Smart Money Concepts analysis for USD/JPY intraday trading.
Website: smcpulse.com | Built with Claude Code
Three questions the system answers each trading day:
- Should I be long or short? — Modules 01-07 aggregate macro, technical, positioning, and cross-asset signals into a weighted directional bias. Technicals drive the score (±3); macro and cross-asset provide context (±1 each).
- Where do I enter? — Module 08 identifies Smart Money Concepts (order blocks, FVGs, liquidity levels) across 4 timeframes and outputs a graded entry plan with ATR-based stops and position sizing.
- What are the realistic paths? — The 12h Playbook projects 3 probability-weighted price scenarios with distance-based intervention risk modeling.
The system also knows when to say no trade — if the best zone is too far away and there's no session catalyst, it outputs "NO SETUP" instead of forcing an entry.
External data flows through Python analysis modules, gets pushed to Supabase (data) and served via Vercel (dashboard) for delivery. Scheduling is handled by local cron.
| # | Module | Frequency | Data Source | Purpose |
|---|---|---|---|---|
| 01 | Macro Regime | Daily | FRED, MOF Japan | US/JP rate differential, JGB curve, DXY, VIX risk sentiment |
| 02 | Policy & Politics | Weekly | Web search | BOJ/Fed stances, intervention risk, political developments |
| 03 | Technicals | Daily | Yahoo Finance | SMA, RSI, MACD, Ichimoku |
| 04 | Positioning | Weekly | CFTC | COT net speculative position, crowding signal |
| 05 | Cross-Asset | Daily | FRED, Yahoo | Correlations (Nikkei, Gold, VIX, Oil), energy risk |
| 06 | Seasonality | Weekly | Reference data | Seasonal bias, flow events, trade balance |
| 07 | Checklist | Daily | Modules 01-06 | Intraday-weighted signal aggregation → direction + confidence |
| 08 | SMC Entry | On-demand | Yahoo Finance | Order blocks, FVGs, entry zones, 12h playbook |
| 09 | Scenario Monitor | Auto | Yahoo Finance | Live check + scorecard |
| 10 | Real-Time Alerts | Every 15m | Yahoo Finance | Zone approach, scenario triggers, intervention risk via Telegram |
| EA | SMC_USDJPY (MT5) | Real-time | Exness broker | MQL5 Expert Advisor — semi-auto and full-auto SMC execution |
The SMC_USDJPY EA ports the Module 08 analysis engine to MQL5 for real-time execution on Exness MT5. It runs on a USDJPY M5 chart and reads D1/H4/H1/M15 for multi-timeframe analysis.
| Feature | Details |
|---|---|
| Two modes | Semi-auto (pending orders for review) and full-auto (market orders with trade management) |
| Self-contained bias | D1+H4 structure determines direction — no external dependencies. Optional BiasOverride input. |
| Killzone filtering | Entries restricted to London (16-19 JST), NY (21-01 JST), Asian (09-12 JST, Scenario D only). Dead zone 00-07 JST. |
| Displacement required | Full-auto requires candle body > 1.5x ATR to confirm institutional intent |
| Zone types | Order blocks, FVGs, breaker blocks (flipped mitigated OBs), inversion FVGs (flipped traded-through FVGs) |
| OB refinement | H1 zone refined on M5 for tighter entries and smaller stops |
| Structure-based SL | At T1, SL trails M15 swings — not arbitrary breakeven |
| Internal/external liquidity | Targets alternate: sweep external → fill internal → sweep next external |
| 4 scenarios | A: Intervention Bounce, B: Trend Retracement, C: Liquidity Sweep (wick+close back), D: Tokyo Fix Fade |
| Safety | One position at a time, daily loss circuit breaker, spread check, magic number isolation, time-based exit |
| Chart visuals | Semi-transparent zone rectangles with overlap dedup, proximity filter (25 pips), max visible caps, on-chart dashboard panel |
Files: MQL5/Experts/SMC_USDJPY/ (7 files, ~4,700 lines)
SMC_USDJPY.mq5 ← Main EA: inputs, OnTimer pipeline, OnTick monitoring
SMC_Structure.mqh ← Swing points, BOS/ChoCH, market structure, bias
SMC_Zones.mqh ← Order blocks, FVGs, breaker blocks, inversion FVGs
SMC_Liquidity.mqh ← EQH/EQL, session levels, Tokyo fix, sweep detection
SMC_Scoring.mqh ← Confluence scoring, premium/discount, killzones
SMC_Trade.mqh ← Entry plans, position sizing, order management
SMC_Visual.mqh ← Chart drawing, dashboard panel
Setup: Copy MQL5/Experts/SMC_USDJPY/ to your MT5 data folder → Compile in MetaEditor → Drag onto USDJPY M5 chart. Start with Semi-Auto on a demo account.
