'])
    print(f'{year} added')

df = pd.DataFrame(values)
df['x'] = pd.to_datetime(df['x'], unit='s')  
df.set_index('x', inplace=True)  
df.rename(columns={'y': 'fees'}, inplace=True)

Output:

import numpy as np

df.iloc[np.r_[0:2, -2:0]]

                 fees
x                    
2009-01-17   0.000000
2009-01-18   0.000000
2024-10-12  44.805855
2024-10-13  24.719396

Plot:

import seaborn as sns
import matplotlib.pyplot as plt

sns.set_theme(rc={'figure.figsize':(20, 10)})
sns.set(style="ticks")
sns.lineplot(data=df['fees'])
plt.grid()
plt.show()

Compare the chart on the website:

Actually, the chart on the website is itself skipping days, if you set it to 'all'. E.g.:

Compare that to the raw data retrieved:

df.loc['2017-12-20':'2017-12-23']

                   fees
x                      
2017-12-20   911.591743
2017-12-21  1128.761670 # matches value as shown in snippet website (1,128.762)
2017-12-22  1495.946477 # peak, not shown on website
2017-12-23  1065.615595