Londoncalling001

LONDON CALLING is a once-a-day mean-reversion EA that trades only the aftermath of the 16:00 London currency fix (WM/Reuters fix). Before the fix, real-money flows tend to push price in one direction; after the fix, part of that run-up tends to revert. LONDON CALLING trades this structural bias only on days when the pre-fix run-up over 45 minutes exceeds 2.5x ATR. It enters once, 2 minutes after the fix, against the run-up; take-profit is a 40% retrace of the run-up, stop-loss sits at the run-up origin plus an ATR buffer, and any open trade is closed after 45 minutes. Maximum one position per symbol, one trade per day — no martingale, no grid, no averaging down, ever. ■Fourteen pairs from a single chart: attach the EA to one chart (GBPJPY M1 recommended) and it simultaneously monitors and trades every pair you enable via ExtUse_ among the fourteen built in. Each pair has its own magic number and its own max-spread (x ATR) and slippage settings. One common parameter set for all pairs — no per-pair curve fitting. ■Six pairs are ON by default: in my testing (raw-spread data, M1, Mar 2020 – Aug 2026, $10,000 start, $7/lot round-turn commission deducted; past results never guarantee future performance) only pairs profitable in BOTH summer and winter DST seasons are enabled (individual results at 0.5% risk per trade): EURGBP PF 1.57, +$2,281, 154 trades; CHFJPY PF 1.33, +$893, 125; AUDJPY PF 1.28, +$2,777, 313; GBPJPY PF 1.19, +$1,935, 356; USDCHF PF 1.14, +$4,154, 507; AUDUSD PF 1.11, +$2,197, 423. Running all six defaults simultaneously at a fixed 0.10 lot gave PF 1.28, +$1,660, 1,878 trades, max closed-trade drawdown $321. ■Eight more pairs ship OFF by default: EURJPY (PF 1.09 but summer-skewed), NZDUSD (PF 1.11 but thin profit), EURUSD/GBPUSD/USDJPY/USDCAD (below threshold at $7/lot), CADJPY/NZDJPY (PF 2.9/2.1 but only 29/16 trades — sample too small). Yet at a $4/lot round-turn commission ALL FOURTEEN built-in pairs were profitable. The lower your commission, the more pairs you can switch on. ■Transparency: BTCUSD and GOLD were also tested and rejected (BTCUSD: the spread gate blocks every entry; GOLD: PF 1.01 even aligned to the 15:00 LBMA afternoon auction). I do not ship markets where the edge does not exist. ■Account requirement (important): the edge per trade is small, so a RAW/ECN low-spread account is mandatory. On a standard account with wide spreads the edge disappears (the spread gate will simply skip the trades). ■How to use (detailed): 1) open a GBPJPY M1 chart on a RAW/ECN account; 2) attach the EA — the six default pairs run as-is; 3) check FixHourServer: the default 18 matches 16:00 London year-round on brokers whose server time is GMT+2/+3 with US DST switching. On other brokers, verify which server hour equals 16:00 London in each season and adjust (the EA prints a startup warning to remind you); 4) set ExtSymbolSuffix to your broker's symbol suffix if any (the EA automatically falls back to the plain symbol name when the suffixed one does not exist); 5) if your commission is $4–5/lot or less, consider enabling EURJPY and others via ExtUse_; 6) enable algo trading. The free version trades a fixed 0.01 lot, no compounding. ■Development & validation: built with Anthropic's Claude Code as a development partner. Twelve scalping logics were tested across seven pairs (~200 backtests); London-fix reversion was the only survivor, then extended to a 16-symbol screening, commission and slippage stress tests, entry-hour distribution checks, summer/winter season-split validation, and a trade-for-trade identity check between the multi-symbol build and the single-symbol engine (356/356 identical). In addition, a 10-year out-of-sample test on 2010–2020 — a period never used for pair selection — showed ALL SIX default pairs profitable after the $7/lot commission (e.g. GBPJPY PF 1.52, USDCHF 1.34, CHFJPY 2.35); the full 16.6-year default portfolio (2010–2026, fixed 0.10 lot) delivered PF 1.32, +$4,638 over 4,281 trades. ■Note: this is a low-frequency EA — at most 6 trades/day with defaults (about 1.1/day on average), and some days have none. Past performance does not guarantee future results. Always test on a demo account first. A VPS for 24h operation is recommended.