Configuration¶

This guide covers all configuration options for the RQF-ML system.

Pine Script Settings¶

Core Settings¶

Parameter	Default	Range	Description
`pivotLen`	5	2-20	Pivot detection sensitivity for H/L labels
`structureLen`	20	5-50	Structure detection for BoS/MSS
`lookback`	250	50-500	Bars to track for drawings

Timeframe Guidance

Scalping (1-5m): pivotLen=4, structureLen=15
Intraday (15m-1H): pivotLen=5, structureLen=20
Swing (4H-D): pivotLen=7, structureLen=30

Equal Highs/Lows¶

Parameter	Default	Description
`showEQHL`	true	Show EQH/EQL lines
`eqThresholdATR`	0.3	ATR multiplier for "equal" definition
`eqLookbackBars`	50	Bars to search for equal levels

RQF Sweep System¶

Parameter	Default	Description
`showSweep`	true	Show S1, S2... labels
`showTrigger`	true	Show TH/TL trigger lines
`showFailedMss`	true	Show f1, f2... failed MSS labels
`showEntrySignal`	true	Show entry signals
`sweepTimeout`	100	Max bars to wait for sweep/MSS

Fibonacci Settings¶

Parameter	Default	Description
`showFib`	true	Show Fib levels
`showDiscountBox`	true	Show zone boxes
`fibExtendBars`	50	Bars to extend Fib

Confluence (OB/FVG)¶

Parameter	Default	Description
`showOB`	true	Show Order Blocks
`showBreakers`	true	Convert broken OBs to Breakers
`showFVG`	true	Show Fair Value Gaps
`fvgFilterZone`	true	Only FVGs in discount/premium
`fvgMaxCount`	3	Maximum FVGs displayed
`obExtendBars`	30	Bars to extend OB box
`obMaxSearch`	50	Bars to search for OB
`fvgMinSize`	0.5	Minimum FVG size (ATR mult)

Python Configuration¶

Edge Config (`config/edge_config.yaml`)¶

# Confluence scoring weights
edge_stack:
  quality_weight: 0.4    # Pattern quality (OB, FVG presence)
  time_weight: 0.2       # Session-based edge
  sweep_weight: 0.2      # Sweep quality (wick ratio)
  nested_weight: 0.1     # Nested OB bonus
  htf_weight: 0.1        # HTF alignment bonus

# Backtest settings
backtest:
  initial_capital: 10000
  commission_per_trade: 0.50
  slippage_ticks: 1
  position_size: 1

# ML model settings
model:
  type: lightgbm
  params:
    n_estimators: 200
    max_depth: 6
    learning_rate: 0.05
    min_child_samples: 20
    subsample: 0.8
    colsample_bytree: 0.8

# Validation gates
gates:
  min_sharpe_oos: 0.8
  max_drawdown: 0.08
  min_win_rate: 0.50
  min_profit_factor: 1.3
  walk_forward_pct: 0.60

Risk Configuration¶

Create config/risk_limits.yaml:

# Position limits
max_position_size: 2          # contracts/lots
max_concurrent_positions: 3
max_daily_trades: 10

# Loss limits
max_daily_loss_pct: 3.0       # % of capital
max_total_drawdown_pct: 6.0   # % of capital
consecutive_loss_limit: 5

# Per-trade limits
max_risk_per_trade_pct: 1.0   # % of capital
min_rr_ratio: 1.5             # minimum reward:risk

# Kill switch triggers
kill_switch:
  enabled: true
  daily_loss_trigger: 3.0     # % loss triggers kill switch
  spread_anomaly_mult: 5.0    # spread > 5x normal
  connection_timeout: 60      # seconds

Environment Variables¶

Create .env from .env.example:

# API Keys
TRADINGVIEW_WEBHOOK_SECRET=your_secret_here
MT5_LOGIN=your_login
MT5_PASSWORD=your_password
MT5_SERVER=your_broker_server

# Notification channels
DISCORD_WEBHOOK_URL=https://discord.com/api/webhooks/...
TELEGRAM_BOT_TOKEN=your_bot_token
TELEGRAM_CHAT_ID=your_chat_id

# Data sources
FIRSTRATE_API_KEY=your_api_key

# Environment
ENVIRONMENT=development  # development, paper, live
LOG_LEVEL=INFO

Security

Never commit .env to version control. It's in .gitignore by default.

Instrument-Specific Settings¶

Futures - Equity Index¶

Instrument	Tick Size	SL Buffer (ticks)	Session
MNQ, NQ	0.25	5	NY AM
MES, ES	0.25	5	NY AM
MYM, YM	1.0	5	NY AM
M2K, RTY	0.10	5	NY AM

Futures - Commodities¶

Instrument	Tick Size	SL Buffer (ticks)	Session
GC, MGC	0.10	10	London/NY
SI	0.005	10	London/NY
CL, MCL	0.01	10	NY

Crypto¶

Instrument	Tick Size	SL Buffer (ticks)	Session
BTCUSD	0.01	50	24/7
ETHUSD	0.01	50	24/7
SOLUSD	0.01	30	24/7

Forex¶

Instrument	Tick Size	SL Buffer (pips)	Session
EURUSD	0.00001	10	London/NY
GBPUSD	0.00001	12	London/NY
USDJPY	0.001	10	Tokyo/London

Session Configuration¶

Kill Zones (NY Time)¶

sessions:
  asia:
    start: "20:00"
    end: "00:00"
    quality: low
    color: blue

  london:
    start: "02:00"
    end: "05:00"
    quality: medium
    color: red

  ny_am:
    start: "08:30"
    end: "11:00"
    quality: high
    color: teal

  ny_lunch:
    start: "12:00"
    end: "13:00"
    quality: avoid
    color: yellow

  ny_pm:
    start: "13:30"
    end: "16:00"
    quality: medium
    color: purple

Best Sessions

NY AM (08:30-11:00 ET) has the highest edge based on backtesting.

Claude Code Skills Configuration¶

Skills are configured in .claude/skills/[skill-name]/SKILL.md.

Frontmatter Options¶

---
name: skill-name              # Required
description: "When to use"    # For auto-invocation
disable-model-invocation: true  # Manual only
user-invocable: false         # Hide from menu
argument-hint: [args]         # Autocomplete hint
allowed-tools: Read, Grep     # Tool restrictions
context: fork                 # Isolated context
agent: Explore                # Subagent type
model: sonnet                 # Force model
---

See Creating Skills for full documentation.