MA2830 — Portfolio Hedging (Tableau Assignment)

Last updated: Nov 5, 2025, 23:00 SGT

Contents: Finance Crash Course · Upfront EDA · Get Live Data · Tableau Steps · Deliverables · Grading Rubric

Finance Crash Course (Start Here)

Open/Close Crash Course

Motivation — Why hedge?

Ticker choice: Any liquid portfolio ticker with a sensible hedge is acceptable. Any tickers shown are examples only, not required.

Big losses kill compounding. A hedge aims to reduce the worst falls so you can stay invested.

We compare three paths: Unhedged, Static (fixed ratio), Dynamic (rolling beta).

Key ideas

Return = today/prev − 1; Volatility = day‑to‑day wiggle; Drawdown = drop from prior peak.

Correlation (−1→+1) and Beta (slope) tell how portfolio moves vs hedge.

Static hedge: one fixed ratio daily (e.g., 1.0). Dynamic hedge: use recent beta each day.

Sharpe proxy = Annual Return / Annual Volatility (return per unit of wiggle).

Math you’ll implement

Daily return: today/prev - 1

Static hedged: port_r - ratio * hedge_r

Dynamic hedged: port_r - beta90_clamped * hedge_r

Equity: 100 * EXP(RUNNING_SUM(LOG(1 + r)))

Drawdown: equity / WINDOW_MAX(equity) - 1

Annual return: EXP(AVG(LOG(1+r))*252)-1; Annual vol: STDEV(r)*SQRT(252)

Static vs Dynamic — When each wins / loses

Quick idea: Static = fixed hedge size (your Hedge Ratio parameter). Dynamic = use the recent Rolling Beta (90d) (clamped 0–1.5) as the hedge size.

Static tends to win when…

Regime is steady: Relationship between portfolio & hedge is stable (beta ~ constant).

Transaction costs / friction matter: You prefer no daily resizing.

Signal is noisy: Rolling beta is jumpy; fixed size avoids over‑reacting.

Dynamic tends to win when…

Regime drifts: Beta trends up/down across months; dynamic tracks it.

Market mix changes: Correlation/volatility shift (e.g., rotations, macro shocks).

Risk focus: You want the hedge to scale automatically with current sensitivity.

Dynamic can lose when…

Very fast shocks: Beta estimate lags reality → mis‑sized hedge in the moment.

Beta flips sign briefly: Noise can produce negative/huge betas — clamp (0–1.5) prevents blow‑ups.

Overfitting short windows: Too small a window chases noise; keep ~90 days.

Read your dashboards: Use D2 Rolling Beta to spot regime drift; compare KPIs in D3 Static vs Dynamic; confirm protection in shock windows via D5 Stress.

Tiny example: If 90‑day beta ~0.6, Dynamic hedges less than Static 1.0, keeping more upside in calm periods. If beta rises toward 1.2, Dynamic hedges more, cutting drawdowns.

Upfront Assessment (Exploratory Visuals)

Goal: Build quick intuition about your chosen Portfolio and Hedge before hedging.

Monthly Return Heatmap (Highlight Table) — rows = Year, cols = Month, color = Avg Monthly Return of Portfolio.

Daily Return Box Plot by Month — distribution of Portfolio daily returns by Month.

Daily Return Histogram — frequency of Portfolio daily returns.

Rolling 90‑day Correlation Heatmap — Portfolio Return vs Hedge Return (−1→+1).

Upfront EDA — Guiding Questions (answer in captions)

Which months are typically strongest/weakest for the Portfolio? (Heatmap)

Which months show the widest spread or most outliers? (Box plot)

Is the return distribution symmetric or skewed? (Histogram)

When does correlation with the hedge strengthen or weaken? (Rolling correlation)

Given these patterns, what hedge ratio might work better: lower, ~1.0, or higher?

Get Live Data (choose one)

A. Google Sheets (no code)

Create tabs: PORT, HEDGE, MERGE. Put tickers in PORT!A1 and HEDGE!A1.

