Skip to content
MEDGE
MEDGECapital

Stress Lab

Stress test your portfolio against the scenario you actually fear.

Set the shocks yourself — on indices, sectors, countries, rates, commodities or currencies — and MEDGE moves the rest of the market the way it has historically moved with them. You see the portfolio, every holding against its index, sector and country, the past episodes that looked most like your scenario, and which assets inside or outside the portfolio would have held up.

Build a scenario Read the method

How it works

1. Pick a market type and a horizon

Bear, neutral or bull decides which days the relationships between assets are estimated on — correlations rise in weak markets and hedges behave differently. Horizon from one week to one year.

2. Set the shocks you care about, one at a time

Drag any of 45 benchmarks — indices, US sectors, countries, styles, rates and credit, commodities, currencies — or one of your own holdings. Every benchmark you leave alone moves to where its history with your shocks puts it, and you can override any of them.

3. Read the portfolio and every holding

Expected portfolio move with a band, where the result comes from shock by shock, each holding next to its index, sector and country peers, and the historical windows that looked most like your scenario.

4. Change the mix and watch the outcome move

Adjust weights or add a benchmark or stock from outside the portfolio; the scenario result recalculates instantly in the browser.

Why adding up betas gets scenarios wrong

The quick way to stress test is to estimate the portfolio’s beta to each factor separately, multiply by each shock and add the results. It breaks as soon as two shocks are correlated. An S&P 500 fall and a Nasdaq 100 fall are, for the most part, the same market decline: the portfolio’s beta to each already contains it, so summing them counts the decline twice.

The Stress Lab treats all the shocks as one joint observation. For any asset a and the set of shocked benchmarks S with shocks s, the expected move is the conditional expectation of a multivariate normal:

μa = ΣaS ΣSS−1 s   ·   residual variance = h·(σ²a − ΣaS ΣSS−1 ΣSa)

It is the same answer a multiple regression on all shocked benchmarks together would give, so correlated shocks share their effect. Benchmarks you leave alone get their implied move, the portfolio band comes from the residual covariance, and the relationships are estimated on the market type you chose (weakest, middle or strongest third of market conditions), blended with the full history so a handful of unusual days cannot drive the result.

What you get from one scenario

  • •Portfolio expected move with an approximate 90% band, and the same shocks evaluated in bear, neutral and bull conditions side by side.
  • •Where the result comes from, shock by shock — correlated shocks share their effect instead of stacking.
  • •A plausibility reading (how unusual the combination is) and a coherence check that flags shocks contradicting each other.
  • •Each holding next to its closest index, its sector ETF and sector peers, its country ETF, and the assets it moves most like.
  • •Per-holding strengths and weak spots against its index: beta in bear vs bull phases, up/down capture, and the index’s worst months in which it gave way.
  • •The five historical windows most similar to your scenario, with what each holding and the portfolio actually did.
  • •A ranking of what holds up in the scenario — inside and outside the portfolio — with its unexplained risk and how much it diversifies what you already own.
  • •A live weight editor: add a candidate at 5% or rebalance, and the outcome updates instantly. Scenarios can be saved in your browser.

45 benchmarks you can shock

Indices

S&P 500 (SPY) · Nasdaq 100 (QQQ) · Russell 2000 (IWM) · Dow 30 (DIA) · MSCI ACWI (ACWI) · MSCI EAFE (EFA) · MSCI Emerging (EEM)

US sectors

Technology (XLK) · Semiconductors (SMH) · Communication (XLC) · Cons. discretionary (XLY) · Cons. staples (XLP) · Financials (XLF) · Health care (XLV) · Industrials (XLI) · Materials (XLB) · Energy (XLE) · Utilities (XLU) · Real estate (XLRE)

Countries & regions

Eurozone (EZU) · Germany (EWG) · United Kingdom (EWU) · Japan (EWJ) · China large-cap (FXI) · India (INDA) · Brazil (EWZ) · Canada (EWC)

Styles

US growth (VUG) · US value (VTV) · Momentum (MTUM) · Min volatility (USMV)

Rates & credit

Treasury 1–3y (SHY) · Treasury 7–10y (IEF) · Treasury 20y+ (TLT) · TIPS (TIP) · US aggregate bonds (AGG) · IG corporate (LQD) · High yield (HYG)

Commodities

Gold (GLD) · Silver (SLV) · Crude oil (USO) · Broad commodities (DBC)

Currencies

US dollar (UUP) · Euro (FXE) · Japanese yen (FXY)

Your own holdings and any tickers you add as candidates can be shocked too.

