Skip to main content
Sensitivity Analysis is the ultimate tool for avoiding Overfitting. Overfitting occurs when you “fine-tune” your strategy parameters (like setting a 14-day RSI, or a 20.5% trailing stop-loss) so perfectly to past data that the strategy stops being a generalized system and becomes a memorized historical chart. If your strategy is only profitable at a 14-day window, but loses money at a 13-day or 15-day window, you have overfitted your model. It will almost certainly fail in live, forward-moving markets.

​
The Robustness Test

When you execute a Sensitivity Analysis, Kalpi varies your core input parameters (e.g., shifting asset weights or rebalancing windows) by ±10% to create a performance heatmap.

​
How to read the heatmap:

  • The “Green Plateau”: If you see a large, flat area of profitability across multiple parameter variations, you have found a Robust Strategy. It is flexible, consistent, and highly likely to work in live trading.
  • The “Lone Peak”: If you see a single, tiny pixel of high profit surrounded by a sea of losses, you have an Overfitted Model. Discard this strategy immediately—it is a mathematical illusion.
Always aim for strategy parameters that sit in the middle of a “Green Plateau,” rather than chasing the absolute highest peak of a “Lone Star” parameter.