Pandas for strategies

import pandas as pd
import numpy as np
from finlab import data
from finlab.markets.us import USMarket

data.set_market('us')

# DataFrame: US adjusted close prices
df = data.get('price:adj_close')
df = df[df.index.dayofweek < 5]  # remove weekends
df = df[df.index <= '2024-07-01']

# Series: single stock
series = df['AAPL']

df = data.get('price:adj_close')
df = df[df.index.dayofweek < 5]  # remove weekends

# head(n): get the first n rows, default is 5
demo_head = df.head(3)

# tail(n): get the last n rows, default is 5
demo_tail = df.tail(3)

# index: get the row index (dates)
demo_index = df.index

# columns: get the column labels (stock symbols)
demo_columns = df.columns

# values: get the underlying numpy array
demo_values = df.values

# Boolean indexing: filter rows by condition
# Example: get data from 2023 onwards
demo_boolean_indexing = df[df.index > '2023']

# loc: label-based selection
# Example: select all stocks for a specific date range
demo_loc1 = df.loc['2024-01-02':'2024-01-08']

# Example: select specific stocks (two equivalent methods)
demo_loc2 = df.loc[:, ['AAPL', 'MSFT', 'GOOGL']]
demo_column_select = df[['AAPL', 'MSFT', 'GOOGL']]

# Example: select specific date range AND specific stocks
demo_loc3 = df.loc['2024-01-02':'2024-01-08', ['AAPL', 'MSFT', 'GOOGL']]

# iloc: integer position-based selection
# Example: select rows 6 through 10
demo_iloc1 = df.iloc[5:10]

# Example: select columns 6 through 10
demo_iloc2 = df.iloc[:, 5:10]

# Example: select rows 6-10 AND columns 6-10
demo_iloc3 = df.iloc[5:10, 5:10]

# nlargest: get top n values
# Example: get the 10 highest-priced stocks on the most recent day
demo_nlargest = df.iloc[-1].nlargest(10)

# nsmallest: get bottom n values
# Example: get the 10 lowest-priced stocks on the most recent day
demo_nsmallest = df.iloc[-1].nsmallest(10)

df = data.get('price:adj_close')

# Show all rows
pd.set_option("display.max_rows", None)

# Show all columns
pd.set_option("display.max_columns", None)

# Show at most 20 rows
pd.set_option("display.max_rows", 20)

# Show at most 20 columns
pd.set_option("display.max_columns", 20)

# Reset all display options to defaults
pd.reset_option("^display")

df = data.get('price:adj_close')
df = df[df.index.dayofweek < 5]  # remove weekends

# Create a boolean DataFrame: True where price > 100
price_above_100 = df > 100

# sum: aggregate along an axis (axis=0 sums each column, axis=1 sums each row)
# Example: count how many stocks are priced above $100 each day
demo_sum = price_above_100.sum(axis=1)

# mean: average value (commonly used for computing moving averages)
demo_mean = df.mean()

# median: middle value
demo_median = df.median()

# std: standard deviation (commonly used for Bollinger Bands, deviation rate)
demo_std = df.std()

# max: maximum value
demo_max = df.max()

# min: minimum value
demo_min = df.min()

# cumsum: cumulative sum
# Example: cumulative dollar change
demo_diff = df.diff()
demo_cumsum = demo_diff.cumsum()

# cumprod: cumulative product
# Example: compute cumulative return
demo_pct_change = df.pct_change() + 1
demo_cumprod = demo_pct_change.cumprod()

# cummax: running maximum (commonly used for drawdown calculation)
demo_cummax = df.cummax()

# cummin: running minimum
demo_cummin = df.cummin()

# quantile: get the value at the nth percentile
# Example: get the 90th percentile price on the most recent day
demo_quantile = df.iloc[-1].quantile(0.9)

# corr: pairwise correlation
demo_corr = df.iloc[:, :5].corr()

# describe: summary statistics
demo_describe = df.describe()

df = data.get('price:adj_close')
df = df[df.index.dayofweek < 5]  # remove weekends

# shift: move data forward/backward in time
# Example: shift prices down by 1 row (yesterday's price)
demo_shift1 = df.shift()
# Example: shift prices up by 1 row (tomorrow's price -- use with caution to avoid look-ahead bias)
demo_shift2 = df.shift(-1)

# rolling: apply a function over a sliding window
# Example: 20-day moving average (NaN for the first 9 rows, partial calculation starting from row 10)
demo_rolling = df.rolling(20, min_periods=10).mean()

# diff: row-to-row difference
# Example: daily price change in dollars
demo_diff = df.diff()

# pct_change: row-to-row percentage change
# Example: daily return
demo_pct_change = df.pct_change()

df = data.get('price:adj_close')
df = df[df.index.dayofweek < 5]  # remove weekends

# 10-day moving average for AAPL and MSFT
df[['AAPL', 'MSFT']].average(10)

# Top 5 stocks by price over the last 10 days
df.iloc[-10:].is_largest(5)

# Bottom 5 stocks by price over the last 10 days
df.iloc[-10:].is_smallest(5)

# Did the price rise compared to yesterday?
df.iloc[-10:, 10:15].rise()

# Did the price fall compared to yesterday?
df.iloc[-10:, 10:15].fall()

# Did the price rise for 2 consecutive days?
df.iloc[-10:, 10:15].rise().sustain(2)

# Get the 90th percentile price across all stocks for the last 10 days
df.iloc[-10:].quantile_row(0.9)

df = data.get('price:adj_close')
df = df[df.index.dayofweek < 5]  # remove weekends
aapl = df.iloc[-400:]['AAPL']

mean_20 = aapl.rolling(20).mean()
std_value = aapl.rolling(20).std()

up = mean_20 + std_value * 2
down = mean_20 - std_value * 2
up.plot(label='upper band', legend=True)
down.plot(label='lower band', legend=True)
mean_20.plot(label='MA 20', legend=True)
aapl.plot(
    title='AAPL Bollinger Bands', label='close', legend=True, figsize=(20, 8), grid=True
)

df = data.get('price:adj_close')
df = df[df.index.dayofweek < 5]  # remove weekends
check_list = ['AAPL', 'MSFT', 'GOOGL', 'AMZN', 'NVDA', 'META']
demo_corr = df.iloc[-600:][check_list].corr()

demo_corr.style.background_gradient(cmap='viridis').set_properties(
    **{'font-size': '20px'}
)