October 31, 2026

[Court Session #8] A Thorough Verification of TSE ETFs Based on “Net Proceeds After Taxes and Fees”—Out of 67 ETFs Never Before Examined, How Many Will Survive the Gate of …

Introduction

“All-Country or TOPIX alone is boring,” “Let’s mix in bonds or gold for defense,” “If you want more growth, use leverage, high-dividend, or thematic ETFs”—if you open brokerage feature pages, YouTube, or social media, such attractive sales pitches are lined up every day. At first glance, they all sound very plausible.

So, what happens if we don’t cherry-pick TSE-listed ETFs based on human subjectivity, but instead line up everything that meets the criteria and put them through a simultaneous screening based on the “true net proceeds” after strictly deducting taxes (20.315%) and trading costs (commissions/spreads) that we, as individual Japanese investors, actually pay in our specific accounts?

When I took this shelf that everyone believes “adding to makes it better” and placed it into a gate that was fixed before seeing a single yen of profit, a shocking conclusion that made me doubt my own eyes emerged. I will show you the entire record.

1. Rules of the Experiment—The “Gate of Net Proceeds and Statistics” Fixed Before Seeing a Single Yen of Profit

The experiment number is EXP-0015. The data is a snapshot from the TSE official J-Quants from October 2006 to October 2026, and I have locked and not read anything after April 1, 2026.

The subjects are 67 TSE ETFs that had never been examined in the sterile room before. In accordance with the periods for which data was available, I verified them across two eras and a total of 111 sets: approximately 11 years from 2014 (Period A: 2014-01-06 to 2024-12-30, 46 ETFs) and approximately 6 years from 2019 (Period B: 2019-01-04 to 2024-12-30, 65 ETFs) (since I also lined up previously examined ETFs for comparison, the number of ETFs in each period is higher than the number of “unexamined” ones). The cumulative number of times the sterile room has tested up to this experiment is 243 times, and the DSR discount is calculated based on this number.

The passing line was written in the pre-registration and fixed before seeing a single yen of profit (2026-10-27).

The meaning of “survival” for someone who already holds All-Country and BTC is having net performance that cannot be explained by chance, and also moving differently from both.

Net proceeds were calculated as follows.

Commissions are 0.10% one-way. Spreads are the larger of the trading value tier or the value estimated from the price
20.315% on dividends. For foreign-domiciled ETFs, a 10% withholding tax is deducted first, and foreign tax credits are not used
20.315% on capital gains
Evaluation is daily: “the amount remaining in hand if you quit and sell on that day”

One more thing, I will write this honestly. Before starting this experiment, I had a prior expectation (hypothesis) in my head that “stocks, bonds, high-dividend, and thematic types would all be wiped out, and only pure gold (1540) would remain.” However, I did not move the passing line or set the design values to match that prior expectation at all (recorded in §8 of the pre-registration).

2. Results of the Judgment—Why did all 67 disappear before the gate?

In Period A (approx. 11 years), the SPA test, which checks if even one item in the whole exceeds chance, passed with p = 0.002. In the StepM test, which checks if individual items can be said to be “better than chance,” 4 items including NASDAQ-100 (1545) and NY Dow (1546) passed (42 items were eliminated in StepM).

However, what stood in the way was the DSR gate, which discounts by the 243 trial counts. All 46 ETFs in Period A were all eliminated at this gate. Furthermore, NASDAQ-100 and NY Dow also failed the correlation line with stocks. The judgment is that “for someone who already holds global stocks, it is merely a repetition of the same price movement”—in other words, it does not become a source of new diversification.

In Period B (approx. 6 years), the p-value of the SPA test was 0.015, which did not reach the line of 0.0125. Since it could not be distinguished from chance as a whole, all 65 items failed at the SPA gate.

There were 0 ETFs that survived the strict statistical gate to the end. The confirmation period (January 2025 to March 2026) has not yet opened.

3. Why did even “Pure Gold (1540)” fall?

1540 was one of the few that passed the correlation-only sieve (correlation < 0.4, 2020-10 to 2024-12) in EXP-0010, which preceded this experiment. Since it is an ETF I have already looked at, I did not include it in the count of “not previously viewed” ETFs this time, and recorded it in a separate table.

1540 left more in hand than TOPIX (1306). Even so, it fell.

4. The “Ranking of What Increased the Most in Hindsight” and “Evidence of Being Able to Choose in Advance” are Completely Different

This line is not a line for choosing “what increased the most in hindsight.” It is a line that asks whether it exceeded chance enough to say it could have been chosen in advance, and whether it moves differently from what is already held. A Sharpe ratio of 0.86 is a good number, but it is not enough to say “it is not by chance” after 243 trials. According to the pre-registration estimate, an annualized rate of approximately 1.4 or higher is required for period A.

I will compare it to a lottery. Lining up the “winning numbers” after the drawing is finished is completely different from being able to choose those numbers with a basis before the drawing. Many of the “this much increased in the past” tables seen in popularity rankings or on social media are these “tables of winning lottery tickets after the fact.” They do not account for the discount based on how many items were chosen from (N = 243 in this sterile room).

