Categorising the Trading Strategies: The Big Picture by Prodipta Ghosh

About the webinar video:

In the ever-evolving world of algorithmic and quantitative trading, understanding the landscape of systematic strategies is critical for professionals and aspirants alike. In a recent talk, Prodipta Ghosh—Vice President at QuantInsti—shared a detailed framework for categorising, building, and validating trading strategies in today’s dynamic markets.

In Prodipta’s words:

“If you can’t clearly define the DNA of your strategy, you’ll never be able to scale or validate it.”

About Prodipta Ghosh:

Prodipta Ghosh is the Vice President at QuantInsti, where he leads the development of fintech products and platforms. With a strong background in quantitative trading, he has worked at Deutsche Bank as a Vice President, at Standard Chartered Bank as an Associate, and as a Scientist at the Defence Research & Development Organisation.

At EPAT, he teaches programming and financial computing, helping participants relate theory to market-relevant applications using Python, R, and quantitative techniques.

“The EPAT classroom is where practical meets academic. You get to build, test, and deploy models that work in markets,” says Prodipta.

Read more about Prodipta here.

Let’s walk through the highlights of his discussion.

Classification of Trading Strategies

At a high level, trading strategies can be classified along two primary dimensions:

1. The underlying trading view or risk factor, such as momentum, mean reversion, carry, or event-driven.
2. The trading style, which could be fundamental, technical, or quantitative.

What’s fascinating is how the same core idea, like momentum, can look very different depending on the style of execution. For example:

• A technical trader may use price charts or moving averages.
• A fundamental investor might focus on trends in earnings or analyst upgrades.
• A quant could distinguish between time-series momentum (performance of one asset over time) and cross-sectional momentum (performance relative to other assets).

“Momentum isn’t just a charting pattern or a buzzword, it’s a deep-rooted behavior of markets, and each style interprets it in its own language,” Prodipta explained.

Quick Quiz:

Same Strategy, Different Style: Examples Across the Board

One of the most insightful parts of Prodipta Ghosh’s talk was how the same core trading idea can look dramatically different when approached through technical, fundamental, or quantitative lenses. This understanding is crucial for anyone transitioning between styles or looking to blend multiple perspectives in their strategy development.

“A quant doesn’t just ask what’s moving, but how consistently and relative to what benchmark,” Prodipta emphasized, highlighting the depth that quantitative analysis brings to even seemingly simple concepts.

Let’s break it down by strategy type:

1. Momentum Strategies

These aim to capture trends—assets that are rising tend to keep rising, and falling ones tend to keep falling. But the way you define and exploit momentum depends on your toolkit.

• Technical Approach: A technical trader identifies momentum through price action. This could involve using indicators like moving averages (e.g., 50-day crossing above 200-day), trendlines, or breakout patterns from support/resistance zones.
• Fundamental Approach: Here, momentum isn’t about price movement but underlying value drivers. For instance, traders may look at earnings upgrades, revenue growth, or analyst sentiment trends as a sign that an asset is gaining long-term favor.
• Quantitative Approach: Quants differentiate between:

Time-series momentum: Is this asset doing better than its own past?

Cross-sectional momentum: Is this asset outperforming its peer group? Using historical data and statistics, quants quantify these patterns and automate entries/exits.

2. Mean Reversion Strategies

These assume that prices eventually revert to their historical average or fair value after short-term deviations.

• Technical Approach: Traders use tools like Bollinger Bands, RSI, or candlestick reversal patterns to identify overbought/oversold conditions, aiming to profit from bounces or pullbacks.
• Fundamental Approach: This resembles value investing. A fundamentally cheap asset (based on metrics like P/E or P/B) is expected to revert upward in price over time as the market corrects its mispricing.
• Quantitative Approach: Quants often use pairs trading or statistical arbitrage, where they model the historical relationship between two or more assets. When the price relationship diverges from the statistical norm, they bet on convergence.

Example: A quant might model the price spread between Coke and Pepsi. If one spikes while the other doesn’t, they may short the outperformer and long the underperformer.

3. Carry Strategies

These aim to earn returns by holding positions that benefit from differences in yield, interest rates, or premiums—often in calm or predictable markets.

