diff --git a/ideas/idea-069-eurgbp-late-session-mean-reversion-5m.md b/ideas/idea-069-eurgbp-late-session-mean-reversion-5m.md
new file mode 100644
index 0000000..044f502
--- /dev/null
+++ b/ideas/idea-069-eurgbp-late-session-mean-reversion-5m.md
@@ -0,0 +1,44 @@
+# Idea 069: EURGBP Late-session mean reversion
+
+## Summary
+
+Document a PyP Quant strategy candidate for EURGBP on 5m candles using a HistGradientBoostingClassifier approach focused on late-session mean reversion.
+
+## Market Hypothesis
+
+EURGBP may show repeatable behavior when late-session mean reversion conditions align with volatility, trend, and recent structure filters. The first implementation should stay conservative and treat this as an educational research candidate, not a production trading recommendation.
+
+## Candidate Features
+
+- Recent return over 3, 6, and 12 bars.
+- ATR-normalized candle range and close location value.
+- Rolling volatility percentile over 50 and 200 bars.
+- Distance from EMA 20, EMA 50, and EMA 200.
+- Session flag or market-hours bucket where applicable.
+- Prior swing high and swing low distance.
+
+## Label Design
+
+- Predict whether the forward 5m move exceeds an ATR-normalized threshold.
+- Use neutral labels when the forward return is too small to justify risk.
+- Tune the horizon separately for trend, range, and scalp variants.
+
+## Risk Controls
+
+- Require minimum confidence before emitting UP or DOWN.
+- Block signals during abnormal spread, missing candles, or low-liquidity windows.
+- Cap exposure per instrument and stop after a daily drawdown limit.
+- Validate stop-loss and take-profit levels with PPE simulation before live use.
+
+## Backtest Checklist
+
+- Compare against HOLD and simple moving-average baselines.
+- Run walk-forward splits with no future leakage.
+- Inspect trade count, profit factor, max drawdown, win rate, and average adverse excursion.
+- Review behavior by session, volatility regime, and weekday.
+
+## Implementation Notes
+
+This idea is intentionally Markdown-only. A future template can add `strategy.py`, `quant.config.json`, and a focused README once PPE results justify turning the idea into executable code.
+
+Closes #137