Building a Systematic Small-Cap Catalyst Model Using Hybrid Quant and AI Portfolio Intelligence

A data-driven, multi-factor framework for navigating India’s small-cap markets with discipline, robustness, and AI-guided risk control.

A data-driven, multi-factor framework for navigating India’s small-cap markets with discipline, robustness, and AI-guided risk control.

Small-caps are a paradox: they offer some of the highest long-term returns in public markets, yet they are also the most unforgiving during drawdowns. Traditional discretionary investing struggles with this volatility; meanwhile, pure quantitative systems often fail in small-caps due to liquidity traps, data noise, and rapid regime shifts.

To address this, I built the Smallcap Catalyst Model—a hybrid quant + AI portfolio intelligence system that systematically identifies high-quality small-cap opportunities while enforcing institutional-grade risk management.

This post breaks down the model, backtest performance, portfolio intelligence layer, failure modes, robustness tests, and deployment considerations.


Strategy Essence

A disciplined, multi-factor catalyst engine applied to the NSE SmallCap 250 universe, augmented with AI-driven risk oversight.

What the System Does

Universe:
NIFTY SmallCap 250 (refreshed monthly)

Scoring Engine (7 Fundamental Catalysts):
Weighted composite score (0–10):

FactorWeightPurpose
Revenue Growth YoY20%Detect top-line acceleration
PAT Growth YoY20%Validate operational efficiency
Operating Margin Trend10%Capture profitability improvement
ROE10%Reward capital efficiency
Debt/Equity10%Penalize leverage risk
Promoter Holding Change20%Insider conviction proxy
FII Holding Change10%Institutional flow signal

Technical Entry Timing Filter:
Buy only if price is ≤15% above 200DMA (avoids euphoria).

Portfolio Rules:

  • Entry: Buy stocks score ≥ 4.0
  • Exit: Hard stop at -15%; hold winners
  • Rebalance: Quarterly
  • Typical portfolio size: 5–7 names
  • Sector cap: Max 2 stocks per sector
  • Liquidity filters: Min ₹1Cr daily volume; position <5% daily volume

Performance (Backtest: Jul 2023 – Nov 2025)

MetricStrategyBenchmark (Nifty Smallcap 250)Alpha
CAGR37.15%~18%+19.15%
Total Return88.17%~40%+48.17%
Sharpe Ratio1.060.60+0.46
Sortino Ratio2.44~1.0+1.44
Calmar Ratio1.93~0.7+1.23
Max Drawdown-19.20%-25%Better
Win Rate75%
Turnover0.1/quarterLow

Summary:
A high-CAGR, moderate-risk, tax-efficient small-cap engine that materially outperforms the benchmark.


1. Strategy Overview

A multi-factor model layered with AI-driven risk intelligence.

AttributeDetails
NameSmallcap Catalyst Strategy
TypeSystematic fundamental + flow model
UniverseNSE SmallCap 250
RebalanceQuarterly
Holding Period90–120 days; target >12 months for LTCG
Portfolio Size5–7 stocks
ObjectiveCapture fundamental catalysts and institutional flows early

Unlike large-cap quant models, which rely on factor orthogonality, small-caps require catalyst-based factor construction—growth, margins, flows—with dynamic weights.


2. Performance Summary

A. CAGR, Risk, Efficiency

MetricValueInterpretation
CAGR37.15%Strong alpha capture
Sharpe1.06Good risk-adjusted returns
Sortino2.44Excellent downside management
Max Drawdown-19.2%Controlled for small-caps
Turnover0.1Tax-efficient

B. Robustness Across Start Dates

Stress-testing the strategy starting at six random dates:

Start DateCAGRMax DDVerdict
Jul 2023+38.8%-19.2%Excellent
Oct 2023+38.8%-19.2%Excellent
Jan 2024+1.2%-38.6%Marginal
Apr 2024+11.8%-17.4%Good
Jul 2024-4.6%-26.0%Poor
Oct 2024-5.5%-31.7%Poor

Mean CAGR: +13.4%
Worst-case CAGR: -5.5% during the chokepoint mini-bear of late-2024.

Insight

The strategy is highly regime-sensitive—it shines in trending or stable environments but requires risk overlays in volatile or declining markets.


3. Catalyst Engine Architecture

A composite score capturing growth, profitability, balance sheet strength, and ownership flows.

Growth (40%)

  • Revenue Growth YoY
  • PAT Growth YoY

These two explain the majority of dispersion among small-cap winners.

