NumPy for Traders

# NumPy basics for traders -- everything starts with creating an array.
import numpy as np

# An array from a Python list: five daily closes of RELIANCE (in rupees).
closes = np.array([1305.0, 1298.5, 1310.4, 1318.6, 1312.7])
print("From a list :", closes)

# Handy ready-made arrays:
print("arange      :", np.arange(1, 6))           # 1..5, like range()
print("zeros       :", np.zeros(5))               # five 0.0s (empty P&L slots)
print("ones        :", np.ones(3))                # three 1.0s
print("full        :", np.full(4, 100))           # four 100s (e.g. lot sizes)
print("linspace    :", np.linspace(100, 200, 5))  # 5 evenly spaced, 100 to 200

# Every array knows its own shape, size and data type.
import numpy as np

closes = np.array([1305.0, 1298.5, 1310.4, 1318.6, 1312.7])

print("Array    :", closes)
print("shape    :", closes.shape)    # (5,) -> 5 elements in one row
print("ndim     :", closes.ndim)     # 1  -> one dimension
print("size     :", closes.size)     # 5  -> total count
print("dtype    :", closes.dtype)    # float64 -> decimal numbers

qty = np.array([10, 25, 15])
print("int dtype:", qty.dtype)       # int64 -> whole numbers (e.g. quantities)

# Reach into an array by position -- the same indexing you used for lists.
import numpy as np

# Ten days of NIFTY closing values.
nifty = np.array([25900, 25960, 25840, 25990, 26050,
                  26010, 25880, 25930, 26100, 26075])

print("First day   :", nifty[0])
print("Latest day  :", nifty[-1])     # -1 is "most recent" -- used constantly
print("Days 2 to 4 :", nifty[1:4])    # positions 1,2,3 (the stop is excluded)
print("Last 5 days :", nifty[-5:])
print("Every 2nd   :", nifty[::2])    # a step of 2

# Vectorisation: do the math on the WHOLE array at once -- no loops.
import numpy as np

closes = np.array([1305.0, 1298.5, 1310.4, 1318.6, 1312.7])

# Broadcasting: combine an array with a single number.
print("After 5% rise :", np.round(closes * 1.05, 2))
print("Minus 10 rs   :", closes - 10)

# Two arrays, multiplied element by element: price x quantity held.
qty = np.array([10, 10, 20, 20, 15])
value = closes * qty
print("Holding value :", value)
print("Total value   :", value.sum())

# The summary functions a trader reaches for every day.
import numpy as np

closes = np.array([1305.0, 1298.5, 1310.4, 1318.6, 1312.7])

print("mean   :", round(closes.mean(), 2))    # average close
print("max    :", closes.max())               # highest
print("min    :", closes.min())               # lowest
print("std    :", round(closes.std(), 2))     # spread -- a volatility proxy
print("median :", np.median(closes))
print("sum    :", closes.sum())
print("cumsum :", np.cumsum(closes))          # running total

# Ask a yes/no question of every element, then filter or count.
import numpy as np

closes = np.array([1305.0, 1298.5, 1310.4, 1318.6, 1312.7])
avg = closes.mean()

mask = closes > avg                        # True/False for each day
print("Above average?:", mask)
print("Those closes  :", closes[mask])     # keep only the True ones
print("How many days :", mask.sum())       # True counts as 1

# np.where labels every element in one stroke (1 = above avg, 0 = not).
tags = np.where(closes > avg, 1, 0)
print("Tags          :", tags)

# A grab-bag of popular NumPy functions you'll use on price data.
import numpy as np

closes = np.array([1305.0, 1298.5, 1310.4, 1318.6, 1312.7])

print("diff    :", np.diff(closes))               # day-to-day change
print("abs     :", np.abs(np.diff(closes)))       # size of move, sign dropped
print("round   :", np.round(closes, 0))           # round to whole rupees
print("sqrt    :", np.round(np.sqrt(closes), 1))
print("maximum :", np.maximum(closes, 1310))      # floor each value at 1310
print("sort    :", np.sort(closes))               # ascending
print("unique  :", np.unique([5, 5, 10, 20, 20])) # distinct values, sorted

# A 2D array is a table: here three stocks (rows) over four days (columns).
import numpy as np

# rows = RELIANCE, TCS, INFY ; columns = Mon..Thu closes
prices = np.array([
    [1305.0, 1298.5, 1310.4, 1312.7],
    [2068.0, 2090.5, 2075.0, 2081.25],
    [1052.0, 1041.3, 1055.8, 1048.5],
])

print("shape        :", prices.shape)        # (3, 4) -> 3 rows, 4 columns
print("INFY row     :", prices[2])           # the third stock
print("Thu column   :", prices[:, -1])       # last day, every stock
print("Per-stock avg:", np.round(prices.mean(axis=1), 1))  # across each row
print("Per-day avg  :", np.round(prices.mean(axis=0), 1))  # down each column
print("Reshaped     :", np.arange(6).reshape(2, 3))