Module 07 uses intraday-weighted scoring — technicals are the primary driver:
| Module | Weight | Rationale |
|---|---|---|
| Technicals (03) | ±3 (HIGH=3, MED=2, LOW=1) | Primary driver for intraday entries |
| Macro (01) | ±1 | Background context, changes weekly |
| Cross-Asset (05) | ±1 | Regime context |
| Policy (02) | ±1 | Weekly background |
| COT (04) | ±1 | Contrarian flag |
| Seasonality (06) | ±1 | Monthly context |
Daily max: ±5. Full weekly max: ±8. Daily conviction is scored independently — missing weekly modules don't cap conviction.
Module 08 includes integrated risk management:
| Feature | How It Works |
|---|---|
| ATR-based stops | max(ATR_1H × 0.5, 5 pips) — adapts to session volatility |
| Position sizing | Lots computed from account size × risk% × grade scaling factor |
| Grade scaling | A: 100% risk, B: 50%, C: 25%, D: 0% (paper trade) |
| No-trade filter | Zone >30 pips away + no catalyst → NO SETUP |
| Kill switch | FOMC/BOJ decision days suppress entries |
| Intervention probability | Distance-based: <158 → 1%, 158-160 → 3-8%, 160-162 → 10-20%, >162 → 25% |
| Grade | Score | Action |
|---|---|---|
| A | ≥5.0 | Full size entry |
| B | ≥3.0 | Half size entry |
| C | ≥2.0 | Quarter size (caution) |
| D | ≥1.0 | Paper trade only |
| NO SETUP | <1.0 | No entry |
Every report is cross-checked against 3+ external sources within 3 minutes of generation:
| Source | Indicators | Method |
|---|---|---|
| Yahoo Finance | Spot, SMA, RSI, MACD, Ichimoku, US 10Y, correlations | API + computed |
| Investing.com | RSI, SMA, MACD, JP 10Y yield | Web scrape (3-retry) |
| TradingView | Spot, RSI, SMA, MACD, Ichimoku | Scanner API |
| Derived | Rate spread (US 10Y - JP 10Y) | Computed cross-source |
16 indicators validated, 0 SKIPs. Results pushed to Supabase and displayed on smcpulse.com/validation.
- Setup (A/B/C/D) = market condition classification. A: Intervention Bounce, B: Trend Retracement, C: Liquidity Sweep, D: Tokyo Fix Fade.
- Scenario (Primary/Alternative/Tail Risk) = forward-looking playbook projections with probability weights.
Live at smcpulse.com
| Page | Purpose |
|---|---|
| Dashboard | Hero card, MTF alignment, risk alerts, liquidity levels |
| Reports | Daily + weekly modules with charts (date navigation) |
| SMC Analysis | Entry plan, 12h playbook, scenario paths, active zones |
| Scorecard | Module 09 accuracy tracking |
| Validation | Cross-reference results by module with PASS/WARN/FAIL/SKIP counts |
| Journal | Trade log with report context and performance stats |
| Pricing | Signal subscription tiers (Free / Alerts $29/mo / Full $79/mo) |
Features: auto dark/light theme, English + Traditional Chinese, mobile-responsive, real-time data from Supabase.
