The Strategy Research Process

This is the complete guide to researching a trading strategy on Quanthop. It covers every stage from initial idea to live monitoring.

Most strategies fail. The purpose of research is to discover this before risking capital — not after.

Stage 1: Idea Generation

Every strategy starts with a hypothesis about market behavior:

• "Trend-following with moving average crossovers captures directional moves"
• "RSI extremes predict mean-reverting price action"
• "Volatility breakouts from Bollinger Bands signal continuation"

The hypothesis does not need to be novel. It needs to be testable.

Write your idea as a QSL strategy using the three required functions: define(), init(), and onBar().

Stage 2: Baseline Testing

Run a single backtest with default parameters to see if the idea has any merit at all.

What to look for:

• Does the strategy produce trades? (If not, your entry logic may be too restrictive)
• Are results clearly better than random? (Compare to a buy-and-hold benchmark)
• Is the equity curve reasonable? (Smooth growth, not one lucky trade)

What to ignore at this stage:

• Exact return numbers (parameters are not optimized yet)
• Small differences between intervals or symbols

If the baseline is clearly unprofitable across multiple configurations, the idea may not be worth pursuing. Move on.

Stage 3: Parameter Optimization

If the baseline shows promise, use optimization to search the parameter space systematically.

Key principle: You are not looking for the best parameters. You are looking for stable regions where many nearby parameter values produce similar results.

A sharp peak in the optimization surface means the strategy only works with one exact configuration — this is overfitting.

A broad plateau means the strategy works across a range of values — this is robustness.

Quanthop provides:

• Plateau detection to identify stable zones
• Stability scoring to rank parameter clusters
• Baseline scan for multi-asset validation

Stage 4: Walk-Forward Analysis

Walk-Forward Analysis (WFA) is the most important validation step.

WFA divides your data into rolling windows. For each window:

1. The optimizer finds the best parameters on in-sample data
2. The strategy runs with those parameters on out-of-sample data it has never seen
3. The window rolls forward and repeats

The out-of-sample results are the ones that matter. If a strategy performs well in-sample but fails out-of-sample, it is overfitted.

What good WFA results look like:

• Positive aggregate out-of-sample returns
• Consistent parameter selection across windows
• In-sample and out-of-sample results in the same direction

Stage 5: Stability Analysis

After WFA, analyze the robustness of your results:

• Do the same parameters work across different markets?
• How sensitive is performance to small parameter changes?
• Are there enough trades for statistical significance?

If the strategy passes these checks, it has earned a degree of confidence. If not, it joins the majority — safely rejected before any capital was at risk.

Stage 6: Adaptive Flow Monitoring

A validated strategy does not stay validated forever. Markets change. Adaptive Flow monitors deployed strategies and detects degradation:

• Tracks live performance against historical expectations
• Flags regime changes that may affect the strategy
• Suggests reoptimization when parameters drift

This creates a continuous feedback loop: validate, deploy, monitor, reoptimize.

Summary

Related Concepts

• What Is Walk-Forward Analysis? — The theory behind WFA
• Strategy Overfitting Explained — Why most backtests lie
• Parameter Stability — What makes parameters robust