Why We Didn’t Go Aggressive in a “Risk-On” Small-Cap Market: An Empirical Test of Portfolio Concentration

Should risk-on markets mean aggressive concentration? We tested 8–10 stock portfolios across regimes using live and backtested data. The result surprised us: concentration destroyed Sharpe. Here’s what the data actually says.


The Question We Needed to Answer

We received a sharp and fair critique of our SmallCap Catalyst strategy:

“This is a high-quality, capital-preserving small-cap portfolio that is labeled as aggressive but behaves defensively. It avoids mistakes well, but is less convincing at exploiting opportunity when conditions are favorable.”

At face value, the critique pointed to a regime mismatch:

  • The system declared RISK_ON
  • But the portfolio remained diversified, capped, and conservative

The implicit assumption was clear:

If the regime is RISK_ON, shouldn’t the portfolio be more concentrated and aggressive?

Rather than debate this philosophically, we chose to test it empirically.


The Hypothesis

We formalized the question as a falsifiable hypothesis:

When regime conditions are favorable and conviction quality is high, concentrated portfolios (8–10 stocks) should deliver higher risk-adjusted returns than diversified portfolios (15+ stocks).

If this were true, our defensive posture in RISK_ON would indeed be a design flaw.

If false, the defensiveness would be correct—even in favorable regimes.


How We Tested It

We ran a controlled validation using the production version of our system.

Test design highlights:

  • Universe: NSE SmallCap 250
  • Period: Nov 2022 – Nov 2025 (12 quarters) + latest live RISK_ON quarter
  • Rebalance: Quarterly
  • Same conviction scorer, same regime detection, same data
  • Portfolio sizes tested: 5, 8, 10, 12, 15, 20, 25 stocks
  • Sizing methods: Equal-weight, score-weighted, conviction-tiered
  • Metrics evaluated:
    • Information Coefficient (IC)
    • Top-N vs average performance
    • Sharpe ratios across concentration levels
    • Concentration premium (Sharpe of ≤10 stocks vs ≥15 stocks)

This was not a parameter tweak. It was a structural test of whether concentration is earned.


What We Expected to See

For aggressive concentration to make sense, three conditions must hold:

  1. Ranking quality must be strong
    IC should be meaningfully positive (>0.05)
  2. Top-ranked stocks must outperform
    Top-N returns should exceed the portfolio average
  3. Concentration should improve Sharpe
    Not just returns, but risk-adjusted returns

If these conditions were met, concentrated RISK_ON portfolios would be justified.


What the Data Actually Showed

1. Ranking Quality Was Weak — and Sometimes Inverted

Across regimes:

  • NEUTRAL regime: IC ≈ +0.02 (noise)
  • RISK_ON regime (latest quarter): IC ≈ −0.06 (inverted)
  • RISK_OFF regime: IC weak and inconsistent (limited data)

In the most important case—RISK_ON—higher conviction scores were associated with worse forward returns.

This is not a “needs more data” result.
It is a do not scale exposure result.


2. Top Picks Underperformed the Average

Consistently:

  • Top-5, Top-10, and Top-15 stocks underperformed the portfolio average
  • In the live RISK_ON quarter, top-decile stocks lost significantly more than bottom-decile stocks

This means:

  • The model does differentiate stocks
  • But the differentiation does not map to future returns

Differentiation ≠ prediction.


3. Concentration Destroyed Risk-Adjusted Returns

When we compared portfolio sizes:

  • 15-stock equal-weight portfolios had the highest Sharpe
  • Portfolios with ≤10 stocks suffered a ~0.14 Sharpe penalty
  • This result held across sizing methods

In other words:

Concentration did not unlock upside. It amplified ranking errors.

The Key Insight (And the Hard One)

The original critique assumed that opportunity existed at the top of the ranking and was being diluted by diversification.

The data showed the opposite:

There is no exploitable right tail to concentrate into—yet.

The portfolio is defensive not because the designers are timid, but because the signal demands it.


What This Changed in Our Thinking

1. Concentration Is Not a Style Choice

It is a statistical privilege.

Without sufficient IC, aggressive sizing is not conviction—it is leverage on noise.


2. “Risk-On” Does Not Automatically Mean “Aggressive”

In our system:

  • RISK_ON means more signals are enabled
  • It does not mean the system has earned the right to concentrate capital

Those are separate decisions.


3. Portfolio Construction Was Not the Bottleneck

The empirically optimal configuration was:

  • ~15 stocks
  • Equal-weight
  • Controlled concentration

The real bottleneck is ranking efficacy, especially in favorable regimes.


What We Are Not Doing (Yet)

  • We are not implementing aggressive 8–10 stock RISK_ON portfolios
  • We are not increasing position caps
  • We are not forcing aggression to “look risk-on”

Doing so would knowingly destroy value.


What We Are Doing Next

  1. Making concentration a gated capability
    Aggressive construction will activate only if ranking IC exceeds a hard threshold consistently.
  2. Refocusing research on signal quality
    Improving factor selection, interactions, and regime-specific behavior—especially where IC inverted.
  3. Improving exit mechanics
    Progressive exposure decay instead of binary exits—this critique remains valid and independent of concentration.

The Bigger Lesson (For PMs and Allocators)

A portfolio that refuses to become aggressive in favorable markets is not necessarily flawed.

It may simply be honest about what its signals can and cannot do.

“Aggression is not a mandate. It is something a model must earn.”

Our takeaway was not that concentration is bad—but that it is premature.

And learning that before deploying capital is exactly what quantitative research is supposed to do.


Subscribe to RevDog Quantitative Research

Don’t miss out on the latest issues. Sign up now to get access to the library of members-only issues.
jamie@example.com
Subscribe