Quantitative analysis (lens)

Definition

The quantitative lens evaluates instruments and strategies mathematically, statistically, and systematically: hypotheses are expressed as testable rules over data, evaluated over samples, and judged by out-of-sample behavior rather than narrative plausibility. It is both a source of theses (factor and statistical strategies) and the platform’s method for grading every other lens’s claims.

How it works / structure

• Inputs: any structured data — prices, fundamentals, positioning, macro series — plus a precise rule definition and a sample.
• Core operations: hypothesis formulation, backtesting with explicit costs, cross-validation / out-of-sample splits, significance assessment adjusted for multiple testing, and ongoing live-vs-backtest monitoring.
• Output shape: distributions and test statistics, not narratives — expected return, volatility, drawdown, hit rate, turnover, capacity, and the confidence attached to each.
• Relationship to the engine: this lens’s discipline (parameterized rules, golden vectors, replay) is the shape the simulation engine executes; management strategies (pillar 5) and indicators (pillar 6) are written to be quantitatively testable by construction.

When it applies

Whenever a claim can be stated precisely enough to test — which the platform’s falsifiable-thesis format requires. It is the primary lens for systematic strategies (momentum, mean reversion, pairs, carry) and the referee lens for everything else: a fundamental or sentiment thesis becomes gradeable when restated quantitatively.

Risk profile & failure modes

• Overfitting / multiple testing: search enough parameters and something will backtest well by chance; Harvey, Liu and Zhu (2016) argue most published factors fail appropriately raised significance bars.
• Post-publication decay: McLean and Pontiff (2016) measure substantial average decay of predictors after publication.
• Regime breaks: stationarity assumptions fail exactly when it is most costly (crisis correlations, volatility spikes).
• Look-ahead and survivorship bias: silent data errors that manufacture fake edges.
• Capacity and costs: paper edges that do not survive spread, slippage, and market impact at size.

Evidence & limits

The factor literature is the lens’s best-documented output: value and size (Fama and French 1993) and momentum (Jegadeesh and Titman 1993) are the canonical cross-sectional results, replicated across markets and decades — and still debated as risk premia vs mispricing, with attenuation after publication (McLean and Pontiff 2016). The meta-finding matters as much as any factor: Harvey, Liu and Zhu (2016) estimate that a large share of published findings are false positives under multiple-testing corrections. Quantitative method does not guarantee edge; it guarantees that the absence of edge is detectable.

Falsifiable-thesis examples

Illustrations only, not signals:

• “A 12-1 month momentum rank portfolio over universe U will have a positive information ratio vs the universe average over the next 126 trading days” — falsified by a non-positive realized IR.
• “The 60-day realized correlation between X and Y will stay above 0.5 for the next quarter” — falsified by any 60-day window below 0.5 in that span.

Cross-references

• Strategies graded natively by this lens: strategy-momentum, strategy-mean-reversion, strategy-pairs-trading, strategy-futures-trend-following
• Risk machinery: risk-kelly-criterion, risk-volatility-targeting, risk-scenario-analysis
• Adjacent lenses: lens-technical (rule inputs), lens-fundamental (factor inputs), lens-risk (distributional outputs)