Home/Systematic Trading/Systematic Macro

Systematic Macro

regime-detectionLevel 3 — Advanced

Prerequisites

Regime DetectionTrend FollowingCarry StrategiesEdge Identification

What It Is

Systematic macro combines multiple quantitative signals (carry, momentum, sentiment, curve dynamics, macro fundamentals) across global asset classes to express directional views without discretionary override — translating macro themes into diversified, model-driven portfolios.

Correct Execution

  • Cluster models by what they're actually measuring (run correlation analysis on signal returns) rather than by what they're theoretically supposed to capture
  • Maintain 100+ individual models to avoid idiosyncratic model risk; cluster into 10-20 independent "investment themes"
  • Never call alpha something that's really beta — distinguish directional factor exposure from genuine information advantage
  • Build models at multiple time horizons (fast momentum, slow trend, carry) and let them run simultaneously — do not force one time horizon
  • Cross-sectional thinking is critical: the edge is knowing which signals are working NOW vs on average

Progression Levels

Diagnostic Tree

Coaching Cues

  • "We call our themes 'bin one, bin two' — it's really the clustering and relationship between models that matters, not the label." — Asif Noor, FWM S6E9
  • "Run a clustering algorithm on your models. Whatever comes out, those are your true investment themes. Don't fight the data." — Asif Noor, FWM S6E9
  • "Alpha is a factor nobody else has found yet. Once everyone knows about it, it becomes a factor and the premium shrinks." — Giuseppe Paleologo, Odd Lots, 2025-06-23

Common Errors

  1. Organizing signals by theoretical category rather than empirical correlation: Theoretical category (carry, momentum) does not equal empirical independence → Cluster signals by realized return correlation, not by the label you put on them
  2. Over-labeling factor returns as alpha: Carry in currencies is a well-known factor; if your "alpha" is just unlabeled carry, you will be surprised in the next risk-off event → Run full factor decomposition; own up to factor exposure
  3. Building too few models to achieve true diversification: A program with 5 signals can easily lose 3 of them in a regime change → Build 100+ models and let the clustering reveal independent sources of return

Edges

💎 Elite-Only Behavior

Theoretical Signal Categories Are Wrong — Let the Data Cluster Signals

regime-detectionsystematic-macro →

Systematic macro programs are typically organized by theoretical signal type: momentum, carry, value, sentiment. When you run actual correlation analysis on the signal returns (not the signal definitions), the empirical clusters rarely match the theoretical categories. "Fast momentum 3-month" and "medium trend 6-month" may cluster together; "option-market sentiment" and "equity momentum" may be nearly identical empirically. The theoretical taxonomy creates false confidence in diversification that doesn't exist — and misses genuine independence that does.

What most people do
Organize signal diversification by theoretical category; report "we have momentum, carry, value, and sentiment signals" as evidence of diversification.
What the best do
Run an unsupervised correlation algorithm on all model return series; let the empirical clusters define the investment themes regardless of theoretical label; size by independent risk budget derived from actual correlations.
Why it's an edge: Programs that believe they have 6 independent sources of return based on labels, but actually have 3 correlated clusters, are misallocating risk budget. Programs that size based on empirical independence have genuine diversification.
How to exploit: Take all signal return series (even if you have 50+ models) and run k-means or hierarchical clustering on the pairwise correlations. Note where theoretical categories split or merge. Rebuild your signal bucket framework to match the empirical clusters. Re-weight by cluster, not by signal count.
Cross-domain parallel
In algorithmic trading factor research, academic factor categories (value, momentum, quality) are highly correlated empirically when applied to the same universe. LASSO-based factor selection routinely discovers that 3-4 empirically orthogonal factors replace 10+ theoretical ones.
Asif Noor, "Modern Systematic Macro," FWM S6E9, 2023-06-26 — "take your 110 models and run a simple clustering algorithm. The bins that come out will surprise you."
⚡ Conventional Wisdom Is Wrong

Alpha and Beta Are Regime-Dependent Constructs — The Same Strategy Is Both

regime-detectionsystematic-macro →

Systematic macro programs routinely distinguish "alpha strategies" (short-term, idiosyncratic, proprietary) from "beta strategies" (carry, momentum, well-known factors). This distinction collapses under scrutiny: an "alpha" strategy that works during risk-on and fails during risk-off has hidden directionality — it is really a beta to market regime in disguise. Calling it alpha because it's short-horizon or unlabeled does not remove the latent exposure. Conversely, a well-known factor (trend-following) can generate genuine alpha if you have better execution, better signal calibration, or better portfolio construction than competitors.

What most people do
Label trend/carry/momentum as beta and everything faster-moving or proprietary as alpha; size "alpha" strategies larger because they're supposed to be uncorrelated to market direction.
What the best do
Factor-decompose ALL strategies — including "alpha" strategies — to identify latent directionality; accept that some well-known factors generate genuine returns and size them accordingly rather than inflating returns with hidden beta.
Why it's an edge: Inflating "alpha" with hidden beta leads to overconfidence and under-hedging; discovering the latent exposure before a regime change — not after — is the actual risk management skill.
How to exploit: For any strategy labeled as alpha in your program, compute its conditional returns during three regimes: (1) sustained equity bull, (2) equity bear, (3) rate dislocation. If the strategy performs asymmetrically across regimes in a way consistent with having market beta, it has hidden directionality and should be sized as a partial beta allocation.
Cross-domain parallel
In sports betting, a "sharp action" signal that only works in heavy public-action games is really a market-regime-dependent signal, not pure information. The "alpha" is conditional on a specific market structure.
Asif Noor, FWM S6E9, 2023-06-26; Giuseppe Paleologo, Odd Lots, 2025-06-23
💎 Elite-Only Behavior

100 Models Is Not a Lot — It's the Minimum for True Signal Diversification

regime-detectionsystematic-macro →

Practitioners building systematic macro programs with 5-15 signals believe they are diversified because they have "many different approaches." At 5 signals, the idiosyncratic risk of any single model failure is enormous — if one signal fails in a new regime, you lose 20% of your program. At 100+ models, no single model can cause a material failure. More importantly, 100 models across genuinely different time horizons, asset classes, and signal types is what enables the empirical clustering approach to reveal the true structure of the return space.

What most people do
Build 10-15 "best" signals; focus on model quality over model count; believe signal count beyond 15 is diminishing returns.
What the best do
Build 100+ models including many that are "individually unimpressive" but contribute genuine diversification at the portfolio level; let the clustering reveal which are truly independent.
Why it's an edge: The diversification benefit of adding a mediocre independent signal to a portfolio is mathematically positive even when the standalone Sharpe is near zero. Most practitioners don't build enough models to realize this benefit.
How to exploit: When building a systematic macro program, set a minimum model count target of 50 before evaluating performance. Build fast, medium, and slow versions of every signal type across every liquid market. Prune only models that are genuinely correlated to existing clusters — not models that are individually weak but empirically independent.
Asif Noor, FWM S6E9, 2023-06-26 — "we maintain 100+ individual models to achieve true diversification; then cluster into 14 themes."

Sources

  • Asif Noor, "Modern Systematic Macro," Flirting with Models S6E9, 2023-06-26 — multi-signal systematic macro construction; empirical signal clustering; alpha vs beta distinction
  • Sandrine Ungari, "Alternative Risk Premia," Flirting with Models S3E10, 2021-04-10 — trading premia vs academic premia; macro factor neutrality in ARP strategies
  • Giuseppe Paleologo, "Quant Investing at Multi-Strat Hedge Funds," Odd Lots, 2025-06-23 — factor vs alpha distinction; crowding dynamics in systematic strategies