Shwetank Singh

GritSetGrow - GSGLearn.com

DATA AND AI

WINDOW FUNCTIONS
WITH
REAL-TIME EXAMPLES
& EXPLANATIONS
www.gsglearn.com
Introduction to Window Functions
Window functions perform calculations across a set of rows related to the current row,
while retaining the original granularity of the dataset. They are essential for:

Time-series analysis (e.g., stock price trends).

Rankings (e.g., top-performing stocks).
Cumulative metrics (e.g., running totals of trades).
Comparative analysis (e.g., daily price changes).

Unlike aggregate functions (e.g., SUM , AVG ), window functions do not collapse rows
into summaries. Instead, they compute values for each row based on a defined window of
rows.

2. Finance Dataset Design

We create a detailed dataset to simulate real-world financial analysis. The table
includes stock prices, dividends, and trading volumes for three companies (AAPL,
GOOGL, MSFT) over 10 trading days.

Table Structure: stock_data

Column Description Data Type

stock_symbol Company ticker (e.g., AAPL) VARCHAR(10)

date Trading date DATE

open_price Opening stock price DECIMAL(10,2)

high_price Highest price during the day DECIMAL(10,2)

low_price Lowest price during the day DECIMAL(10,2)

closing_price Closing stock price DECIMAL(10,2)

volume Number of shares traded INT

dividend Dividend per share (if applicable) DECIMAL(10,2)

Sample Data

-- Create table
CREATE TABLE stock_data (
stock_symbol VARCHAR(10),
date DATE,
open_price DECIMAL(10,2),
high_price DECIMAL(10,2),
low_price DECIMAL(10,2),
closing_price DECIMAL(10,2),
volume INT,
dividend DECIMAL(10,2)
);

-- Insert data for AAPL, GOOGL, MSFT (10 days)

INSERT INTO stock_data VALUES
('AAPL', '2023-10-01', 148.00, 152.00, 147.50, 150.00, 100000, 0.00),
('AAPL', '2023-10-02', 150.50, 153.00, 149.80, 152.00, 120000, 0.00),
...
('GOOGL', '2023-10-01', 2795.00, 2810.00, 2785.00, 2800.00, 50000, 0.00),
...
('MSFT', '2023-10-05', 308.00, 312.00, 307.50, 310.00, 95000, 2.50);

3. Types of Window Functions with Detailed Use Cases

3.1. Ranking Functions

a. ROW_NUMBER()
Assigns a unique sequential integer to rows within a partition.
Use Case: Identify the sequence of days when a stock's closing price increased
consecutively.

SELECT
stock_symbol,
date,
closing_price,
ROW_NUMBER() OVER (
PARTITION BY stock_symbol
ORDER BY date
) AS trading_day_sequence
FROM stock_data;

Output Explanation:

AAPL on 2023-10-01 gets 1 , 2023-10-02 gets 2 , etc.

Helps track trading patterns over time (e.g., streaks of gains/losses).

b. RANK() vs. DENSE_RANK()

RANK() : Skips ranks after ties (e.g., 1, 2, 2, 4).

DENSE_RANK() : No rank gaps (e.g., 1, 2, 2, 3).

Use Case: Rank stocks by daily trading volume.

SELECT
date,
stock_symbol,
volume,
RANK() OVER (PARTITION BY date ORDER BY volume DESC) AS volume_rank,
DENSE_RANK() OVER (PARTITION BY date ORDER BY volume DESC) AS dense_volume_rank
FROM stock_data;

Output Example (2023-10-01):

date stock_symbol volume volume_rank dense_volume_rank

2023-10-01 AAPL 100000 1 1

2023-10-01 MSFT 80000 2 2

 2023-10-01 GOOGL 50000 3 3

c. NTILE(n)
Divides rows into n buckets (e.g., quartiles, deciles).
Use Case: Segment stocks into quartiles based on closing price volatility.

SELECT
stock_symbol,
date,
closing_price,
NTILE(4) OVER (
PARTITION BY stock_symbol
ORDER BY (high_price - low_price) DESC
) AS volatility_quartile
FROM stock_data;

Insight:

Stocks in quartile 1 have the highest daily price swings (high volatility).

3.2. Analytic Functions

a. LAG() and LEAD()
LAG(column, offset) : Accesses data from a previous row.

LEAD(column, offset) : Accesses data from a subsequent row.

Use Case: Calculate day-over-day (DoD) price changes and 5-day momentum.

SELECT
stock_symbol,
date,
closing_price,
LAG(closing_price, 1) OVER (PARTITION BY stock_symbol ORDER BY date) AS
prev_close,
closing_price - LAG(closing_price, 1) OVER (PARTITION BY stock_symbol ORDER BY
date) AS dod_change,
LEAD(closing_price, 5) OVER (PARTITION BY stock_symbol ORDER BY date) AS
next_5day_close
FROM stock_data;

Output Analysis*:

dod_change shows daily price fluctuations.

next_5day_close helps anticipate future prices for momentum strategies.

b. FIRST_VALUE() and LAST_VALUE()

FIRST_VALUE() : Returns the first value in the window.

LAST_VALUE() : Returns the last value in the window.

Use Case: Compare the first and last closing prices of a stock in a 7-day
window.
 SELECT
stock_symbol,
date,
closing_price,
FIRST_VALUE(closing_price) OVER (
PARTITION BY stock_symbol
ORDER BY date
ROWS BETWEEN 6 PRECEDING AND CURRENT ROW
) AS week_start_price,
LAST_VALUE(closing_price) OVER (
PARTITION BY stock_symbol
ORDER BY date
ROWS BETWEEN 6 PRECEDING AND CURRENT ROW
) AS week_end_price
FROM stock_data;

Insight*:

Track weekly performance trends (e.g., +5% increase from Monday to Friday).