| Service | Role |
|---|---|
| Supabase | PostgreSQL database + file storage (Tokyo region) |
| Vercel | Next.js hosting + CLI deploy |
| macOS launchd | Scheduled report generation + Supabase push (11 agents) |
| Telegram | Real-time alerts + subscriber bot (onboarding, /bias, /status) |
| LemonSqueezy | Subscription billing ($29/mo Alerts, $79/mo Full) |
| Command | Description |
|---|---|
/usdjpy-daily |
Full daily analysis (Modules 01, 03, 05, 07) |
/usdjpy-weekly |
Full weekly analysis (all Modules 01-07) |
/usdjpy-entry |
Module 08 SMC entry zones + 12h playbook + chart |
/usdjpy-levels |
Quick reference: active zones and liquidity levels only |
/usdjpy-fix |
Tokyo Fix Fade check (best at 09:50 JST) |
/usdjpy-check |
Pre-trade checklist from cached data (no API calls) |
/usdjpy-cot |
Standalone CFTC COT positioning analysis |
/usdjpy-cb |
Standalone central bank policy analysis |
/usdjpy-monitor |
Module 09 live check against active playbook |
/usdjpy-scorecard |
Module 09 post-session accuracy scoring |
/usdjpy-alerts start |
Start real-time alert engine (Telegram + email) |
/usdjpy-alerts status |
Show armed zones, scenario tracking, alerts sent today |
/usdjpy-alerts test |
Send test alert to verify delivery channels |
/usdjpy-alerts reset |
Clear alert state and rearm zones |
/usdjpy-journal import |
Import trades from Exness CSV export |
/usdjpy-journal open |
Manual trade entry with auto-attached signals |
/usdjpy-journal close |
Close trade with pips/R:R calculation |
/usdjpy-journal review |
Performance summary and bias alignment analysis |
Reports run via macOS launchd agents (run_local.sh). Data is pushed to Supabase after each run.
| Time (JST) | Job | Days | Reports |
|---|---|---|---|
| 08:00 | morning |
Mon-Fri | Daily + SMC (+ Weekly on Monday) |
| 08:30 | validation |
Mon-Fri | Cross-reference validation (after morning push) |
| 09:00 | scorecard |
Tue-Sat | Module 09 scorecard (previous afternoon SMC) |
| 13:30 | monitor |
Mon-Fri | Module 09 live check (morning SMC) |
| 15:30 | afternoon |
Mon-Fri | SMC afternoon refresh |
| 15:33 | validation |
Mon-Fri | Cross-reference validation (3 min after refresh) |
| 19:30 | scorecard |
Mon-Fri | Module 09 scorecard (morning SMC) |
| 21:30 | monitor |
Mon-Fri | Module 09 live check (afternoon SMC) |
| Always | alerts |
24/7 | Real-time Telegram alerts (zone approach, scenarios, intervention) |
| Always | telegram-bot |
24/7 | Subscriber onboarding bot |
Scheduling is managed by 11 launchd agents in ~/Library/LaunchAgents/com.usdjpy.*.plist. The alert and telegram-bot daemons run continuously with KeepAlive and auto-restart on login.
run_local.sh waits for the Supabase project URL to be reachable (up to 5 min) before running any job, refuses to push reports that contain only N/A values, and retries each Supabase push 3× with backoff — so jobs that fire just after a wake-from-sleep no longer poison the dashboard with empty data.
Logs: ./logs/cron_<job>_<date>.log. Manual run: ./run_local.sh <job>.