In PORT!A3 paste:
=QUERY(GOOGLEFINANCE($A$1,"close",EDATE(TODAY(),-36),TODAY()), "select Col1,Col2 where Col2 is not null label Col1 'Date', Col2 'Close'",1)

Repeat same in HEDGE!A3.

In MERGE!A2 paste:
=QUERY({PORT!A3:B, ARRAYFORMULA(IFNA(VLOOKUP(PORT!A3:A, HEDGE!A3:B, 2, FALSE)))}, "select Col1,Col2,Col3 where Col3 is not null label Col1 'Date', Col2 'PortfolioClose', Col3 'HedgeClose'",0)
Download MERGE as prices_clean.csv.

B. Python (yfinance, ~10 lines)

import yfinance as yf, pandas as pd PORT, HEDGE = "QQQ", "SPY" end = pd.Timestamp.today().normalize(); start = end - pd.DateOffset(months=36) def pull(t): df = yf.download(t, start=start, end=end, interval="1d", auto_adjust=True)[["Close"]] return df.reset_index().rename(columns={"Close":t}) df = pull(PORT).merge(pull(HEDGE), on="Date", how="inner") df = df.rename(columns={PORT:"PortfolioClose", HEDGE:"HedgeClose"}) df.to_csv("prices_clean.csv", index=False)

Tableau Steps (calculated fields)

Portfolio Return: ( SUM([PortfolioClose]) / LOOKUP(SUM([PortfolioClose]), -1) ) - 1

Hedge Return: ( SUM([HedgeClose]) / LOOKUP(SUM([HedgeClose]), -1) ) - 1

Parameter: Hedge Ratio (0.00–1.50, step 0.10)

Hedged Return (Static): [Portfolio Return] - [Hedge Ratio] * [Hedge Return]

Rolling Beta (90d): WINDOW_COVAR([Portfolio Return],[Hedge Return], -89, 0) / WINDOW_VAR([Hedge Return], -89, 0)

Beta Clamped: MIN(1.5, MAX(0, [Rolling Beta (90d)]))

Hedged Return (Dynamic): [Portfolio Return] - [Beta Clamped] * [Hedge Return]

Equity (per series): 100 * EXP(RUNNING_SUM(LOG(1+[Return Series])))

Drawdown (per series): [Equity] / WINDOW_MAX([Equity]) - 1

KPIs (per series): Ann Return, Ann Vol, Max DD, Sharpe = Return/Vol. Compute Using: Date.

Implementation (Tableau‑ready)

Purpose: exact fields and settings so your Tableau workbook works on first try.

0) Expected columns

Date (continuous, ascending)

PortfolioClose (your chosen asset’s close)

HedgeClose (hedge benchmark’s close)

1) Daily returns

[Portfolio Return] = ( SUM([PortfolioClose]) / LOOKUP(SUM([PortfolioClose]), -1) ) - 1 [Hedge Return] = ( SUM([HedgeClose]) / LOOKUP(SUM([HedgeClose]), -1) ) - 1

Compute Using: Date (table down). First row null is expected.

2) Rolling beta (90‑day)

[Rolling Beta (90d)] = WINDOW_COVAR( [Portfolio Return], [Hedge Return], -89, 0 ) / WINDOW_VAR ( [Hedge Return], -89, 0 )

Place on Rows vs Date. Compute Using: Date. Optional regime color: IF [Rolling Beta (90d)] > 1.2 THEN "High (>1.2)" ELSE "OK" END

3) Strategy returns

[Hedged Return (Static)] = [Portfolio Return] - [Hedge Ratio] * [Hedge Return] [Hedged Return (Dynamic)] = [Portfolio Return] - [Rolling Beta (90d)] * [Hedge Return]

Create a numeric parameter Hedge Ratio (default 1.0), show parameter control.

4) Equity indexing (per series)

Equity = 100 * EXP( RUNNING_SUM( LOG( 1 + [Return Series] ) ) )

PortEquity → [Return Series] = [Portfolio Return]

StaticEquity → [Return Series] = [Hedged Return (Static)]

DynamicEquity → [Return Series] = [Hedged Return (Dynamic)]

Compute Using: Date. Ensure 1 + [Return Series] > 0.

