Chapter 4: Architecting Your Edge: Building Initial Strategies
With a solid foundation of correctly configured and imported data, you’re now ready to step into the role of a strategy architect. This is where you leverage StrategyQuant’s power to generate an initial pool of potential trading strategies. This phase is about casting a wide, intelligent net to uncover promising candidates that will later undergo rigorous testing. For someone exploring a new career in quantitative trading, this is where the creative yet systematic work truly begins.
Section 4.1: Defining Your Battlefield: Market and Timeframe
First, you need to specify the market and timeframe for which StrategyQuant will build strategies.
- Backtest Engine:
- Choose the engine that matches your intended trading platform (e.g., MetaTrader, TradeStation/NinjaTrader). They handle data and execution simulation differently. While this guide leans towards forex, the principles apply broadly.
- Symbol:
- Select the currency pair (e.g., EUR/USD, GBP/JPY) or other financial instrument. For beginners, highly liquid major forex pairs like EUR/USD are often recommended due to typically tighter spreads and abundant data.
- Period (Timeframe):
- Choose the chart timeframe your strategies will operate on (e.g., M15, H1, H4, D1).
- The H1 (1-hour) timeframe is a popular choice, offering a good balance between a sufficient number of trading signals and reduced sensitivity to minor execution variables like slippage and spread widening compared to very short timeframes.
- Start Date and End Date (Initial Development Range):
- Define the historical period for this initial strategy generation. It’s crucial to reserve distinct blocks of data for later Out-of-Sample (OOS) testing.
- Example for 2025: If your data extends to December 2025:
- Initial Build Period (In-Sample + 1st OOS): January 2018 – June 2024
- Reserved for 2nd OOS Test: July 2024 – December 2024
- Reserved for 3rd OOS Test: January 2025 – December 2025
- Out-of-Sample Period (During Initial Build):
- StrategyQuant allows you to specify a portion of your “Initial Build Period” as Out-of-Sample. A common and effective setting is 25-33% OOS.
- The strategy is “discovered” and optimized on the In-Sample (IS) portion (e.g., Jan 2018 – Dec 2022 if using ~30% OOS on the 2018-mid 2024 range).
- Its performance is then immediately checked on the 1st OOS part (e.g., Jan 2023 – June 2024) – data it hasn’t seen during development. This is a first, quick filter for over-optimization.
- Test Precision:
- Crucial “No-Nonsense” Setting: For reliable backtests, especially on H1 or higher, set test precision to “1 minute data” (or the highest available if not M1). This ensures that price movements within each main candle (e.g., within an H1 candle) are simulated based on the more granular M1 data, leading to far more accurate modeling of stop-loss, take-profit, and pending order executions.
- “Trade On Bar Open” is a specific setting where trades only occur at the open of a new bar. This might be suitable for some systems but is less common for dynamic forex strategies.
- Ensure other settings like Spread are realistic for your chosen symbol and broker.
Section 4.2: Core Strategy DNA: Essential Options
Here, you set basic operational parameters for the strategies.
- Trade Directions: Typically, you’ll want to build strategies that can trade both Long and Short to capture opportunities in all market directions.
- Most other settings here can often be left at their defaults when starting. Functions like “Limit signals to time range” can be explored later if you want to build strategies that only trade during specific market sessions (e.g., London open).
Section 4.3: The Algorithm’s Ingredients: Indicators & Logic
This is where you define the “building blocks” – the indicators, price conditions, and logical operators – that StrategyQuant will use to construct the trading rules.
- Entry Rules Building Blocks:
- Select from a wide array of technical indicators (Moving Averages, RSI, Stochastics, Bollinger Bands, etc.), price values (Open, High, Low, Close), and logical conditions (e.g., MA1 crosses above MA2, RSI < 30).
- No-Nonsense Approach: While you can enable many indicators, be mindful that too many can lead to overly complex and curve-fitted strategies. Start with common, well-understood indicators.
- Volume: For decentralized forex markets, volume data can be unreliable. It’s often wise to disable volume-based indicators for forex strategy generation.
- Order Types & Exit Building Blocks:
- Define how strategies can enter (Market, Stop, Limit orders) and exit trades.