The reason why gold moved together with stocks and BTC has not been verified from the records. It is possible that due to the weak yen, gold and foreign stocks viewed in yen rose at the same time (this is a guess; verification excluding currency factors is outside the scope of this experiment).

5. What if you had deposited 10 million yen for 11 years? — The Cruel “Net Proceeds Gap” with TOPIX (1306)

Apart from the statistical gate, let me show you the most vivid “reality of net proceeds” as an individual investor. I compared the net proceeds when buying 10 million yen in a lump sum at the closing price on 2014-01-06 and selling it on 2024-12-30, with the case of holding 1306 (TOPIX) under the same rules.

1306 ended with net proceeds approximately 13.11 million yen higher than the 10 million yen invested (effective CAGR 8.0%, effective maximum DD 28.4%).

Median by type (only ETFs not previously viewed)

Contents of the median (this is important)

The “approximately 18.11 million yen” for leverage/inverse is the midpoint between the winners and the losers. When you line up the 7 items, they split into two as follows.

Since it was 11 years where Japanese stocks rose, ETFs that bet upward won big, and ETFs that bet downward lost big. The 3 downward-betting ETFs ended with net proceeds 7.89 to 8.22 million yen less than the 10 million yen invested. Which one should have been bet on could not have been known as of 2014.

I will also write down the contents of the median for others.

The “approximately 10.23 million yen” for bonds is a value for only one item. It is the Listed Index Fund International Bond (1677, foreign bonds without currency hedge), which ended with net proceeds approximately 2.88 million yen higher than the 10 million yen invested. The effective Sharpe ratio was +0.39, which is not negative. The maximum DD was 15.5%, which was shallower than the 28.4% of 1306. The result is that while it fulfilled its defensive role, it lost significantly to 1306 in terms of growth. Domestic bonds and hedged US bond ETFs were not among those that met the conditions for this period.
The 16 stocks (broad indices) also vary widely. NASDAQ-100 (1545) left about 53.89 million yen more than 1306. On the other hand, Brazil (1325) and Mothers (1563) left 11.68 to 14.25 million yen less.
The 3 REITs left 2.27 to 8.55 million yen less, and their real Sharpe ratios were 0.32 to 0.48.

“Scars in the record” due to fixing classification rules in advance (things left as they are without correcting them with hindsight)

The method of categorization was determined mechanically by the words in the ETF names and fixed before looking at the numbers. If I re-categorized them after looking, it would become a “classification method tailored to the results,” so I did not correct them. As a result, the following discrepancies remain.

Daiwa’s TOPIX (1305) and Nikkei 225 (1320) are included in “Sectors/Themes” (because the mechanical classification reacted to the word “Securities” in the “Daiwa Securities” part of the brand name).
The Nikkei VI Futures ETN (2035) is included in “Stocks (Broad Indices)” (because the mechanical classification reacted to the word “Nikkei”).
Shinkin Central Bank’s preferred securities (8421) are included under “Unclassified” (they are not ETFs).
In the table after looking, SPDR S&P 500 (1557) is included in REITs (because the mechanical classification reacted to the “REIT” part of “State Street”).Street (Note: The Japanese text had a split word here).
The real Sharpe ratio of 1.67 for the Natural Gas ETF (1689) is suspected to be an appearance caused by the tick price bouncing at a low price range. The DSR is 0.210, which is below the line, so the judgment does not change.

6. What can be said from these results regarding “Bonds for defense” and “Leverage for growth”?

What can be said is as follows.

Just lining up the ETFs that increased the most after the fact is not proof that they could have been “chosen.” Both the Nikkei 2x and NASDAQ-100 significantly outperformed 1306, but when discounted by the number of trials, they cannot be distinguished from chance. Sales methods that show tables of past performance do not apply this discount.
For ETFs that bet on a direction, the result is almost entirely determined by whether that direction is correct. On the same “Leverage/Inverse” shelf, there are products that left about 19.7 million yen more than 1306 and products that left about 21.33 million yen less.
The VIX short-term futures ETF decreased significantly in a rising market. It ended with net proceeds about 9.98 million yen less than the 10 million yen invested (real maximum DD 99.8%).
The one bond ETF lost significantly in terms of growth in exchange for shallower declines. Whether this is expensive or cheap as the price of “defense” depends on the holder’s purpose. One cannot say “do not mix in bonds” based on this record alone.

I will also write down what cannot be said.

These 11 years have been a period where Japanese and foreign stocks rose and the yen weakened. In a period of continuous decline, the wins and losses in the table could be reversed.
There is only one bond ETF and one unclassified item, so they cannot be said to be representative of their types.

7. Is this standard too strict?—That is why my core assets are still in “All Country”

It is strict. At the time of pre-registration, it was written that “this line is strict, and the design is such that it is easy for zero to survive.”

Therefore, “0” does not mean that “no ETF is worth buying.” What it means is that for someone holding All Country World Index (ACWI) and BTC, after 243 trials, no ETF that could be added as a new, non-coincidental source was found from these 11-year and 6-year net proceeds records.

