The diagnostic values and performance as an EA were separate things
EA for EURUSD "Boundary - KYOKAI -" is completed: what was tried, what failed, and what remained. This is a record of the validation process.
» Go to the product page for EA "Boundary - KYOKAI -"
What I did at the end of the validation was to decide not to use the discovered strong correlations.
The entry and exit shapes had almost solidified. From here, there was still a sense that it could be improved. However, from previous tests, adding conditions does not necessarily improve things. So in the final stage I did not add trading conditions; I only added diagnostics. The trading logic remained the same just by looking at it.
There was a clear difference in Pivot movement speed
What I looked at was how fast the Pivot (high/low) that formed the horizontal line moved just before. I examined 41 cases where, in 2024 BUY, it was decided which of the two directions, ±40 pips, arrived first.
| Pivot Low speed > 0 18 wins out of 27 |
66.7% | |
| Pivot Low speed < 0 4 wins out of 14 |
28.6% |
※ The bar length corresponds to the win rate.
The moving speed of Pivot Low for winning trades was +1.02 pips/hour at the median. For losing trades, it was almost zero. When the low was rising, buying tended to win; when it wasn’t moving, it tended to lose. A fairly clean distinction.
If this criterion had been included, backtest results should have improved. Still, I did not include it.
- Small sample size. Of 41 cases, only 14 had negative speed
- Only results for 2024 BUY; unclear whether it applies to other years or SELL
- Adding conditions makes it fit past data too well. I wanted to avoid overfitting
The correlation found and what can be included in trading conditions are not the same thing.This diagnostic was not included in the final version.
EURUSD-specific, M15, horizontal-line First Touch reversal EA "Boundary - KYOKAI -". All conditions used are listed in the manual.
» See the product page for Boundary - KYOKAI -
Where will it extend after +2 ATR?
Another diagnostic concerns the exit. After profit reaches +2 ATR and trailing starts, how far will the price extend further? The exit itself was not changed.
| 2024 | 2025 | |
|---|---|---|
| Percentage that extended to +2.5 ATR | 60.7% | 61.8% |
| Percentage that extended to +3 ATR | 37.7% | 41.2% |
| Median time to reach +2.5 ATR | About 23 minutes | About 50 minutes |
| Median time to reach +3 ATR | About 1 hour 26 minutes | About 2 hours 8 minutes |
※ Diagnostic for extension after reaching +2 ATR.
After reaching +2 ATR, extending to +2.5 ATR happens in about 60% of cases. Reaching +3 ATR drops to around 40%.
Even with this result, I did not change the exit. It trails from +2 ATR, keeping the distance at 1 ATR. Just because the diagnostic revealed something, I did not immediately alter the specification. That was the final decision.
What remained in the final version and why
| What remained | Why it was kept |
|---|---|
| First Touch after line formation | The M15 +5 pips attainment rate was 63.7% at drawing time, while First Touch was 81.1% (Episode 1) |
| M15 | First Touch response was slightly better than M5 (Episode 1) |
| BB2.9・ADX22.5 | RSI and Stochastics dropped out, BB・ADX combination had PF 1.22 (Episode 2) |
| Limited higher-timeframe filters | Total ban on BUY decreased by about 40% in 2025. Limiting some improved all three years (Episode 2) |
| H4 ATR-driven SL and trailing | Extending TP lowered PF, adding averaging widened losses by 1.5x, fixed-amount trailing yielded roughly the same profits (Episode 3) |
| EURUSD-only | PFs of the other three currencies were 1.06, 1.07, 0.94. EURUSD had the lowest ATR and the highest rate of reaching 2 ATR first (Episode 4) |
What I learned through 5 episodes
- Touch alone did not provide enough edge.When focusing on the returning First Touch, the difference finally appeared
- Increasing conditions did not make the EA stronger.What remained was BB and ADX; H1 PO had no effect
- Win rate alone could not determine things.There were cases where PF rose even as win rate fell
- Good numbers did not guarantee adoption.Averaging additions or Pivot speed were such cases
- Diagnostic values and EA performance were not the same.Currencies with small diagnostic differences showed large differences when including the exit
EA formed after five validation rounds
Boundary - KYOKAI -
EURUSD-specific EA aiming for First Touch on horizontal lines with contrarian logic. It has taken its present shape after the validations described in this series.
| Target | EURUSD-only, M15 |
| Entry | First Touch reversal on horizontal lines (R1, R2, S1, S2) |
| Filters | Bollinger Bands・ADX・Higher-timeframe filters |
| Exit | H4 ATR-driven SL and trailing |
Thank you for reading this far.
This series is now concluded as a record of Boundary's validation. Going forward, I will continue with forward-test records and report the results.
※These are past validation results and do not guarantee future performance.