- Essential Exits: Always include Stop Loss. Profit Target is also common.
- Simplicity is Often Key: While options like “Move Stop Loss to Breakeven” or “Trailing Stop” exist, sometimes simpler exit logic (fixed Stop Loss and Profit Target) can lead to more robust strategies. You can explore more complex exits later. The goal here is to give StrategyQuant a clear set of tools to work with.
Section 4.4: The Evolution Engine: Understanding Genetic Options
This section controls StrategyQuant’s “genetic algorithm,” the evolutionary process it uses to discover and refine strategies. It starts with a random population of strategies and iteratively improves them based on your performance criteria, mimicking natural selection. For newcomers, the default settings here are usually well-optimized and a good starting point. Understanding the deep mechanics isn’t essential to begin, but know that this is the “engine room” of strategy discovery.
Section 4.5: Risk Blueprint: Money Management Essentials
This defines how trade sizes are determined during backtesting.
- Fixed Size: Each trade uses a constant size (e.g., 0.1 lots). Good for initial, unbiased evaluation of the strategy’s raw edge.
- Fixed Amount (Fixed Monetary Risk): Position size is calculated so each trade risks a specific cash amount (e.g., $100). This is a robust way to compare trades as each carries similar risk.
- Risk Fixed % of Account:
- No-Nonsense Warning (Crucial for Development): Avoid this for initial strategy generation and testing. If a strategy has a lucky winning streak, the account size grows, and subsequent trades risk larger and larger absolute amounts. This skews performance metrics like drawdown (in percentage terms) and makes it hard to assess the strategy’s inherent stability. While percentage risk is common for live trading portfolio management, for discovering a strategy’s core viability, fixed size or fixed amount risk is preferred.
- Initial Build Recommendation: Often, for the very first generation run, you might select “No money management” or a very simple fixed size. This allows you to focus purely on the predictive power of the entry/exit signals. Money management can be applied more rigorously in later retesting stages.
Section 4.6: Defining Success: Key Performance Metrics for Ranking
This is a pivotal step. Here, you tell StrategyQuant how to judge the “fitness” of the strategies it generates and which ones to save for further scrutiny.
No-Nonsense Metrics – Focus on Ratios and Consistency:
Avoid being swayed by large absolute profit numbers in USD, as these can be easily manipulated by position sizing. Instead, focus on robust ratios and consistent performance across both In-Sample (IS) and the initial Out-of-Sample (OOS) periods.
- Profit Factor (IS and OOS):
- Ratio of gross profit to gross loss. Minimum 1.3 is a good starting point. Crucially, look for similar values in IS and OOS. A big drop in OOS Profit Factor is a major red flag.
- Return/DD Ratio (IS and OOS):
- Total net profit divided by maximum drawdown. Measures reward relative to risk.
- Target at least 0.5 per year of data (e.g., for a 5-year IS period, aim for Ret/DD >= 2.5). OOS Ret/DD should also be strong and reasonably proportional.
- % Wins (Win Rate):
- While not the only factor, a win rate below 30-35% can be psychologically challenging and may indicate a less stable edge. Aim for a reasonable minimum.
- Number of Trades (IS and OOS):
- Ensure enough trades for statistical significance. For H1, perhaps 200+ in IS and 100+ in OOS. Adjust for different timeframes (fewer for D1, more for M15).
- Average Trade (Net P/L per trade):
- Should be sufficiently positive after considering spread and commission to ensure profitability.
Deferring Intensive Robustness Tests:
During this initial building phase, you typically don’t enable the full suite of intensive robustness tests (Monte Carlo, etc.). These are applied later to the promising strategies selected from this generation run.
Launching the Build:
With all settings configured, click “Start.” StrategyQuant will now begin its work. This can take many hours, even overnight, as it generates, backtests, and ranks thousands, if not millions, of potential strategy variations.
The next morning, or when the process completes, you’ll have a databank of strategies that met your initial criteria. These are your raw recruits, ready for the rigorous boot camp of robustness testing. This systematic generation is a far cry from manual, intuition-based trading – it’s the foundation of a quantitative career.