Do fallen stocks bounce back? What I learned from testing contrarian swing trading

In this study, I tested swing trading with a holding period of several days, using indicators such as deviation from the moving average, deceleration of price declines, and historical price movement cycles.

The concept is a contrarian approach: selecting stocks that are declining but appear likely to reverse, and buying them at a low price. I set profit-taking and stop-loss levels, as well as maximum holding days, while also accounting for the number of shares that could actually be ordered and the limits of available capital.

What I felt after seeing the results was that “it is difficult to buy individual stocks at a low price.”

For the period from April to September 2026, using the conditions selected from the previous six months—’take profit +3%, stop loss -2%, maximum 5 business days’—the result was approximately +478,000 yen after costs. This is not a figure based on selecting the best conditions after the six-month period ended. However, it does include unrealized gains at the end of the period.

Looking at this period alone, I feel there is sufficient potential.

However, when I updated the settings every six months and tested the past seven periods, only three periods were profitable. In an operation that continues holding across half-year periods, the cumulative profit from April 2023 to September 2026 was approximately +282,000 yen, with a maximum drawdown of approximately 17.2%.

Even if there were profitable half-year periods, I needed to confirm separately whether it was possible to consistently accumulate profits over a long period.

The win rate for this half-year was about 48%. While this figure does not directly represent the ‘probability that a declining stock will fall further,’ it means that even when buying with the expectation of a reversal, about half of the trades ended in a loss.

Even including periods where the TOPIX rose, providing a tailwind for buying across the market, individual stocks that have fallen do not necessarily bounce back. I felt that if I were to use the same contrarian approach for selling, it would be even more difficult in a bull market.

I also tested filters to avoid consecutive losses.

Buying only on days when the TOPIX is above its 5-day moving average. Buying only stocks whose sector has shown positive recent price movement. Limiting to stocks with positive historical evaluations.

Looking at recent losses, one wants to add such conditions. In fact, there were conditions that could reduce losses in specific periods.

However, when applied to a long historical period, these conditions sometimes erode profits as well. The condition of the TOPIX being above the 5-day line worsened the cumulative profit and loss for the entire period. The condition of limiting to stocks with positive evaluations improved cumulative profit and drawdown, but it could not prevent the consecutive losses in October.

It is necessary to confirm not only whether losses were avoided, but also how much profit was missed by sitting on the sidelines.

I thought this also applies to the rules I usually set for my trading, such as ‘I won’t buy on days like this in the future’ or ‘I won’t enter under these conditions.’

Were the rules I decided on after recent failures effective in the past as well? Are they conditions that only avoid this recent loss? Verification is necessary even when adding rules.

This study is basically a contrarian swing trade assuming a range-bound market. I have not yet tested trend-following swing trades.

A major takeaway from this study was realizing that this strategy has strict conditions for stable operation before committing real capital.

Beyond just searching for conditions that generate profit, it is important to understand the weaknesses of the strategy you intend to adopt. This verification process was useful for that purpose as well.

I used up my 5-hour ChatGPT limit on this verification. Although I ended up splitting the analysis between the morning and afternoon, I could not find an effective strategy.
However, since I was able to obtain valid data, I am leaving this note as a record.