There were numbers that seemed nice lined up, but not a single one remained

It's a shame, but I had never encountered a verification as many times as this where that word "惜しい" (near miss) comes to mind, and yet not a single one survives until the end.

Last time → [Episode 2] When I tried to examine everything, the calculations never finished

A recap of what happened up to the last time

Last time, I explained the process of attempting to search the entire time axis at shorter time scales (M15, M30) as a whole, where the calculations did not finish at all, and how narrowing down to a recent fixed period allowed them to fit into a realistic time frame.

This time, after completing the narrowing, I will share the results of actually running the search.

Looking at just the numbers alone, the result didn't look bad

When I tested one of the shorter time axes, I looked at the numbers that appeared on the screen at first and honestly felt my expectations rising. Of the 29 currency pairs, more than half of the pairs had a single indicator (the number that measures the series’ ability to consistently produce results, as I’ve conveyed repeatedly in this series) that exceeded the level that had been deemed “passable” in prior verifications.

“There might be many that get found.” While thinking that, I waited for the results of the robustness test (a process that persistently checks from five different angles whether the apparent numbers are genuine).

【Image: 34_looks_good_vs_survives.png to be inserted here (owner will attach manually)】

The result was 0 cases

However, when I opened the lid, the result was 0 cases.

The numbers that seemed promising all collapsed at some point in one of the five robustness tests. For all 29 currency pairs, there were no candidates that survived until the end.

In previous verifications, I had experienced annihilation as a result. But this time it felt a little different. When looking at the single numbers, there were many that, compared with candidates that had passed before, were not inferior in their level, and not just one or two but many that stood out. Yet in the end, not one remained. I was honestly puzzled by this gap.

【Image: 34_zero_result.png to be inserted here (owner will attach manually)】

Why there were 0 cases, I still don’t know

Here, I have to be honest with you.As of now, there is no clear reason why this time it ended up with 0 cases.

In previous series, I repeatedly explained the hypothesis that the reason performance does not improve on shorter time scales is “each price movement tends to become smaller relative to trading costs.” However, this time that explanation alone does not seem to fit. The individual numbers themselves weren’t bad.

Another possibility is the effect of the narrowing I described last time. The time axis tested this time differs from other time axes in the length of the calendar period that can be used for verification. The shorter the period, the more likely it is that whether the robustness test (especially tests that split the data into early and late parts) passes is influenced by luck. However, this is only one hypothesis and has not been confirmed yet.

【Image: 34_unknown_reason.png to be inserted here (owner will attach manually)】

Saying that we don’t know, and saying it’s unknown

“Numbers that looked promising lined up, yet none remained.” It would have been easy to assign a plausible reason and pretend to understand. But that would be contrary to the posture this research institute has always valued.

What is clear is that numbers that look promising and the candidates that actually survive are completely different things. And the fact that the reason is still unknown is something I believe is valuable to honestly record in this verification's documentation.

Disclaimer

• The content published on this note does not advocate a specific buy/sell or provide investment advice. It is only sharing information.
• Verification results based on past data do not guarantee the same results in the future. Markets are always changing.
• Whether you actually buy or sell should be decided at your own responsibility.

Next time, I will share a story about encountering the same face again while testing a short time axis. If you’d like, you can follow so you’ll receive the next article as well.

※Currently, the official LINE account is under preparation. We will announce here as soon as it opens.

Next: [Episode 4] I encountered the same face again. But I’m not surprised anymore