Methodology

Methodology

This page explains exactly how our numbers are produced, so you can judge them yourself.

1. Coverage

All listed Japanese equities (3,900+) are screened every night. There is no manual pre-selection of "interesting" names — the same rules run against the entire market.

2. Point-in-time (PIT) discipline

Every historical study uses only the data that was publicly available on that date: prices as adjusted at the time, filings as of their disclosure timestamp, financials as first reported. Restatements are recorded separately rather than back-fitted. This is what makes an 18-year track record meaningful instead of decorative.

3. Audit stamps

Reports and datasets carry SHA hash stamps generated at publication. If a historical claim were later edited, the stamp would no longer match — making silent revision detectable by readers.

4. What the signals are (and are not)

Our labels (watchlist entries, warning signals, valuation and quality tiers) describe why a company is worth observing. They are not buy or sell recommendations, and we do not publish price targets or ratings.

5. Verification and negative results

Hypotheses are pre-registered before testing, and results are published whether or not they support our priors. Findings that failed are listed alongside those that held — for example, event studies where positive news produced no measurable excess return.

6. Human review

Generation is automated; publication is not. A human reviews every issue before it is sent, and is accountable for it.

Tsukiyo Research provides data and analysis for informational purposes only. Nothing here is investment advice or a recommendation to buy or sell any security. We are not a registered investment advisor in any jurisdiction. Past performance does not guarantee future results. Do your own research.