Does Limiting Investment in Stocks That Move Alike Improve Results?

Even if you hold many stocks, they can fall together when their price movements are similar. So does keeping investment from concentrating in stocks that move alike lead to better results?

At AlpAI, we simulated a rule that sorts stocks with similar price movements into groups and limits how much can be invested in the same group at the time of purchase. From these results alone, we could not conclude that setting a cap improves results. There were numerical differences between the 25% and 30% caps, but not enough to decide which one to use.

This is a historical-data simulation run on September 27, 2026. It does not align the stocks covered, costs, or trade timing with actual operations, so these results alone cannot tell us what effect the rule would have in live investing. Below, we explain how the cap works, what was compared, and what the results can and cannot settle.

What is limited is the amount added to stocks in the same group

A 25% cap means that the combined amount already held and newly purchased in the same group must stay within 25% of total account assets. It is not a cap on a single stock or on a single purchase amount.

If a purchase would push the group over the cap, the purchase amount is reduced. If the group has already reached the cap, additional purchases in that group are skipped. If a group's share rises above the cap because of price gains or similar, this rule alone does not trigger a sale.

To form the groups, we used price movements over the 120 trading days before each month began, and created 11 groups every month. These are groups based on how similarly prices move, not an industry classification. The test uses price-based groups rather than historical industry classifications. Its results therefore do not establish how industry caps would perform.

We compared "whether to set a cap" and "25% versus 30%"

We examined two questions using simulations under different conditions. Both cover 2020–2025 and compare 10 variations, with the start date shifted by 5 trading days each time.

In the 513-stock comparison, 3 stocks with short histories were excluded for the entire period. In the 516-stock comparison, eligibility for purchase was determined month by month, and only 25% and 30% were compared. The two comparisons cannot be linked to measure the effect of introducing a cap.

The results of the two tests cannot be compared directly with each other. Within each test, we compare results under the same conditions.

Compared with no cap, the improvement at 25% was not consistent

In the supplementary 513-stock test, the results were as follows. The annualized return difference is the average of the comparisons for each start date across the 10 variations.

The 25% cap was positive on average, but it outperformed no cap in only half of the start dates. The 30% cap was negative even on average. Under these conditions, the results do not consistently show that "setting a cap improves results."

Simulations with shifted start dates include a lot of overlapping periods. "5 of 10" does not represent 10 independent runs, nor a 50% win rate in the future. Note also that the unit of the difference is "percentage points," not "%."

25% outperformed 30%, but that does not yet make it the better setting

Next is the 516-stock test. After aligning the eligible stocks for each month, we compared caps of 25% and 30%.

Numerically, 25% came out ahead. However, this comparison has no no-cap result. Nor can the 10 start dates be treated as independent samples. This result alone cannot settle whether to introduce a cap or whether 25% is superior.

We also checked the size of the declines. Maximum drawdown is the largest percentage drop from the previous peak. The averages for the two conditions were nearly the same, but depending on the start date, 25% was worse, by as much as 1.70 percentage points. The effect of reducing declines was also not seen for every start date.

What these results can and cannot decide

From these simulations alone, we cannot decide to adopt a cap as a better rule than no cap, nor choose between 25% and 30%. This does not mean we have shown that the two produce the same results, nor that diversification in general has no effect.

What we tested is this specific rule, which limits additional purchases in groups built from price movements. If we investigate further, the starting point would be to compare no cap, 25%, and 30% using the same eligible stocks and the same trading conditions.

Remaining limitations in reading the simulations

• This does not test caps by industry. Results for groups based on price movements cannot be applied as-is to an industry-based rule.
• The test does not establish that every input was available before the corresponding trade. Groups use prices from before each month began. Even so, the test leaves a risk of using information that a trader would not yet have had.
• This is not a comparison under the same conditions as actual operations. The stocks covered, purchase conditions, measured trading costs, and decision timing have not been aligned with actual operations. Some parts use prices from the same day on which the conditions were evaluated. The simulation therefore does not establish that those trades could have been executed at those prices.

This article shares the content and limitations of the simulation and does not recommend that readers use a 25% or 30% cap. Past simulations do not guarantee future results.

Related public materials

General cautions for reading backtests and investment performance are also available in the U.S. Securities and Exchange Commission's investor guidance. This material is a general reference for reading such results and is not a source that verified AlpAI's simulation results.

Japanese original.

Browse all English research