Financial Instruments Business Operator Kanto Finance Bureau Director (Kinsho) No. 1960 / Membership Association Investment Trusts Association, Japan Membership No. 012-02323 Go to GogoJungle Home
icon List
Yohaku
10/26/2026 18:30
Release： 10/26/2026 18:30
Update： 10/26/2026 18:30
Yohaku
Investment style : Day Trade
Investment career : More than 10 years
Investment amount : 5 million - 10 million JPY
We do not build tools to predict the market. We are creating MT4/MT5 tools to reduce trading effort and mistakes, such as lot sizing, migrating to MT5, and monitoring restarts. All of these are things I thought would make trading easier when I was actively trading FX. Tools that handle signals determine only from confirmed bars, and before issuing, they are verified through automated testing and continuous operation on a demo account.
Article cover image

We tested GOLD's standard logic in 26 variations, using 13 years of 1-minute bars with trading costs included: in this procedure, we did not find an edge that would be ready to sell.

This is Yohaku, creating MT4/MT5 indicators and useful tools. The things I create are decided from GoGoJan reviews and questions, as well as the complaints I hear on X.

I wondered if I could create GOLD signals, and checked 26 common shapes using the same procedure. The result is that none of them could be considered suitable for selling. Although the validation didn’t go well, I think how I tested and where I drew the line can be useful as a way of looking at signal selection, so I’ll write everything down.

First the conclusion

  • On the 1-minute chart of XAUUSD, I tested 26 shapes including trading costs. When aligned over 2010 to 2022 (13 years), the profit factor (total profits ÷ total losses) ranged from 0.61 to 1.07, and none of them exceeded the line I had set (such as 1.25 or higher).
  • Shapes with more occurrences tended to fall below 1.0, and the good numbers for shapes with fewer occurrences could be explained by adjusting for the number of attempts.
  • With this procedure, these 26 shapes did not yield a positive result. This does not mean the signals themselves have no meaning.

Conditions tested

  • Data: 1-minute bars of XAUUSD from MT5 demo account (June 2004 to October 2026, 7,065,883 bars). The 13-year comparison used January 2010 to December 2022.
  • Time normalization of data: The server time of this account was New York time plus 6 hours until May 6, 2011; during weeks when daylight saving changes in the US and Europe caused a shift (roughly two weeks in March and one week from late October to early November each year), it was an extra -1 hour, so I aligned all periods to New York +7 hours and counted (recounted on October 15, 2026).
  • Judgments were only for confirmed bars. Entries occur at the opening price of the next 1-minute bar after the signal bar closes.
  • Buys enter at the ask price and exit at the bid price; sells enter at the reverse. Spreads use the values recorded on the 1-minute bars, and I shifted the stop executions slightly unfavorably.
  • If both stop loss and take profit were hit within the same 1-minute bar, it was counted as a stop loss.
  • Only prices on the higher time frame that were confirmed at that time were used (no looking ahead).
  • Immediately after sharp moves, when price action is unusually volatile, or when the spread exceeds three times the usual, or during thin trading periods, I decided not to enter any shape.

Shapes tested

  • Range of Tokyo time with moves in the direction of the London early session
  • After a pullback in line with the higher-time-frame trend, break of the most recent high or low
  • After the price range shrinks, move in the breakout direction (two methods to measure Bollinger Band width and previous day's range)
  • Break of the previous day’s high or low
  • Revert toward the mean if far from the moving average or move further away if already distant
  • Break of the week's high or low (trend-following)
  • When the timing bar of US indicators moves significantly, follow its direction
  • Ride the direction of the early-day move into the latter part of the day
  • Enter at a fixed time and exit at a fixed time (time-of-day bias)

Some entries were counted as 26 shapes by varying stop loss and take profit widths and measurement periods across 3–5 variations. The first eight shapes were verified and dropped from consideration in 2018–2022, and the remaining 18 shapes were predetermined with numbers before testing, then compared over 2010–2022 once. All attached trades are aligned to 13 years for all 26 shapes.

Results

  • Shapes with more occurrences: follow-the-trend pullbacks in higher time frames occurred about 2,500 times with PF 0.89 to 0.93, and shapes entered and exited by time occurred about 2,200 to 3,200 times with PF 0.76 to 0.96.
  • The best results were pullbacks aligned with the higher time frame trend (one of the first eight shapes): 256 trades with PF 1.07, and one shape that followed the direction when US indicators’ time bars moved significantly: 231 trades with PF 1.06. In both cases, when the spread was doubled, PF dropped below 1.0 (to 0.94 and 0.97), and there were years within the 13-year period where losses occurred for more than four years.
  • The stability of daily profit and loss (Sharpe ratio) adjusted for the number of attempts (26 shapes, adding a re-counted 26 that accounted for time, totaling 52) using the Deflated Sharpe Ratio, did not reach a level that could be described as “not by chance.”

Why this happened

  • GOLD has large price moves and spreads move as well, so shapes with small win/loss differences are almost erased by trading costs.
  • Even if a good number appears with a shape that has few occurrences, testing 26 shapes makes one or two look good by chance. The adjustment is to account for that.
  • As an average, even if there is a slope, a few large moves can pull the average, and when stop loss and take profit are placed, the rest remains unprofitable.