Misoka001

MISOKA (Japanese for "the last day of the month") trades **only the last 3 business days of each month**. At month-end, passive funds, pensions and corporates crowd their monthly rebalancing into the 16:00 London WM/Reuters fix; the pre-fix overshoot reverts after the fix — and this structural bias is strongest at month-end. In my measurements, the same fix-reversion logic earns about +$6 per trade on normal days but **+$30 per trade at month-end — roughly 5x**. MISOKA extracts only those richest three days: a low-frequency, high-purity design. ■Logic: only when the 45-minute pre-fix run-up exceeds 2.5x ATR, enter once against it 2 minutes after the fix; TP at a 40% retrace, ATR-based SL, 45-minute time stop. One trade per symbol per day; completely idle outside month-end. No martingale, grid, or averaging. ■Twelve pairs from one chart: ten validated pairs ON by default (CHFJPY/GBPJPY/AUDUSD/GBPUSD/AUDJPY/USDCHF/USDJPY/EURJPY/USDCAD/EURUSD). ■Results (raw-spread data + $7/lot round-turn deducted, Jan 2010 – Aug 2026, $10,000, 0.5% risk): 10-pair portfolio **PF 1.60, +$36,860, 1,938 trades**. Individually CHFJPY PF 2.48, GBPJPY 2.06, AUDUSD 1.78, etc. Honest note: 2012, 2014 and 2024 were slightly negative years. ■Double validation: every trade was reproduced on REAL TICKS (actual price paths), refunding the actual spread and charging RAW + $7/lot — **all 10 pairs positive, aggregate PF 1.45** (929 trades). This rules out the M1-model optimism bias. A trade-for-trade identity check between the multi-symbol build and the single engine also passed (158/158). ■Theory and transparency: month-end fix flow concentration is documented in academic literature. Built with Anthropic's Claude Code as a development partner; six sibling hypotheses (quarter-end amplification, fiscal repatriation, Friday weekly fix, pre-fix momentum, turn-of-month reversal, extended holds) were all tested and honestly rejected as adding no independent edge (extended holding is available via parameters: raising MaxHoldMin/RetraceTarget lowers PF but raises profit per trade). GOLD has no month-end fix flow and is deliberately not included. ■Account: a RAW/ECN low-spread account is required. Default FixHourServer 18 matches 16:00 London year-round on GMT+2/+3 servers with US DST switching. ■How to use: attach to any M1 chart (GBPJPY recommended), the ten defaults run as-is, enable algo trading. It does not trade outside month-end (3 trading days per month, ~20 trades/month on average). The free version trades a fixed 0.01 lot, no compounding. Past performance does not guarantee future results. Test on demo first; VPS recommended.