• Technical Approach: While rare, some technical traders simulate carry exposure by engaging in short volatility trades or gamma scalping, attempting to profit from small price moves or time decay.
• Fundamental Approach: A classic example is the FX carry trade, where traders borrow in low-interest currencies (e.g., JPY) and invest in high-yielding ones (e.g., ZAR), profiting from the interest rate differential.
• Quantitative Approach: Quants implement carry via systematic high-frequency strategies such as market making, where they profit from the bid-ask spread and inventory carry, often relying on low-latency infrastructure.

“Carry strategies don’t require price movement to generate returns—but they demand disciplined risk management because they can suffer in volatile markets,” Prodipta noted.

4. Event-Driven Strategies

These revolve around exploiting market-moving events, such as earnings releases, M&A announcements, or macroeconomic data.

• Fundamental Approach: Traders use discretion and qualitative research to analyze the impact of corporate events. For example, they may anticipate stock movement based on merger rumors or regulatory changes.
• Quantitative Approach: Quants use natural language processing (NLP) to analyze news headlines, earnings call transcripts, or social media in real-time. This feeds into event-based models that trigger trades based on sentiment or keyword analysis.

Example: A model could scan for phrases like "beats expectations" in an earnings report and instantly go long.

• Technical Approach: This style is rarely applied to event-driven strategies because price charts alone don’t offer enough predictive power during sudden news-driven volatility. However, some may trade technical breakout patterns immediately following an event.

Quick Quiz:

Anatomy of a Systematic Trading Strategy

Building a successful systematic trading strategy is not just about having a great idea—it’s about designing a structured, repeatable process that can operate in real-time markets. Prodipta Ghosh breaks this process down into five essential layers, each critical to the overall robustness of the system.

“A trading strategy isn’t a black box—it’s a layered architecture where each component must work seamlessly with the others,” explained Prodipta.

Here’s a deeper look at each layer:

1. Input Layer

What it is: This is the foundation. It consists of raw data that feeds the system.

Examples:

• Market data: Open, High, Low, Close, Volume (OHLCV)
• Macroeconomic data: Interest rates, GDP, inflation
• Fundamental data: Earnings reports, financial ratios
• Alternative data: Social media sentiment, analyst forecasts, news feeds

Insight: “Good strategies begin with clean, diverse inputs,” Prodipta stressed. More variety in input data often leads to more resilient signals.

2. Data Processing Layer

What it is: This layer converts raw data into features or indicators that a model can use.

Examples:

• Moving averages
• Volatility bands
• Technical or fundamental indicators (such as RSI or PE trends)
• Custom transformations like z-scores or cross-sectional rankings

Why it matters: This is where domain knowledge plays a critical role. Poorly designed features result in poor model performance, no matter how sophisticated the algorithm.

3. Intelligence Layer

What it is: This is the core of the system—the layer that generates trading signals.

Examples:

• Rule-based logic (for example, moving average crossovers)
• Statistical models
• Machine learning models such as random forests or neural networks

Key point: Whether simple or complex, the intelligence layer must be consistent, interpretable, and grounded in logic.

“The intelligence layer doesn’t have to be complex, but it must be consistent and explainable,” Prodipta noted.

4. Order Management Layer

What it is: This layer translates signals into position sizes and portfolio-level decisions, while managing risk.

Functions include:

• Position sizing
• Capital allocation
• Portfolio balancing
• Application of stop-losses and risk controls

Why it’s crucial: Even the best signals can be undermined by poor money or risk management. This layer ensures discipline and portfolio integrity.

5. Execution Layer

What it is: The final step—placing trades in the market while minimizing transaction costs and slippage.

Techniques used:

• Smart order routing
• VWAP/TWAP-based execution
• Latency-sensitive execution engines
• Real-time trade monitoring

Takeaway: Execution can make or break a strategy. “A delay or wrong route in execution can turn a winning trade into a loser,” Prodipta warned.

Quick Quiz:

What Inputs Should You Use?

Diverse inputs lead to stronger strategies. Here are some key categories:

• Market data: OHLCV (Open, High, Low, Close, Volume)
• Positioning data: Investor flows, long/short exposure
• Fundamentals: Financial ratios, earnings reports
• Macroeconomic indicators: Interest rates, GDP, inflation
• Alternative data: Sentiment, social media signals, analyst ratings

Quick Quiz:

How Are Trading Rules Built?