3.3. Aggregate Functions with Window Frames

a. Moving Averages
Use Case: 7-day moving average to smooth price volatility.

SELECT
stock_symbol,
date,
closing_price,
AVG(closing_price) OVER (
PARTITION BY stock_symbol
ORDER BY date
ROWS BETWEEN 6 PRECEDING AND CURRENT ROW
) AS moving_avg_7day
FROM stock_data;

Output:

stock_symbol date closing_price moving_avg_7day

AAPL 2023-10-07 158.00 152.86

b. Cumulative Sum
Use Case: Track cumulative dividends paid over time.

SELECT
stock_symbol,
date,
dividend,
SUM(dividend) OVER (
PARTITION BY stock_symbol
ORDER BY date
ROWS BETWEEN UNBOUNDED PRECEDING AND CURRENT ROW
 ) AS cumulative_dividend
FROM stock_data;

Insight:

MSFT's cumulative dividend reaches $10.00 by Q4 2023.

c. Rolling Total Volume

SELECT
stock_symbol,
date,
volume,
SUM(volume) OVER (
PARTITION BY stock_symbol
ORDER BY date
ROWS BETWEEN 2 PRECEDING AND CURRENT ROW
) AS rolling_3day_volume
FROM stock_data;

Business Use:

Identify spikes in trading activity (e.g., rolling volume > 300,000 shares).

3.4. Statistical Functions

a. PERCENT_RANK()
Computes the percentile rank of a row within a partition.
Use Case: Identify days where AAPL's closing price was in the top 10% historically.

SELECT
stock_symbol,
date,
closing_price,
PERCENT_RANK() OVER (
PARTITION BY stock_symbol
ORDER BY closing_price
) AS price_percentile
FROM stock_data
WHERE stock_symbol = 'AAPL';

Interpretation:

A percentile of 0.9 means the price is higher than 90% of historical values.
b. CUME_DIST()
Calculates the cumulative distribution of a value (proportion of rows ≤ current
value).
Use Case: Analyze the distribution of GOOGL's daily trading volumes.

SELECT
stock_symbol,
date,
volume,
CUME_DIST() OVER (
 PARTITION BY stock_symbol
ORDER BY volume
) AS volume_cume_dist
FROM stock_data
WHERE stock_symbol = 'GOOGL';

Output:

A CUME_DIST of 0.7 means 70% of days had volumes ≤ current day's volume.

4. Advanced Window Function Techniques

4.1. Custom Window Frames

Define the range of rows for calculations using ROWS , RANGE , or GROUPS .
Use Case: Compare a stock's closing price to the average of the prior 3 days
(excluding current row).

SELECT
stock_symbol,
date,
closing_price,
AVG(closing_price) OVER (
PARTITION BY stock_symbol
ORDER BY date
ROWS BETWEEN 3 PRECEDING AND 1 PRECEDING
) AS avg_prior_3days
FROM stock_data;

4.2. Multiple Window Partitions

Combine partitions for granular analysis.
Use Case: Rank stocks by volume within each sector (assuming a sector column
exists).

SELECT
stock_symbol,
sector,
date,
volume,
RANK() OVER (
PARTITION BY sector, date
ORDER BY volume DESC
) AS sector_volume_rank
FROM stock_data;

5. Real-World Finance Scenarios

Scenario 1: Identifying Support/Resistance Levels

Goal: Find price levels where a stock frequently reverses direction.

Approach: Use LAG and LEAD to detect local minima/maxima.
WITH price_changes AS (
SELECT
stock_symbol,
date,
closing_price,
LAG(closing_price) OVER (PARTITION BY stock_symbol ORDER BY date) AS
prev_price,
LEAD(closing_price) OVER (PARTITION BY stock_symbol ORDER BY date) AS
next_price
FROM stock_data
)
SELECT
stock_symbol,
date,
closing_price AS potential_support_resistance
FROM price_changes
WHERE (closing_price > prev_price AND closing_price > next_price) -- Resistance
OR (closing_price < prev_price AND closing_price < next_price); -- Support

Scenario 2: Dividend Yield Analysis

Goal: Compare dividend yields across sectors.

Approach: Use AVG with partitions.

SELECT
stock_symbol,
date,
dividend,
closing_price,
(dividend / closing_price) * 100 AS dividend_yield,
AVG(dividend / closing_price) OVER (PARTITION BY sector) * 100 AS sector_avg_yield
FROM stock_data;

6. Performance Optimization
Indexing: Add indexes on stock_symbol and date for faster partitioning.
Frame Size: Use narrow frames (e.g., ROWS 10 PRECEDING ) for large datasets.
Avoid RANGE : Prefer ROWS over RANGE for precise control and faster
execution.

7. Common Pitfalls
1. Omitting ORDER BY : Causes incorrect cumulative sums or rankings.
2. Over-Partitioning: Excessive partitions slow down queries.
3. Ambiguous Window Frames: Always define ROWS / RANGE explicitly.

8. Conclusion
Window functions empower financial analysts to:

Track trends with moving averages.

Rank assets by performance.
Compute running totals (e.g., cumulative returns).
 Analyze distributions (e.g., volatility quartiles).

By combining partitioning, ordering, and custom frames, they unlock granular insights
without aggregating data. For instance, a hedge fund might use RANK() to identify
outperforming stocks or LAG() to model mean-reversion strategies. Mastery of window
functions is critical for modern financial analytics.

Thanks, By Shwetank Singh - GSGLearn.com

Thank you

Shwetank Singh
GritSetGrow - GSGLearn.com