What this is
A weekly report from our lab that judges "automated scalping" with measured data, not marketing.
- We test 24 pre-registered micro-session cells (Tokyo open, Tokyo fix, London open, round numbers, rollover, z-score extremes) holding 5–90 minutes. Sub-minute scalping is excluded up front: retail round-trip costs (0.7–1.1 pips) and 60–200ms fills make it arithmetically dead.
- Costs are not advertised spreads: we use hour-of-day spread percentiles measured on our own MT5 account (p75) plus commission auto-derived from deal history.
- Every cell runs SHADOW (virtual fills) first and is promoted to live orders only after a pre-registered statistical gate (n≥100, PF≥1.1, z≥1.65). Every trade, win or lose or skipped, is published here.
Current verdict (24 backtest cells)
Surviving cells: 0 of 24. None of the pre-registered cells cleared "statistically significant positive after measured costs". That is the honest result. The 05:35 JST daily recalibration keeps re-judging as data grows; any survivor automatically enters SHADOW verification.
| Cell | n | Net (pips) | PF | Cost share |
|---|---|---|---|---|
| EURJPY London-open 15min follow | 743 | -737.8 | 0.89 | 18% |
| USDJPY round-number fade | 2523 | -2456.9 | 0.88 | 6% |
| EURJPY round-number fade | 2703 | -3467.9 | 0.83 | 10% |
| EURUSD round-number fade | 1547 | -2008.1 | 0.78 | 8% |
| GBPUSD London-open 15min follow | 745 | -1515.4 | 0.73 | 18% |
| USDJPY London-open 15min follow | 802 | -1678.3 | 0.73 | 18% |
| GBPUSD round-number fade | 2360 | -5083.1 | 0.71 | 10% |
| EURUSD London-open 15min follow | 737 | -1440.8 | 0.68 | 19% |
| USDJPY Tokyo-fix drift | 1449 | -3263.6 | 0.61 | 29% |
| USDJPY post-fix reversal | 1216 | -3330.5 | 0.60 | 25% |
| EURJPY post-fix reversal | 1113 | -3061.5 | 0.59 | 32% |
| USDJPY Tokyo-open impulse fade | 997 | -2699.0 | 0.56 | 29% |
| EURJPY Tokyo-fix drift | 1333 | -4203.9 | 0.52 | 35% |
| USDJPY M1 z-score extreme fade | 1690 | -4116.4 | 0.52 | 29% |
| EURJPY Tokyo-open impulse fade | 887 | -2961.7 | 0.48 | 37% |
| GBPUSD M1 z-score extreme fade | 1416 | -2942.5 | 0.45 | 42% |
| GBPUSD Tokyo-open impulse fade | 868 | -1754.3 | 0.44 | 52% |
| EURJPY M1 z-score extreme fade | 1383 | -4657.5 | 0.41 | 38% |
| EURUSD Tokyo-open impulse fade | 910 | -1763.1 | 0.41 | 48% |
| EURUSD M1 z-score extreme fade | 1475 | -2838.8 | 0.39 | 44% |
| USDJPY post-rollover fade | 752 | -2938.1 | 0.24 | 90% |
| EURJPY post-rollover fade | 785 | -4251.6 | 0.15 | 112% |
| EURUSD post-rollover fade | 750 | -1937.5 | 0.10 | 129% |
| GBPUSD post-rollover fade | 648 | -2363.9 | 0.10 | 144% |
Most instructive finding: the Tokyo-fix drift (USDJPY 9:30→9:55 upward bias) is real (+0.5p/day gross) — but the measured spread in that hour is ~5x larger. The edge exists; costs eat it. That is what scalping actually looks like. London-open follow came closest (PF 0.74–0.90), yet every cell lost across every year block.
Measured cost reality (the real battlefield)
| Pair | Spread now | Measured p75 | RT commission (pips) | Commission source | RT cost (p75) |
|---|---|---|---|---|---|
| USDJPY | 1.20 | 1.20 | 0.00 | measured from deals | 1.20 |
| EURUSD | 0.60 | 0.60 | 0.00 | measured from deals | 0.60 |
| EURJPY | 1.40 | 1.70 | 0.00 | deals (any symbol) | 1.70 |
| GBPUSD | 0.70 | 1.07 | 0.00 | measured from deals | 1.07 |
Measured on: Exness Technologies Ltd (Exness-MT5Trial5, account currency JPY). Spreads are measured percentiles with stale ticks excluded; commission is auto-derived from deal history (falling back to the official spec table).
This week’s trades (SHADOW + live, all of them)
This week had zero trades and zero skips (quiet weeks get reported too).
The per-trade detail (time, cell, shadow/live, side, exit reason, net pips, plus the breakdown of skip reasons) and the reason for every skip are published in our verified Discord channel.
Why it is gated: the detail is the execution record of a real account — the most valuable thing we can hand a reader. We give it to people who opened their account through this site (we only match your account number against our partner ledger; no email, no password). The cumulative record, win/loss split and costs above stay public on this page — we do not show you the wins and hide the losses.
→ See how to open an account and verify
Auto-learning status
Every day at 05:35 JST we re-evaluate the 24 pre-registered cells on incrementally updated M1 data, swapping survivors in and auto-demoting decayed ones (no open-ended hypothesis mining — that is how you overfit). Current survivors: 0. Last re-evaluation: 08/10 05:35 JST.
Disclaimer
This is not investment advice. All numbers are measured on our own verification account and do not guarantee future results. Scalping is the trading style most dominated by transaction costs; in most environments its expectancy is negative.
Related articles
Does Automated Scalping Survive Real Costs? Live Lab Report 2026-W32
We judge 24 micro-session scalping cells daily against measured MT5 costs (hour-of-day spread percentiles + auto-derived commission). Surviving cells: 0. Primary data, every trade published.
Measured broker spreads 2026-W33 — logged hourly from our own MT5 accounts
Not a copy of published spread tables: 1,914 hourly bid/ask samples taken on our own accounts (XM (real money)) over 2026-08-02 → 2026-08-09, broken down by pair and by hour of day.
US-Close FX Live Report 2026-W33 — every order from the Tokyo-open fade and dollar-beta systems
The full record of two systems trading only on the prior-day US500 move: 8 closed to date, averaging +0.22R per trade. Wins, losses and skips all published.