Prepared for: Heightss Investment Committee | Date: December 9, 2025
1. Introduction: The Epistemological Crisis in Modern Asset Management
The Indian capital markets, historically a bastion of relationship-based investing and fundamental earnings tracking, have undergone a seismic structural shift in the mid-2020s. The proliferation of high-frequency trading (HFT), the democratization of algorithmic tools for retail investors, and the institutional adoption of "Quantamental" strategies have created a bifurcated ecosystem. In this new paradigm, the efficacy of traditional stock picking—rooted in the teachings of Graham and Dodd—is increasingly being interrogated by the raw, unsentimental efficiency of Artificial Intelligence (AI) and Machine Learning (ML) models.
This report documents the results of a rigorous, controlled experiment conducted over a 90-day fiscal quarter, specifically from September 8, 2025, to December 8, 2025. The challenge sought to answer a singular, defining question: In a market characterized by high volatility, corporate restructuring events, and operational "black swans," does the agility of AI-driven momentum scoring outperform the valuation-anchored logic of traditional equity research?
"The challenge sought to answer a singular, defining question: Does the agility of AI-driven momentum scoring outperform the valuation-anchored logic of traditional equity research?"
To scrutinize this, a theoretical capital base of $10,000 (approximately ₹8,40,000) was bifurcated equally into two portfolios. Portfolio A, designated "The Machine," utilized a pure quantitative approach based on momentum scoring, durability ratings, and sentiment analysis algorithms similar to those gaining traction in academic finance. Portfolio B, designated "The Human," was constructed strictly adhering to the "Top Picks" and "Buy" recommendations issued by India’s leading brokerage houses—Motilal Oswal, ICICI Direct, and JM Financial—during the first week of September 2025.
2. Methodology and Portfolio Construction
The integrity of any comparative analysis rests on the rigorous definition of its initial parameters. Both portfolios were initialized on September 8, 2025, effectively capturing the market sentiment post-Q1 FY26 earnings and pre-festive season positioning.
2.1 Portfolio A: The AI-Driven Quantitative Model
The construction of Portfolio A was devoid of qualitative judgment. It operated on the premise that market prices discount all known information and that future price movements are probabilistically dependent on recent trends (Momentum) and financial robustness (Durability).
The Algorithm: The selection process mirrored the sophisticated multi-factor models employed by quantitative funds like the UTI Quant Fund and proprietary desks. The specific logic utilized a composite score derived from:
- Trendlyne Momentum Score: A proprietary metric weighing moving averages, RSI, and volume breakouts. The threshold for inclusion was a score > 70.
- DVM Score (Durability, Valuation, Momentum): High durability scores were prioritized to avoid "junk" rallies, filtering for companies with consistent cash flows and low leverage.
- Sentiment Overlay: Utilizing Long Short-Term Memory (LSTM) networks to parse news sentiment, flagging stocks with rising positive mentions in financial media, ignoring valuation multiples like P/E ratios.
Portfolio A Constituents (Entry: Sept 8, 2025):
2.2 Portfolio B: The Traditional Fundamental Model
Portfolio B represented the consensus view of "The Street." This portfolio was constructed by aggregating high-conviction ideas from top-tier brokerage research reports released in early September 2025. The philosophy here was Growth at a Reasonable Price (GARP) and Value Unlocking.
Portfolio B Constituents (Entry: Sept 8, 2025):
3. The Macroeconomic Crucible: India in Q4 2025
To evaluate the performance divergence, one must first understand the macroeconomic currents that defined the Indian equity landscape between September and December 2025. This period was characterized by a transition from euphoria to caution, driven by both global and domestic headwinds.
The Liquidity Tussle: FIIs vs. DIIs: A defining feature of this quarter was the relentless selling by Foreign Institutional Investors (FIIs), driven by the allure of cheaper valuations in China and high US bond yields. This was, however, aggressively counteracted by Domestic Institutional Investors (DIIs), fuelled by record SIP inflows. This tug-of-war created a market that was range-bound with a negative bias.
4. Deep-Dive Performance Analysis: Portfolio A (AI/Quant)
The AI portfolio's performance offers a fascinating case study in the strengths and weaknesses of momentum investing. By ignoring narratives and focusing on price, the algorithm successfully identified the leaders of the quarter but failed to exit the losers in time due to the lagging nature of some momentum signals.
The Winners: Riding the "Super-Momentum"
Bajaj Finance (+37.5%): The AI Edge. While fundamental analysts debated the compression of Net Interest Margins (NIMs), the AI model detected a divergence: price volume action was bullish despite neutral news flow. The stock's Relative Strength Index (RSI) signaled a reversal from oversold zones.
Bharat Electronics Ltd (+35.4%): The Sector Play. The AI model identified BEL not just on its own merit but as part of a broader sectoral breakout. It effectively piggybacked on the government's indigenization narrative without "knowing" the policy details—it simply tracked the capital flows.
The Losers: The Trap of Past Performance
Trent Ltd (-21.1%): The Valuation Ceiling. The AI fell victim to the "recency bias" inherent in momentum models. It bought at the top of a multi-year rally. As consumption data slowed, the stock's exorbitant P/E multiple contracted.
5. Deep-Dive Performance Analysis: Portfolio B (Fundamental)
The Fundamental portfolio was designed to be robust, utilizing the collective wisdom of India's top equity researchers. However, it suffered from what can be termed the "Consensus Trap"—when a stock is universally liked, the positive news is often already priced in.
The Tata Motors Demerger: A Case Study in Value Complexity (-1.0%)
Brokerages anticipated a re-rating of the PV business (JLR + EV). However, slowing global auto sales and JLR's capital expenditure needs capped the upside. The "sum of the parts" did not become greater than the whole in the short term. Retail investors panic-sold on the price drop, creating volatility that fundamental targets could not cushion.
The Swiggy IPO: The Lock-in Liquidity Shock (-6.0%)
Swiggy was a consensus "Buy". However, following its IPO, Swiggy faced the expiry of the lock-in period for pre-IPO investors. Early investors exited to book profits, overwhelming demand. Analysts focused on revenue growth while the market shifted focus to profitability.
6. P&L Attribution and Comparative Metrics
7. Lessons from the Field: The "Quantamental" Future
The underperformance of Portfolio B underscores a critical flaw in traditional brokerage research: The Consensus Lag. By the time a stock like NTPC or Bank of Baroda becomes a consensus "Buy," the smart money has often already positioned itself.
Conversely, while Portfolio A won, it was not without peril. The case of Trent shows that momentum works until it doesn't. An AI model without a strict "Trailing Stop Loss" will ride a stock all the way down just as it rode it up.
8. Conclusion: The Verdict
The $10K Portfolio Challenge of Q4 2025 concludes with a decisive victory for the AI-Driven Quantitative Strategy, which delivered a +13.2% return against a stagnant market. This victory was not due to superior intelligence, but superior adherence to market truth. The AI did not care about "synergies" or "TAM". It cared only about price, volume, and trend.
Actionable Recommendations for Investors:
- Ignore the Noise: Corporate actions often destroy short-term value due to confusion.
- Respect Momentum: Trends tend to persist longer than valuation theory suggests.
- Watch the Lock-ins: For IPO investments, the lock-in expiry is critical.
- Use the Heightss Framework: Build systems that protect you from your own biases. If you can't backtest your thesis, it's just a guess.