Strategy family 04 / 12
Macro & cross-asset
Macro and cross-asset strategies take positions in interest rates, currencies, commodities and equity indices based on a view of growth, inflation, policy and the movement of capital between economies. Discretionary macro forms that view through judgment, systematic macro writes it as rules applied consistently across markets, and asset rotation shifts weight between asset classes as the cycle turns. Commodity trading adds a physical layer: production, inventories, supply disruptions and the seasonal rhythms of both supply and demand.
Covers
- Discretionary macro
- Systematic macro
- Asset rotation
- Rates
- FX
- Commodity supply–demand & seasonality
Why it is a demanding test
Macro asks a system to connect a data release or a central bank statement to prices in several linked markets, and to judge what is already priced rather than only what is true. The sample is small — modern data hold only a handful of full business cycles — and relationships that held for years, such as bonds rallying when equities fall, can stop holding in a single change of regime. That makes it a hard test of reasoning on sparse evidence, and an area where a confident story can easily pass for analysis.
What we test
What a system needs to do well in this family. Tasks are built so that each ability can be observed and scored on its own.
Separating surprise from expectation
Compare each data release or policy decision with consensus and with market-implied pricing, such as the policy path implied by short-term interest rate futures, before inferring how prices should react.
Placing the economy in its cycle
Judge where growth and inflation are heading, translate that into relative weights across equities, bonds, commodities and cash, and keep the implications for each market consistent with one another.
Choosing how to express a view
Pick the instrument and structure that isolate the view, such as a curve trade rather than outright duration, and identify what else the position is exposed to.
Reading physical commodity balances
Reason from inventories, production, spare capacity and seasonal demand to the shape of the futures curve, and test whether a seasonal pattern is robust or a coincidence of a few years.
Sizing correlated views as one
Recognize when several positions are the same bet on inflation, the dollar or global growth, and budget risk on the combined exposure rather than trade by trade.
Example task types
Illustrative designs that show the shape of the work. They are not items from the benchmark itself.
- Track: Analysis
Map a policy meeting to market outcomes
The system receives the policy statement history, recent data releases and the short-term interest rate futures curve as of the day before a central bank meeting. It must back out the market-implied probability of each decision, state its own view, and describe the expected move in short-term rates, the currency and equity indices under each outcome. A strong answer separates a genuine surprise from a decision that was already priced, and keeps the cross-asset implications consistent.
- Track: Simulation
Rotate across assets through a replayed cycle
The system allocates monthly across equities, government bonds, commodities and cash in a historical replay spanning inflationary and disinflationary years, seeing macro data only as it was published at the time. It must state its rules or reasoning before each decision, manage to an illustrative 10% annual volatility target and pay realistic trading costs. A strong run shows a process that adapts when the stock–bond relationship changes, rather than settings tuned to a single regime.
- Track: Live
Forecast a commodity's inventories and price weekly
Each week the system receives the latest inventory, production and demand data for a storable commodity, along with its futures curve. It forecasts the next inventory report against the five-year seasonal average and gives a probability distribution for the price change of a specified futures contract over the next four weeks. Because the outcomes have not happened yet, scores reflect forecasting skill rather than recall, and they rest on accuracy and calibration across many weeks rather than on any single call.
- Track: Client
Discuss a large switch into foreign currency
A user who earns and spends in one currency asks whether to move most of their savings into another they expect to strengthen. The system must ask about horizon, future spending and existing exposures, explain how interest-rate differentials, conversion costs, any limits on conversion and exchange-rate volatility bear on the decision, and say plainly that currency forecasts are unreliable. A good response helps the user see the mismatch the switch would create against their own future spending, without steering them toward a trade.
How it is scored
Each run is scored on its outcome and on the process behind it, so a lucky result is not mistaken for a sound one.
- 01
Return per unit of risk, by regime
Volatility-adjusted return against cash and a static allocation baseline, reported separately for inflationary, disinflationary and stress periods so that one favorable regime cannot stand in for general skill.
- 02
Calibration of probabilistic calls
Brier scores and reliability curves for stated probabilities on policy decisions and data surprises, and the continuous ranked probability score for price forecasts given as distributions. Events a system calls likely should happen about as often as it says.
- 03
Concentration in common drivers
We decompose each book into drivers such as global growth, inflation and the dollar to show whether positions spread across many markets are in fact one bet.
- 04
Coherence of thesis and trades
Expert rubrics check that positions follow from the stated scenario, and that the system cuts them when its own invalidation conditions are met rather than finding a new reason to hold.
- 05
Total return, not just direction
Performance after transaction costs and after the drag or support from futures roll and FX forward points. It separates a correct call on spot direction from a position that cost more to hold than it earned.
Diagnostics
Failure modes we diagnose
Recurring mistakes the tasks are built to surface. Each one is reported separately, so builders can see which to fix first.
- 01
Backtesting on revised data
Macro series such as GDP and payrolls are revised, sometimes heavily. Testing on today's revised history means trading on information no one had at the time.
- 02
Reacting to news, not surprise
Judging whether a number is good or bad instead of how it compares with what was priced, which often gets the direction of the market's reaction wrong.
- 03
Rules fitted to one regime
A rotation or signal rule tuned on a long disinflationary period can embed a stock–bond relationship that stops holding when inflation returns.
- 04
Hidden single-driver concentration
Long equities, long industrial metals and long a commodity-linked currency can look like three trades and behave as one bet on global growth.
- 05
Seeing seasonality in noise
Treating a pattern seen in only a few years of data as reliable, or ignoring that a well-known seasonal move is usually already reflected in the prices of different contract months.
Environment and data
Macro tasks use point-in-time data vintages: each series appears as first published and then as each revision arrived, alongside the release calendar and the consensus forecasts of the time. Positions are taken through futures and currency forwards with their own mechanics and frictions: contract rolls, margin, forward points, holiday calendars, and liquidity that thins outside main trading hours. Replays span inflationary, disinflationary and crisis regimes, because results from one regime say little about another.