- Python 3.11+
- FRED API key (free at fred.stlouisfed.org)
git clone https://github.com/<user>/usdjpy-analyst.git
cd usdjpy-analyst
pip install -r requirements.txt
cp config.yaml.example config.yaml
# Edit config.yaml: set fred.api_key, supabase.url, and risk_management settingsAdd to ~/.zshrc:
export SUPABASE_SERVICE_ROLE_KEY="your_supabase_service_role_key"
export USDJPY_EMAIL_PASSWORD="your_smtp_password" # optional
export TELEGRAM_BOT_TOKEN="your_telegram_bot_token" # for Module 10 alertsEdit config.yaml:
risk_management:
account_size: 50000 # USD
risk_per_trade: 0.01 # 1%
max_daily_risk: 0.03 # 3%
scaling:
A: 1.0 # full risk
B: 0.5 # half risk
C: 0.25 # quarter risk
D: 0.0 # paper trade only/usdjpy-daily # generates daily report
/usdjpy-entry # generates SMC entry + playbook chartusdjpy-analyst/
├── CLAUDE.md # Claude Code instructions
├── config.yaml # API keys, thresholds, risk management
├── run_daily_analysis.py # Daily pipeline (Modules 01, 03, 05, 07)
├── run_cb_analysis.py # Module 02: Central bank policy
├── run_cot_analysis.py # Module 04: CFTC COT positioning
├── run_local.sh # Launchd job runner (morning/afternoon/monitor/scorecard/validation)
│
├── scripts/
│ ├── run_smc_analysis.py # Module 08: SMC + playbook + position sizing
│ ├── smc_engine.py # SMC core: swings, OBs, FVGs, BOS/ChoCH, ATR stops
│ ├── run_scenario_monitor.py # Module 09: live check + scorecard
│ ├── run_alerts.py # Module 10: real-time Telegram/email alerts
│ ├── run_validation.py # Cross-reference validation orchestrator
│ ├── validation_sources.py # External fetchers (Yahoo, Investing.com, TradingView)
│ ├── push_to_supabase.py # Supabase data push
│ └── journal.py # Trade journal
│
├── skills/usdjpy/
│ ├── SKILL.md # Execution flow and caching rules
│ ├── modules/01-10_*.md # Module specifications
│ └── templates/ # Report templates
│
├── MQL5/Experts/SMC_USDJPY/ # MetaTrader 5 Expert Advisor
│ ├── SMC_USDJPY.mq5 # Main EA: types, inputs, orchestration
│ ├── SMC_Structure.mqh # Swing points, BOS/ChoCH, bias
│ ├── SMC_Zones.mqh # OBs, FVGs, breakers, inversions
│ ├── SMC_Liquidity.mqh # Liquidity map, sweeps, sessions
│ ├── SMC_Scoring.mqh # Confluence, killzones, displacement
│ ├── SMC_Trade.mqh # Orders, sizing, SL management
│ └── SMC_Visual.mqh # Chart drawing, dashboard
│
├── MQL5/Scripts/SMC_USDJPY/ # EA test harness
│ └── SMC_Test.mq5 # Module test suite (run in MT5)
│
├── dashboard/ # Next.js web dashboard (smcpulse.com)
│ ├── src/app/ # Pages: /, /daily, /smc, /scorecard, /validation, /journal
│ ├── src/components/ # Shared UI components
│ └── src/lib/ # Supabase client, i18n, providers
│
├── data/raw/ # Cached API responses (gitignored)
└── output/ # Reports, charts, scorecards, journal
| Decision | Rationale |
|---|---|
| Intraday-weighted scoring (tech ±3, context ±1) | Technicals drive intraday entries; macro is background context |
| No conviction cap from missing weekly modules | Daily signals stand on their own; weekly adds nuance, not gatekeeping |
| ATR-based stops, not fixed pips | Adapts to session volatility (Asian ~12p, London ~25p) |
| Distance-based intervention probability | Replaces flat 29% with realistic 1-25% based on price level |
| No-trade filter at 30 pips | Prevents forced entries when no actionable setup exists |
| Grade D = 0 lots (paper trade) | Weak setups generate analysis but not risk |
| VIX in macro module | Risk sentiment overlay — stress/elevated/normal/complacent regime + spike detection |
| Alert bias filtering | Zone alerts only fire when aligned with daily Module 07 bias — reduces noise |
| Validation 3 min after report | Cross-checks data accuracy before it goes stale |
| Wilder's EMA for validation RSI | Matches industry-standard RSI calculation |
| 12h playbook, not 24h | Projects the next 2 trading sessions — the actionable window |
This is a personal learning project. It generates analysis for educational purposes only. Nothing in this system constitutes financial advice. Trading foreign exchange carries significant risk of loss.