In the world of systematic trading, building trading rules is one of the most fundamental steps. There are two predominant approaches to crafting these rules: hypothesis-driven and data-driven. Both have their merits and are commonly used in various strategies, but each requires a different mindset and approach.

1. Hypothesis-Driven Approach

What it is:
This approach is based on domain knowledge—the trader uses their understanding of financial markets, asset behaviour, and economic principles to form a hypothesis or rule. It’s rooted in traditional analysis and is often easier to explain to others.

Example:

• Moving average crossovers: A simple example of a hypothesis-driven rule is when a 5-day moving average crosses above a 20-day moving average, signalling a bullish momentum. This is based on the belief that short-term price movements (reflected by the 5-day MA) will drive prices higher in the long run, and vice versa.
• Another example might be identifying specific chart patterns or applying technical indicators (like RSI or MACD) to define buy or sell points.

Strengths of this approach:

• It’s easier to understand and explain, making it a good starting point for traders who are more familiar with traditional techniques.
• Faster to implement and generally requires fewer resources compared to more complex data-driven models.

Prodipta's Take:

“Hypothesis-driven approaches are often a great way to start. They allow traders to test their market assumptions with real data.”

2. Data-Driven Approach

What it is:
This approach leverages machine learning (ML) techniques to uncover patterns in the data, often with the help of supervised learning. Here, rules are not predefined based on human intuition; rather, the model learns from historical data and creates its own logic based on patterns that have proven successful in the past.

Example:

• Time-series forecasting or classification algorithms (like decision trees or neural networks) can be used to predict future price movements or classify market conditions (e.g., “buy,” “sell,” or “hold”).
• Reinforcement learning models that automatically adjust trading strategies based on feedback and the environment’s response are becoming increasingly popular in high-frequency trading systems.

Strengths of this approach:

• It can uncover non-obvious patterns and correlations that humans may miss.
• As the system is data-driven, it can adapt over time, improving its decision-making abilities based on new data.

Why Both Matter

The distinction between these approaches is important, but Prodipta Ghosh emphasizes that modern quant traders must be proficient in both:

“Quant traders today must be comfortable switching between statistical intuition and ML techniques. One validates the other,” Prodipta advised.
“It’s not about choosing one over the other; it’s about knowing when to use each approach effectively based on the problem you’re trying to solve.”

In other words, hypothesis-driven rules provide a solid foundation for strategy development, but machine learning models can optimize and evolve these rules as new data comes in. By combining both approaches, traders can create more robust and adaptive systems that capitalize on the strengths of each method.

Quick Quiz:

Backtesting vs. Forward Testing: What’s the Difference?

• Backtesting: Simulates how your strategy would have performed on historical data. Most research time is spent here.
• Forward testing: Runs your strategy in real-time (or on unseen data) to validate its predictive power before live deployment.

Important reminder: Backtests can be misleading due to biases like overfitting or look-ahead bias. Forward testing acts as a necessary sanity check.

“If a strategy only works in hindsight, it’s not a strategy—it’s a story,” Prodipta warned.

Quick Quiz:

Key Challenges in Strategy Validation

Even with a well-structured framework for developing systematic trading strategies, validation is far from straightforward. Traders often face several significant challenges when attempting to ensure their strategies are robust and reliable. Let's explore some of the common hurdles encountered during the validation process:

1. Poor or Missing Data

One of the primary challenges in strategy validation is data quality. Inadequate data or missing data points, especially for macro-economic or fundamental inputs, can severely impact the reliability of a strategy. For example:

• Macroeconomic indicators (such as GDP, interest rates, or inflation rates) often have gaps in historical data, particularly in emerging markets or during periods of economic upheaval.
• Corporate earnings reports, fundamentals, or analyst ratings can also be inconsistent, leading to incomplete datasets that fail to capture critical market events or trends.

Without proper data coverage, strategies built on these inputs might fail to perform as expected in real-world scenarios.

“A strategy that works on clean, well-sampled data might not hold up when applied to real-world, messy data. It's crucial to have accurate, complete datasets before running any backtest or forward test.” — Prodipta Ghosh

2. Common Backtest Biases

Another critical issue in strategy validation is the potential for backtest biases. These biases can easily mislead traders into thinking a strategy is more robust than it actually is. Common biases include:

• Lookahead Bias: This occurs when future data is inadvertently included in a model’s training set. In reality, traders cannot access future market data while making decisions, so lookahead bias presents an unrealistic advantage during backtesting.
• Survivorship Bias: This happens when only data from surviving assets (e.g., stocks that are still listed today) are included in the backtest, neglecting assets that failed or were delisted. This leads to over-optimistic results.
• Data Snooping Bias: When the same dataset is used repeatedly to test multiple strategies or models, the results may reflect the peculiarities of the data rather than the true performance of the strategy.