5) What your dashboards should show

D1 Overview: 3 equity lines (indexed to 100).

D2 Rolling Beta: β line with regime highlight (>1.2 = red).

D3 Static vs Dynamic: equity lines + KPIs; call out wins/losses.

D4 Drawdown: 3 drawdown lines via [Equity]/RUNNING_MAX([Equity]) − 1.

D5 Rolling Vol & Sharpe: 30‑day stdev and return/σ per series.

D6 Decision: one clincher chart + three bullets you can defend.

D7 Worst 5‑day loss: Roll5d = WINDOW_SUM([Ret], −4, 0); chart the worst per series.

6) Common mistakes

“Calculation contains errors” → add SUM() when using WINDOW_* with base fields.

Flat line at 100 → table calc not set to Date.

Nulls across chart → missing closes or Date not ascending.

Deliverables

Upfront EDA Dashboard — Monthly Return Heatmap, Box Plot by Month, Histogram, Rolling 90‑day Correlation Heatmap (captions answer EDA questions).

Tableau Storyboard (5–7 story points = 5–7 dashboards). Set Story as default tab; turn off “Show Sheets as Tabs”. Export a PowerPoint (.pptx) of the Story.

Required Dashboards (used as story points):

D1 — Overview — Equity curves: Unhedged vs Static vs Dynamic; Hedge Ratio visible

D2 — Rolling Beta — 90‑day beta line + callouts when beta < 0.3 or > 1.2

D3 — Static vs Dynamic — KPIs table + one example where dynamic helps/hurts

D4 — Sensitivity (Static) — ratios 0.5, 0.8, 1.0, 1.2 (small multiples or parameter view)

D5 — Stress Windows — pick 2–3 shock windows; per‑window bars or mini‑curves

D6 — Decision — one key chart + 3 bullets: Benefit, Cost, Limitation

D7 — Optional — Worst 5‑day loss per strategy (bar chart)

Workbook — .twbx (Tableau Packaged Workbook) ONLY; no Tableau Public links. The .twbx must contain all dashboards and the Story and open offline.

Data file — prices_clean.csv (Date, PortfolioClose, HedgeClose)

Write‑up — 1,500–2,000 words covering context, definitions, method, KPIs, sensitivity, stress windows, decision, and ≥3 credible references

Grading Rubric (100 pts)

Component Pts Evidence

Live data pipeline 15 Working Sheet/script; correct 36‑month window.

Clean CSV 10 Correct columns; aligned dates; no blanks.

Upfront EDA visuals 15 Heatmap, box plot, histogram, rolling correlation heatmap; captions answer EDA questions.

Tableau calcs & settings 25 All fields; Compute Using = Date; Rolling window = 90; parameter visible.

Dashboard quality 15 Readable lines; % formatting; consistent legends; coherent design.

Storyboard (5–7 dashboards as story points) 10 Clear titles; concise captions; logical flow; Story default; PowerPoint (.pptx) exported; .twbx submitted.

Write‑up (1,500–2,000 words) 10 Context & definitions, method justified, sensitivity & stress test, numeric evidence, ≥3 credible references.

© 2025 Professor Dr. Teik Kheong Tan. Educational use by enrolled students only. Redistribution or re‑hosting without written permission is prohibited. Last updated: Nov 5, 2025, 23:00 SGT

Component	Pts	Evidence
Live data pipeline	15	Working Sheet/script; correct 36‑month window.
Clean CSV	10	Correct columns; aligned dates; no blanks.
Upfront EDA visuals	15	Heatmap, box plot, histogram, rolling correlation heatmap; captions answer EDA questions.
Tableau calcs & settings	25	All fields; Compute Using = Date; Rolling window = 90; parameter visible.
Dashboard quality	15	Readable lines; % formatting; consistent legends; coherent design.
Storyboard (5–7 dashboards as story points)	10	Clear titles; concise captions; logical flow; Story default; PowerPoint (.pptx) exported; .twbx submitted.
Write‑up (1,500–2,000 words)	10	Context & definitions, method justified, sensitivity & stress test, numeric evidence, ≥3 credible references.