11 starting points — edit freely

Broad equity bear market

bear · hypothetical

US equities −25% over six months; everything else follows from the estimated relationships.

Tech & semis unwind

bear · hypothetical

Semiconductors −35% and tech −25% while the rest of the market is left to the model.

Credit event

bear · hypothetical

High yield −12%, financials −20%, Treasuries +6% as a flight to quality.

COVID crash (Feb–Mar 2020)

bear · historical

Rounded moves from 19 Feb to 23 Mar 2020.

Rates up, duration sold

neutral · hypothetical

Long Treasuries −15% with a firmer dollar; equities left to the model.

Oil supply shock

neutral · hypothetical

Crude +35% and energy equities +15%.

China slowdown

neutral · hypothetical

Chinese large caps −25%, broad commodities −12%.

Dollar surge

neutral · hypothetical

Dollar index +8%, emerging markets −15%.

Soft landing

bull · hypothetical

US equities +15% with Treasuries +5%: stocks and bonds rally together.

AI melt-up

bull · hypothetical

Semiconductors +40%, Nasdaq 100 +25%.

Reflation & rotation

bull · hypothetical

Value +15%, small caps +18%, commodities +12%, long Treasuries −6%.

“Historical” ideas use rounded moves from the episode named; the others are hypothetical and illustrative.

Limits, stated plainly

  • Relationships are linear and estimated from the price history available for each asset.
  • Normal tails understate extreme moves; the plausibility reading is a guide, not a probability of the future.
  • Correlations can change abruptly in a crisis — the very situation being tested. Comparing with the historical analogues shows how stable they were.
  • Results are illustrative. Nothing on this page or in the tool is investment advice.

Frequently asked questions

What is a portfolio stress test?
A stress test estimates how a portfolio would behave under a specific adverse scenario — a market crash, a rate shock, an oil spike. Historical stress tests replay a past crisis window; scenario stress tests let you define the shock yourself and propagate it through the relationships between assets.
How is this different from replaying 2008 or 2020?
Historical replay tells you what happened to a portfolio in one specific episode, with that episode’s particular mix of moves. A custom scenario lets you ask about the combination you are actually worried about — for example semiconductors down 35% with the rest of the market left to its usual relationships. MEDGE does both: crisis replays live in the Backtesting section, and the Stress Lab also lists the historical windows most similar to your scenario so you can compare the model with what really happened.
Why not just multiply each shock by the portfolio’s beta and add them up?
Because correlated shocks overlap. If the S&P 500 falls 30% and the Nasdaq 100 falls 35%, most of the Nasdaq move is the same market decline; adding a beta-times-shock term for each counts it twice. MEDGE treats the shocks jointly — the conditional expectation of every other asset given all the shocks together — so correlated benchmarks share the effect instead of stacking it.
Can I shock only a sector, a country or a single stock?
Yes. Any of the 45 benchmarks can be shocked on its own, and so can any holding in the portfolio or any candidate ticker you add. Everything you do not set is estimated from its historical relationship with what you did set.
Does the tool tell me if my scenario is realistic?
It reports how unusual the combination is under the model (a Mahalanobis distance with its probability) and flags shocks that contradict each other — for instance a large S&P 500 fall with a Nasdaq rally — showing what the other shocks would imply instead. Both use normal tails, which understate real-world extremes, and the page says so.
Is this a forecast?
No. It is a model of how assets have tended to move together, applied to a scenario you choose. Relationships are linear and estimated from the selected history, and they can break in exactly the crises you are testing. Results are illustrative and not investment advice.
Do I need to connect a brokerage account?
No. You work from tickers and weights. MEDGE never asks for brokerage credentials.

Related

What happens to your portfolio if tech falls 30%?

Paste your tickers, set the shock, read every holding. No brokerage connection. The Stress Lab is included from the Essential plan.

Try the Stress Lab