[Development Note] How EA Handles the Day of the Week — Design considerations for opening gaps, triple-digit swap days, and weekend rollover
【Key Points】
FX has day-of-week tendencies, and in EA design, three aspects—window opening, swaps, and weekend rollover—tend to be linked to the day of the week.
However, days with swap multipliers (e.g., 3x swaps) differ by currency pair and broker. You need to check according to your account specifications.
“Day-of-week bias” is an observation, not a future guarantee. In design, decide in advance whether to adopt a bias-avoiding or bias-utilizing approach.
■ 1. Premise: Structure of the FX market week
FX trading is basically continuous on weekdays. On the other hand, markets stop on the weekend. this “stoppage” is the starting point that creates weekday-specific patterns.
As the table shows, design considerations concentrate on Monday, Wednesday, and Friday. Note that the table reflects general tendencies, and differences may exist depending on broker or instrument.
■ 2. Monday: Gap opening
The Friday close and the Monday open are not necessarily the same. This difference is called a gap. If there is news over the weekend, Monday’s opening price can jump significantly.
People say “gaps will be filled,” but that does not always happen. Therefore, building logic assuming gaps will be filled should be avoided.
There are two main design choices: avoid gaps, or aim for right after gaps.
Lab’s GBPJPY Anomaly v03 falls into the latter.
Buy at 01:30 every Monday and close 60 minutes later, or set a 40-pip stop loss. 2
Backtest from January 2016 to December 2025 shows 298 total trades, a profit factor of 2.93, a win rate of 71.14%, and a maximum drawdown of 0.55%.
About 60% of orders are placed on Mondays, totaling around 30 per year.
Note that these numbers are “past observations.”
Day-of-week biases can be observed but are not guaranteed to continue. Also, the number of trades is not very large—about 300.
We should also check whether the numbers are being shown to look good by sacrificing trades.
Another approach is to base the design on “the first bar when the market reopens” rather than a fixed time, as in WeekOpen Portfolio.
Because it detects weekend closures and moves, it is less dependent on server time drift.
■ 3. Wednesday: Swap 3x day
For many currency pairs, Wednesday rollover adds three days’ worth of swap.
This is because the days spanning the weekend are accumulated on Wednesday. If swaps are positive, you receive 3x; if negative, you pay 3x.
However, the day that gets 3x may differ by currency pair, commodity like gold, indices, or broker.
Do not assert this; you need to check your own account specifications.
In EA design, the impact is larger for logic that assumes rollover. For short-interval settlement logic, the impact is minimal. The lab has a tool (the Devil Wednesday alert) that warns when running on Wednesdays.
However, this is a safety notice, not a measure of a strategy’s superiority.
Swap levels themselves are reviewed by brokers from time to time, so you cannot simply conclude after one check.
If you operate rollover-based logic, regular review of specifications is desirable.
■ 4. Friday: Handling of weekend rollover
Friday design considerations are whether to rollover into the weekend.
If you rollover, you may be affected by Monday’s gap opening.
It can move in favorable or unfavorable directions.
Lab’s USDJPY Squeeze-Momentum v6 is designed with no weekend rollover (weekend flat).
WeekOpen Portfolio likewise has no weekend rollover.
Therefore, the gap is excluded from the start as a risk consideration.
Also, late Friday trading tends to decrease, and the US unemployment statistics occur on the first Friday of each month.
On days with large price movement, spreads tend to widen, so be mindful of transaction costs as well.
■ 5. Weekday x Time: combining biases
Biases by day alone can be too rough.
Lab’s GBPJPY Anomaly v04 defines 110 cells combining day and time, and trades only on cells showing bias.
From January 2016 to December 2025, total trades were 3,025, profit factor 1.51, and forward validation was positive for all 10 cycles.
However, 2016 and 2018 were annual losses, so not every year is profitable.
Combining day and time allows finer trend capture, but as cells increase, the risk of picking up random biases also increases.
This needs separate verification to avoid over-optimization.
■ 6. Server time drift
MT4 server time shifts by one hour between summer and winter time.
For example, GBPJPY Seam Morning v1 uses server time: 01:00 in summer, 02:00 in winter.
Day boundaries and “Monday morning first thing” views are affected by this as well.
Both EA design and operation should assume server time.
■ 7. SummaryThe figures cited in this article are all backtest results and do not guarantee future performance.
Whether day-of-week bias continues must be confirmed in forward testing.
Days of the week are not calendar cutoffs but an element of design.
Whether to avoid gaps or target them. Whether to avoid or accept swaps. The choice appears for each weekday.
Treat weekday tendencies as observed biases, verify them, and record them.
We, Semura Lab., will continue to maintain a stance of verifying biases as facts rather than deciding them in advance.
If you have concerns about EA design or verification approaches, feel free to consult at the All-Purpose Consultation Center.
https://www.gogojungle.co.jp/mypage/community/485