To be honest, as seen in this article, just as all 67 TSE ETFs I hadn’t looked at fell at the gate of the sterile room, even as I write this article, the majority of my core assets remain in “ACWI (All Country World Index).” I am an avid reader of the masterpiece “Just Keep Buying,” and the author, Nick Maggiulli, is also a data scientist; I deeply empathize with his thought process regarding data. My stubborn policy of “not moving my precious core assets from index funds unless statistical superiority is proven in a sterile room” is a conclusion I reached myself after analyzing data, and it is something I have been practicing long before that book was published. In fact, I currently teach data science to students, and I recommend and teach this “Just Keep Buying” book along with the “sterile room mindset” so that they won’t be deceived by anyone in the future. At this moment, I have yet to find anything to which I can entrust my precious core assets with more peace of mind than ACWI. (*This is a reference to my personal asset allocation and a recommendation of a textbook for data science education, not investment advice recommending the buying or selling of specific financial products.)

Next time, to [Court Session #9]—Can you win if you focus on “timing of trades” rather than the “box”?

An ETF is a “box for buying an index.” No matter how much you pick and choose among the 67 boxes lined up on the TSE, not a single one survived the gate of net proceeds and statistics to be added as a new holding.

Then, someone like this will inevitably appear. “It’s no good just buying and holding. If you look at the chart patterns (technical analysis), buy only when it’s going up, and escape when it’s going down, you should be able to crush the market average!”

Next time, [Court Session #9]. From the Golden Cross of moving averages, RSI, MACD, Bollinger Bands, Ichimoku Kinko Hyo, to the combined technique of “environmental recognition x buying the dip,” we will assemble all the “6,402” classic technical trading rules discussed in the world, and put them all to judgment simultaneously based on net proceeds after deducting taxes (20.315%) and fees for a specific account. What is the “shocking truth from 20 years ago” that awaits at the end of that?

Please be sure to “Like (heart mark)” and follow. Your likes will become the electricity for the sterile room’s computer that puts 6,402 classic technical rules to judgment all at once.

Have a break, have a KitSat. ☕🍫

Cryptographic Audit Stamp

Experiment ID: EXP-0015
Dataset Snapshot SHA-256: 9ec681e98e097eddfc368f5c0206f615cc2efbaa43a1eaf229cdc2fe6335dd7e
Result / Rule Artifact SHA-256: 6aaf0f73d2bdb2b7217a749e69e0a66f68182c73f82d9f7e72325998e05df7f4（results.json）
Out-of-Sample Lock: 2026-04-01

▼ [Court Session #0] Why investors who have tasted “victory” once go bankrupt.——Why do 99% of “methods that won in the past 10 years” die in real-world trading? The story of opening a lab for thorough verification in a “sterile room”

▼ [Court Session #1] Does the “200-day moving average” really work?——The results of subjecting 5.5 years and 16 trading signals to “sterile room” statistical testing

▼ [Court Session #2] The truth about “my own bot that earns money while I sleep”——The results of thoroughly verifying 24 grid trading patterns using 3.5 years of 1-minute interval data

▼ [Court Session #3] Is it true that “you make money if you buy on the day of a crash”?——The results of thoroughly verifying 3 types of dip-buying rules using 5.5 years of data

▼ [Court Session #4] Is it true that “you can win if you predict crashes with AI”?——The results of deploying a machine learning model that achieved a Sharpe ratio of 1.88 on “unknown data”

▼ [Court Session #5] Is it true that “mixing Bitcoin into ACWI is the strongest strategy”?——The results of thorough verification incorporating time differences and “Japanese taxes”

▼ [Court Session #6] Is “Just Keep Buying (Dollar-Cost Averaging)” statistically correct?——The results of a direct showdown between monthly accumulation and “monthly rebalancing” in the sterile room

▼ [Court Session #7] Is “Diversifying into BTC, ETH, and XRP is Safe” True? — The Results of a Thorough Verification of “True Diversification Effects” Based on Price Movements During Market Crashes

▼ [Court Session #8] A Thorough Verification of TSE ETFs Based on “Net Returns After Taxes and Fees”—Out of 67 Unexamined Funds, How Many Will Survive the Gate of the Sterile Room?

▼ [Court Session #9] Putting 6,402 “Classic” Japanese Stock Technical Indicators to the Test Based on Net Returns After Taxes and Fees—Exactly How Many Will Survive the Gate of Strict Statistical Analysis?

▼ [Court Session – Final Episode #10] Whose Information Do You Trust the Most? — My Confession on Opening the “Supreme Court” of Investment Information, and Two Proposals for You Who Seek the Truth (Coming Soon)

▼ [Permanent Archive Edition] A Quantitative Analysis Manual for Correctly Interpreting and Citing the “Sterile Room” — From Sharpe Ratios to Multiple Testing (DSR/SPA) and Translation Templates for Secondary Dissemination (Coming Soon)

* This article is a record of the verification of past data and does not recommend the buying or selling of any specific product. Past performance does not guarantee future results.