These biases can make a backtest look far better than it will perform in the live market, which is why it’s essential to apply corrective measures during validation.

“Backtests are useful, but they must be conducted with caution. The absence of biases is crucial, and strategies must be validated through more than just historical data.” — Prodipta Ghosh

3. Difficulty Modeling Discretionary or Qualitative Ideas

In many cases, trading decisions are influenced by discretionary factors, such as a trader’s experience or qualitative insights into market conditions. These factors can be difficult, if not impossible, to model in a quantitative system. For example:

• Market sentiment based on geopolitical events or corporate news can be hard to quantify and incorporate into traditional models.
• Psychological factors like fear and greed, which may heavily influence market behaviour, are often ignored in purely algorithmic systems.

While some machine learning approaches, such as sentiment analysis or natural language processing (NLP), have attempted to quantify these qualitative aspects, systematically modelling human judgment remains a significant challenge.

Quick Quiz:

The Importance of Rigorous Testing

Despite these obstacles, rigorous testing is non-negotiable before deploying any strategy in live markets. Traders must undergo both backtesting and forward testing to validate their strategies thoroughly. Each of these testing approaches has its own strengths:

• Backtesting allows traders to simulate how the strategy would have performed on historical data, offering insight into its potential profitability and risk.
• Forward testing runs the strategy on unseen data or in real-time to check its predictive power and to observe how it behaves in live market conditions.

Only by combining these testing approaches can traders get a true sense of how their strategy might perform in the future. Testing must also account for the full range of potential market conditions—because a strategy that works during one market cycle may fail in another.

“Strategy validation is a tough process, but it's essential. Traders must be meticulous in testing their strategies, using both past data and forward tests. It’s the only way to ensure their strategies are not just luck.” — Prodipta Ghosh

Final Thoughts

The world of systematic trading is vast and nuanced. By understanding how different strategies are conceptualized and implemented across various styles—and by using a structured framework to build and validate them—traders can navigate markets more confidently and systematically.

“Your edge lies in structure. You can’t control the market, but you can control how your model sees and responds to it,” Prodipta concluded.

As Prodipta Ghosh rightly emphasizes, success in systematic trading begins with robust data, clear logic, and disciplined testing.

Ready to take the next step?

Next Steps

After understanding the broad classifications of trading strategies—such as momentum, mean reversion, and style-based approaches—it's essential to dive deeper into their real-world application, indicators, and evaluation techniques.

Begin with the foundational concepts of trend-following in Five Indicators to Build Trend-Following Strategies, where you’ll learn how indicators like RSI, MACD, and moving averages help traders capture market momentum. Then, explore the mechanics of mean reversion strategies through Mean Reversion Strategies: Introduction & Building Blocks, which introduces reversion logic, indicators, and basic risk management principles.

To better understand volatility’s role in market behavior, study Volatility Trading Strategies, a Quantra course that walks you through directional and non-directional volatility setups using implied and historical volatility data.

For those looking to transition from theory to strategy implementation, the Technical Indicators & Strategies in Python course offers hands-on experience building momentum and mean-reversion models in Python. You can also build your foundation with Trading Strategies with Moving Averages, which shows how moving average crossovers are used in trend-based systems.

Newer traders may find value in reading How to Build a Trading Strategy—a step-by-step guide to moving from an idea to a backtested, rules-based model. Complement this with How to Backtest a Trading Strategy, which teaches evaluation techniques for validating strategy performance before deployment.

For a structured learning path that integrates all these elements from strategy classification and development to backtesting, risk management, and deployment, the Executive Programme in Algorithmic Trading (EPAT) provides comprehensive training in both technical and fundamental strategies using Python and advanced statistical methods.

Whether you’re just starting or refining your approach, this collection of blogs and courses will help you develop, test, and evolve your trading strategies more effectively.

Connect with an EPAT career counsellor to discuss the right quant role for your background.