New Here? Start Here

This page explains how to read AlpAI Lab: where to start, which kinds of articles we publish, and how to read our records. AlpAI Lab is a research publication that shares records of AI-assisted company analysis, along with trading and testing done with our own funds. How do we read a company's earnings and disclosures, what do we decide, and what results follow? We check investment ideas not only through analysis, but also through actual trading and data.

For example: "If we let the AI think longer, do investment decisions get better?" or "How do results change when investments in stocks with similar price movements overlap?" In our research articles, we present questions like these together with the comparison conditions, results, and limitations.

What we invest in, how it is allocated, and our trading rules are revised in light of test results and other findings. Decisions and trades at a given point in time appear in dated articles, and whole-account performance can be found on the Track Record page.

The One to Read First

Does limiting investment in stocks that move similarly improve results?

What happens when a purchase-time rule limits the concentration of investments in stocks with similar price movements? Through comparisons including a case with no cap, the article explains the difference between a gap appearing in the numbers and being able to conclude that "this is a good rule."

You can read the full text without registering or logging in.

Choose by what you want to read

To see the performance of the actual account, go to Track Record. On a page separate from the articles, you can check the period covered, the comparison conditions, and interim drawdowns (declines from an earlier peak) together. The S&P 500 shown for comparison is a yardstick for looking at performance over the same period; it does not represent our holdings or allocations.

Four points to keep in mind when reading the records

The target date and the publication date are different. Daily covers a completed US trading day. English editions follow the Japanese issues after review. Check the trading date, the publication date, and the as-of date in the performance section.

A decision and a fill are different. "Deciding to buy" is not the same as a trade being executed (a fill). When we include confirmed fills, we show them separately from the pre-trade decision. Our decision records are not instructions for readers to make the same trades.

Not every method tried in research is adopted. We also present methods we did not adopt as a result of testing. An article is a record of its target period, and we do not necessarily still trade the same stocks, allocations, or trading rules today.

Live trading, simulated trading/score observation, and historical-data testing are different. Live trading means decisions and fills actually made in our own account.

Simulated trading/score observation is tracking of selections or rankings recorded before the results, without real money. Historical-data testing is a calculation that applies a rule to past data. These three sets of figures are not added together. Score and estimate results do not indicate the probability of future price gains or profits.

What you can read

Research articles and the Track Record can be read without registering or logging in. Daily decision records (Daily) and weekly reviews (Weekly) can be read in full for 30 days after publication with a free registration (an email address and answers to a 4-question survey).

Articles open to everyone after the 30-day early-access period. Free members can continue to read older articles.

We currently do not sell any paid plans. For details, see Free Registration Guide.

How we handle information, along with important notices, is covered in Research and Publication Policy and Notices, and the operating company and how articles are made are covered in About Us.

Japanese original of this guide.