[Makedeco] New NISA x Data Analysis Event (2024/2/15) Summary

Makedeco is a Discord community for botters who build analysis and prediction models primarily by utilizing stock-related market APIs, which I co-manage with JPX Market Innovation & Research, Inc. (JPX-R).

I have organized about 18 events for Makedeco so far, but I hadn't been summarizing them. To prevent information from past events from being lost, I have decided to properly summarize the information after each event and post it on this blog from now on.

Seminar Videos (YouTube)

• Makedeco Channel
• Relevant Video

Mr. UKI: Key Points for Trading Individual Stocks in the New NISA

• Materials: Key Points for Trading Individual Stocks in the New NISA

Personal Notes

• He explained in detail how he actually develops strategies while looking at the data. Daken-san's impressions are very clear:

https://twitter.com/daken_in_market/status/1758117245119836668

• There are probably few presentations that explain practical examples of searching for anomalies using monthly data as thoroughly as this.
• I think many people found the approach of examining ideas by scale, sector, and factor to be extremely helpful.

Mr. yoshiso: Tidbits Useful for Portfolio Construction

• Materials: Tidbits Useful for Portfolio Construction

Personal Notes

• By properly verifying portfolio-like approaches with recent data, various insights can be gained.
• GOLD is indeed excellent as an alternative asset. It is clear at a glance that it significantly reduces drawdown without significantly compromising returns.
• The opinion that bonds are superior in a portfolio with stocks is quite dubious. It is not very reliable, as seen in how the stock + bond portfolio was heavily damaged in 2022 (it is better to understand the relationship between stocks and long/short interest rate spreads rather than the relationship between stocks and bonds).
• BTC was an alternative asset equivalent to gold with a correlation coefficient of 0.2 until 2020, but in recent years the correlation coefficient has been 0.6, and the portfolio effect is weak.
• However, backtesting shows that incorporating about 10% of BTC into a stock portfolio increases returns by 1.5 times without increasing drawdown, making it excellent as a return amplification function.
  

Seminar Question List

We answered the following questions:

• Is this effective for altcoins like ETH instead of just BTC?
• Could you share your intent or thoughts behind discussing correlation this time?
• How do you search for ideas for strategies to verify?
• I believe rebalancing was assumed to be at the end of the month, but do institutional investors generally not rebalance at other times?
• To be blunt, are both of you using the New NISA?
• When making actual investment decisions based on optimal portfolio allocation or historical data analysis, I feel like they strictly need to be verified with test data. In such cases, do you use methods like CPCV? Or do you operate the analysis results as an investment strategy directly because the strategy is simple or logically sound?
• A question for Mr. UKI's strategy: Whether to target T (end of month) or T-1 for rebalancing seems like a game of how far one can anticipate. In that case, do you have an image that the volatility caused by this rebalancing will ultimately decrease? Also, how do you imagine it will end up?
• Besides investment prospectuses, is there anything else you always look at for strategy formulation?
• So rebalancing differs even for prop desks. I’ve become interested in the underlying mechanism of end-of-month rebalancing by institutional investors; do you have any recommended articles or data sources?
• Please let me know if there are any rules or investment strategies that institutional investors are bound by that are written somewhere other than prospectuses.
• How do you collect academic papers? (Are there any famous journals, etc.?)
• Question for Mr. yoshiso: Wrap accounts and robo-advisors have become mainstream, but I'd like to hear your opinions or thoughts on other companies' strategies and performance. Thank you.
• Since the factors that are rising are more likely to be selected for rebalancing at the end of the month, would performance improve if the analysis factored in recent price fluctuation rates? For example, targeting semiconductors since they are rising now.
• I am planning to start reading ETF prospectuses; do investment rules differ significantly between each ETF?
• Is Mr. yoshiso's recommended model still CNN?
• Is a strategy of taking positions by reading large orders from the order book or time and sales realistic? I don't have any specific ideas, though...
• End-of-month rebalancing seems like a targetable rule since it's a simple "reduce rising stocks at the end of the month"; don't institutional investors take countermeasures?
• A question about models: what about Transformer or S4?
• In what proportion should Japanese and US stocks be mixed?
• I am interested in both of your perspectives on monthly chart strategies.
• For long-term investment, what is the typical acceptable drawdown?
• To maximize tax benefits, should one invest in high-beta stocks rather than indices?
• Within reasonable limits, I would like to hear your views on how to judge when features have "expired" and your perspective on lot sizes.
• Do you do anything specific during earnings season after looking at the results?