A 250 pip stop loss is a big ask. In our last post we shared how AI discovered that buying GBPJPY and USDJPY at Thursday's close — rather than Friday — nearly doubles the pip yield. The full strategy uses a 250 pip TP and 250 pip SL with a 5-day hold.
But a 250 pip stop on GBPJPY and USDJPY is not a small number. So before we build the algorithm, I wanted to ask a more uncomfortable question:
Can we reduce that risk without destroying the edge?
Back to Claude AI. I asked it to go back through the 6-year dataset and test 7 different trailing stop distances — from 30 pips right out to 200 pips — and measure what happens to total pips, win rate, and year-by-year consistency for each.
The answer was nuanced. And actually quite useful.
a 250 pip fixed stop?
Key Finding
Any trailing stop completely eliminates all hard 250 pip SL hits — zero on both pairs, regardless of trail distance. Over 6 years, the fixed stop got hit 37 times on GBPJPY and 29 times on USDJPY. Those are replaced by smaller trailing exits, not full 250 pip losses.
Pair-by-Pair Breakdown
GBPJPY — Fixed stop wins
The fixed 250/250 stop is still the best performer at 9,813 pips. Trailing stops hurt this pair. GBPJPY makes big weekly moves and tight trails get shaken out regularly. The 37 hard SL hits are painful but the strategy's overall edge outweighs removing them.
USDJPY — Trail 130 pips is the sweet spot
A 130 pip trailing stop delivers 7,187 pips with zero losing years across the entire dataset, eliminating all 29 hard SL hits while barely denting the 7,374 pip fixed baseline. On this pair, the trail genuinely earns its place.
The Conclusion
Run them differently. Fixed stop on GBPJPY. 130 pip trailing stop on USDJPY.
Two pairs. Same entry logic. Slightly different risk management. Both optimised individually rather than treated as one size fits all.
This kind of granular parameter testing — 7 trail distances × 2 pairs × 6 years of data — would have taken days manually. It took under 10 minutes with AI.