Are you not weak to small market fluctuations

Introduction

In the previous article, we explained the meaning of checking by shifting numbers for the fourth test, "Parameter Perturbation." This time is the fifth and final test, "Noise Robustness."

What is Noise Robustness

We take real price movement data and add a tiny random fluctuation (noise), then repeat the verification under the same conditions. The overall trend of price movement remains unchanged, with only small fluctuations deliberately added on top.

Why add noise on purpose

Market data always contains fine up-and-down movements, or "noise," on a daily basis or even within the same day. If the conditions are highly sensitive to these tiny fluctuations, and the signals or performance change drastically with only a small amount of noise, then those conditions are likely to be unstable when used in actual markets.

Conversely, if the performance follows a similar pattern even with some noise added, that condition is considered to have a certain resilience to small price movements.

What becomes visible only when all five tests are together

Thus far we have introduced five robustness tests: separating the observed period from the unobserved period (IS/OOS split), shuffling the order of trades (Monte Carlo shuffle), comparing to random guessing (random comparison), slightly perturbing numbers (parameter perturbation), and adding small noise (noise robustness). Each of these tests challenges, from a different angle, whether the good results are a coincidence. Passing just one is not enough; only after passing all five do we proceed to the next stage.

Impressions so far

Having presented all five tests, what stands out is that “verification is a process of continual doubt.” In the next article, in addition to these five tests, we will also explain another metric we always check, the “SQN,” and how to read its numbers.

Disclaimer

・The content presented in this article does not advocate specific buying or selling, nor does it provide investment advice. It is solely for information sharing.

・Results based on past data do not guarantee the same results in the future. Markets are always evolving.

・Whether you actually decide to trade is at your own risk.

・The indicators and tools introduced in the future are just one of the factors for market analysis, and using them does not guarantee profits.

If you’re interested, please also check out the product page of Indicator Institute.