Home/Systematic Trading/Factor Construction

Factor Construction

factor-investingLevel 2 — Intermediate

Prerequisites

Factor Investing & Quant EquityEdge Identification

Unlocks

Factor Risk Model Construction

What It Is

The technical process of converting a raw investment insight into a tradeable systematic factor — including signal definition, universe construction, weighting scheme, rebalance frequency, and whether to implement as long/short or long-only.

Correct Execution

Practitioner starts with an economic hypothesis for why the factor should generate returns, not with a data-mining exercise. The factor is defined at a point-in-time (no look-ahead bias) using only variables available at the decision date. Long/short construction isolates the pure factor signal; long-only construction mixes factor exposure with market beta. Rebalance frequency is set based on signal half-life (how quickly the signal decays) rather than operational convenience. Factor is tested for persistence across regimes and market structures, not just in aggregate.

Progression Levels

Diagnostic Tree

Coaching Cues

  • "A factor is pervasive, persistent, and interpretable. If it fails any of those tests, it's a theme." — Giuseppe Paleologo
  • "The construction should come from the hypothesis, not from the backtest result."
  • "Capacity kills more factors than bad signals do. Price it in before you launch."
  • "Long/short factor is the pure signal. Long-only factor is the signal plus market beta. Know what you're buying." — Paleologo framework

Common Errors

  1. Mining factor specifications to maximize backtest returns: Choosing the construction that works best historically guarantees over-fit. Specification should come from economic reasoning, not optimization.
  2. Ignoring capacity in factor construction: A factor that works on a $10M portfolio may fail at $100M due to market impact. Capacity analysis is a required step, not optional.
  3. Confusing themes with factors: AI exposure, geopolitical positioning, ESG — these are themes with limited universe and non-persistent behavior. They should not be treated as factors in a systematic model.
  4. Not testing factor orthogonality: Building a "new" factor that is 90% correlated with momentum adds no independent information. Every new factor should be orthogonality-tested against the existing model.

Edges

⚡ Conventional Wisdom Is Wrong

Academic Factor Definitions Are Hypotheses, Not Production Blueprints

factor-investingfactor-construction →

Academic factor papers are designed to prove statistical existence of a phenomenon — they use monthly rebalancing, equal weighting, full universe, and zero transaction costs because that maximizes signal detection power. These are not production-ready constraints. A factor that shows 8% annualized alpha in an academic paper may generate 2% or less in production because academic construction methods are not designed for tradability.

What most people do
Find an academic paper showing a factor works. Implement the paper's exact construction specification in a production system. Attribute live underperformance to "the factor decaying."
What the best do
Treat the academic paper as a hypothesis that a phenomenon exists. Rebuild the factor from scratch with production constraints: liquidity filters, realistic market impact costs, capacity-appropriate universe, point-in-time data. Evaluate the production factor independently of the academic specification.
Why it's an edge: Academic constructions exist in a frictionless world; production factors exist in the real world. The gap between them is systematically exploited by practitioners who rebuild from first principles.
How to exploit: For any academic factor being evaluated, rebuild it with: (1) top-quintile liquidity filter (only tradeable securities); (2) transaction cost of 10-30bps per round trip; (3) market impact model for intended position sizes; (4) capacity ceiling. If the factor remains positive after these adjustments, it may be production-viable. If not, the academic result is not exploitable at your scale.
Giuseppe Paleologo, "Quant Investing at Multi-Strat Hedge Funds," Odd Lots, 2025-06-23
⚡ Conventional Wisdom Is Wrong

Themes Fail The Pervasive-Persistent-Interpretable Test

factor-investingfactor-construction →

Systematic investors routinely mistake themes (AI, tariffs, geopolitics, ESG) for factors. The distinction is empirically testable: a factor must be (1) pervasive — affecting every asset in the universe, not just a sector; (2) persistent — not a temporary phenomenon; (3) interpretable — traceable to an economic mechanism. AI is a theme: it affects only tech-adjacent names (not pervasive), it changes definition year to year (not persistent), and "companies that benefit from AI" is circular, not an economic mechanism (not interpretable).

What most people do
Identify a market narrative with momentum and label it a "factor." Build systematic strategies around themes. Attribute returns to factor exposure when the narrative was the true driver.
What the best do
Apply the three-criteria test to every candidate signal before calling it a factor. Themes fail at least one test; factors pass all three. Themes can be traded opportunistically as high-conviction, short-duration positions — but never sized or managed as structural factor exposures.
Why it's an edge: Prevents systematic exposure to theme-driven positions during narrative reversals, which produce large drawdowns in "factor" strategies that were actually theme bets.
How to exploit: Build a two-column evaluation sheet for every new signal: (1) pervasive? (show it loads on >50% of assets in universe), (2) persistent? (show positive signal for 5+ years across multiple regimes), (3) interpretable? (state the causal mechanism without using the theme name). If the signal fails any test, classify it as a theme with a hard position limit and time horizon.
Giuseppe Paleologo, "Quant Investing at Multi-Strat Hedge Funds," Odd Lots, 2025-06-23
🔑 Hidden Causal Lever

Capacity Kills More Factors Than Bad Signals Do

factor-investingfactor-construction →

A factor that works at $10M typically fails at $100M not because the signal decays but because market impact at scale consumes the expected return. Most factor research is conducted at small notional sizes where impact is negligible, producing optimistic estimates that cannot be achieved at target AUM. Capacity analysis is a required step in factor evaluation, not optional — and it typically produces the most sobering results.

What most people do
Evaluate factor quality based on backtest Sharpe and historical returns at the notional scale of the research. Scale up the strategy based on asset raise without re-evaluating capacity.
What the best do
Calculate the strategy's maximum capacity ($ AUM where market impact costs consume >50% of expected excess return) as part of initial factor evaluation, before any live deployment. Only deploy if capacity significantly exceeds target AUM.
Why it's an edge: Prevents the common pattern of a successful small-scale strategy failing after scaling — which is interpreted as factor decay when it is actually impact-cost erosion.
How to exploit: For any factor under evaluation, calculate: expected annual return × estimated daily volume of universe × desired turnover = capacity estimate. Require capacity > 5× target AUM before proceeding. For strategies where capacity is borderline, build market impact simulation into the backtest explicitly.
Giuseppe Paleologo, "Quant Investing at Multi-Strat Hedge Funds," Odd Lots, 2025-06-23

Sources

  • Giuseppe Paleologo, "Quant Investing at Multi-Strat Hedge Funds," Odd Lots, 2025-06-23 — factor definition, pervasive/persistent/interpretable criteria, themes vs. factors
  • Giuseppe Paleologo, "Multi-Manager Hedge Funds" (Flirting with Models, S7E11), 2024-09-02 — factor model construction, returns vs. characteristics, integration approaches
  • Euan Sinclair, "Positional Option Trading" (Flirting with Models, S3E12), 2021-04-10 — signal construction rigor applied to options factors