Profitability (30%)

  • OPM Trend
  • ROE
  • ROCE Trend

Improving margins signal operating leverage and pricing power.

Balance Sheet (20%)

  • Debt/Equity
  • Promoter Holding Change

Promoter increases consistently outperform promoter decreases.

Institutional Flow (10%)

  • FII stake change

Foreign flows matter disproportionately in small and mid caps.


4. Technical Entry Filter

200DMA Filter: Only buy if stock is ≤10–15% above the long-term trend.

ConditionAvg 90D ReturnSignal
≤10% above 200DMA+8.7%Buy
10–20% above+7.0%Wait
>20% aboveNegativeAvoid

Prevents buying euphoric runaway stocks.


5. Portfolio Characteristics

ParameterValue
Avg Holdings6
WeightingEqual-weight
Sector CapMax 2
LiquidityMinimum ₹1Cr/day
Slippage2.5% entry & exit
Holding Period9.3 months avg

Key Insight:
Even with 60% cost drag (slippage + taxes), the strategy still produced 15.15% post-cost CAGR.


6. Factor Exposures

FactorExposureRationale
QualityHighROE, D/E filters
GrowthHighRevenue/PAT growth
MomentumModerateFII flows
ValueNeutralNo PE filters
SizePure Small-capOnly SC250
Low VolNegativeBy definition

The model is structurally geared toward growth + quality + flows.


7. Risk Management Framework

A mix of rule-based quant filters and AI-generated oversight conditions.

Portfolio-Level Controls

  • Max 7 stocks, Min 5
  • Max 2 per sector
  • Position <5% of daily volume
  • Automatic cash allocation if <5 stocks qualify

Position-Level Controls

  • Hard stop-loss at -15%
  • Soft stop-loss: score <4.0
  • Entry slippage baked into modeling
  • No discretionary overrides

Portfolio Intelligence (AI Layer)

AI evaluates price trend, flows, volatility, and drawdown behavior:

TriggerConditionActionStatus
Technical RiskPortfolio avg <10% above 200DMAExitGreen
Drawdown>15% from peakReduce 50%Green
Flow Reversal3 days FII/DII net sellingExitGreen
VIX SpikeVIX >20Trim 25%Green

These overlays help the system avoid regime traps and momentum crashes.


8. Robustness: Monte Carlo Stock Selection Test

To test sensitivity to individual picks, 30 random portfolios were simulated from the same score-ranked universe (all score ≥4.0).

MetricValue
Best Random Portfolio+65.5% CAGR
Worst Random Portfolio+5.4% CAGR
Mean+38.2%
Median+40.0%
Std Dev14.1%

Key Insight:
Every random portfolio was profitable.

This proves:

  1. The catalyst score is genuinely alpha-generative.
  2. Stock selection matters less than disciplined selection + timing.
  3. Entry regime (e.g., Oct 2024) is critical.

9. Equity Curve & Drawdowns

Starting with ₹100,000, the portfolio grew to ₹188,174 over 2 years.

Drawdown Behavior

  • Max DD: -19.20%
  • Recoveries typically within 1–2 quarters
  • Lower DD than benchmark (-25%)

Risk controls clearly improve drawdown resilience.


Verdict

Rating: B- (Strong performance, moderate robustness, sensitive to market regime)

Strengths

  • High CAGR
  • Excellent risk-adjusted returns
  • Factor engine validated statistically
  • Low turnover and tax efficiency
  • Rule-based, scalable, machine-friendly

Limitations

  • Regime-sensitive; struggles in sudden reversals
  • Concentrated portfolio amplifies volatility
  • Buying late-cycle bull markets reduces effectiveness
  • Technical filter essential—without it, drawdowns rise sharply

Forward-Looking Improvements

  • Introduce volatility-adjusted position sizing
  • Add mid-cap stabilizers during high VIX regimes
  • Extend rebalancing to semi-annual for LTCG optimization
  • Add dynamic sector priors based on institutional flow data

Final Thoughts

The Smallcap Catalyst Model demonstrates that systematic small-cap investing is viable, but only with:

  • High-quality factor engineering
  • Intelligent technical filters
  • Strict liquidity and risk constraints
  • AI-guided oversight to manage regime shifts

This hybrid quant + AI framework offers a robust blueprint for small- and mid-sized funds looking to scale disciplined, repeatable, data-driven investing in one of the world’s most dynamic equity segments.


Subscribe to RevDog AI

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