Alpha Research5

Component Architecture

Branch: rikk_mtf_backtest001 | Companion to: Alpha_Research

This page documents the component-level implementation of the MTFDR (Multi-Timeframe Dynamic Regime) backtesting system. Each component translates trading philosophy into executable Python.

For Developers: Each section maps trading concepts to code. You don't need trading experience—the "why" is explained before the "how".

---

PART 1: Pipeline Architecture

The system processes each 5-minute bar through 5 sequential layers:

     Raw OHLC Data (1D, 4H, 1H, 15M, 5M)
              │
              ▼
┌─────────────────────────────────────────────────────────────────────┐
│  LAYER 1: multi_timeframe_analysis.py                               │
│  "Are all timeframes pointing the same direction?"                  │
│                                                                     │
│  • Analyzes [1D, 4H, 1H, 15M, 5M] via TrendAnalysisCore            │
│  • Calculates weighted alignment score (hierarchical weights)       │
│  • 15-minute confirmation requirement                               │
│  • Output: TimeframeAnalysisResult (aligned/not, direction, score)  │
└─────────────────────────────────────────────────────────────────────┘
              │
              ▼
┌─────────────────────────────────────────────────────────────────────┐
│  LAYER 2: market_state_analysis.py                                  │
│  "What kind of market are we in?"                                   │
│                                                                     │
│  7 Detection Algorithms:                                            │
│  1. Railroad Trend - Strong one-sided move (consistency > 80%)      │
│  2. Creeper Move - Slow grinding price action (avg range < 0.5%)    │
│  3. Volatility - High (>1%), Normal, Low (<0.3%)                    │
│  4. Market State Classification - Trending/Range/Creeper/Momentum   │
│  5. Two-Day Trend - Both days close same direction                  │
│  6. Trend Phase - EARLY/MIDDLE/LATE (based on MA crossover)         │
│  7. Institutional Behavior - Fight, Accumulation, BOS               │
│  Output: MarketStateResult                                          │
└─────────────────────────────────────────────────────────────────────┘
              │
              ▼
┌─────────────────────────────────────────────────────────────────────┐
│  LAYER 3: setup_quality_detection.py                                │
│  "How good is this setup? Grade it A+ to F."                        │
│                                                                     │
│  5-Factor Weighted Scoring:                                         │
│  • Timeframe Alignment (30%) - From Layer 1                         │
│  • Trend Strength (20%) - Penalties: creeper, MA struggle           │
│  • Entry Quality (15%) - Near MA bonus, clean entry bonus           │
│  • Key Level Proximity (20%) - Near support/resistance              │
│  • Risk/Reward (15%) - Lookup table: <1=0, 1-1.5=40, 1.5-2=70...    │
│                                                                     │
│  Penalties Applied: creeper(-50), ma_struggle(-30), no_2day(-30)    │
│  A+ Enforcement: Must pass all 3 criteria or capped at 79           │
│  Output: SetupQualityResult (grade, score, position_size)           │
└─────────────────────────────────────────────────────────────────────┘
              │
              ▼
┌─────────────────────────────────────────────────────────────────────┐
│  LAYER 4: signal_generation_trade_management.py                     │
│  "Should we trade? What direction? Where's entry/stop/target?"      │
│                                                                     │
│  Direction Logic:                                                   │
│  • MA21 rising → LONG direction                                     │
│  • MA21 declining → SHORT direction                                 │
│  • Flat → MTF fallback                                              │
│                                                                     │
│  Gate Filters (JIRA tickets):                                       │
│  • TICKET-12: Hour filter (block 9AM, 10PM, 11PM)                   │
│  • TICKET-19: MA direction alignment                                │
│  • TICKET-20: Probability zone filter (uses probability_zone_analysis)│
│  • TICKET-21: Color change confirmation (currently DISABLED)        │
│                                                                     │
│  Output: Signal (direction, entry, stop, target, grade)             │
└─────────────────────────────────────────────────────────────────────┘
              │
              ▼
┌─────────────────────────────────────────────────────────────────────┐
│  LAYER 5: trade_execution_engine.py                                 │
│  "Execute the trade, manage the position, calculate P&L"            │
│                                                                     │
│  Execution:                                                         │
│  • Signal-to-trade conversion with slippage modeling                │
│  • Commission: Rs 20/lot (Dhan)                                     │
│  • CTT: 0.01% on sell side                                          │
│                                                                     │
│  Position Management:                                               │
│  • TICKET-15/17/20: Trailing stop (ATR-based, 2x ATR)               │
│  • TICKET-13/17: Breakeven stop (activate at 25pt profit)           │
│  • Time-based exit (8 hour max holding)                             │
│                                                                     │
│  Output: Trade objects with complete P&L                            │
└─────────────────────────────────────────────────────────────────────┘

---

PART 2: Component Index

Layer	Component	What It Does	Code Link
1	multi_timeframe_analysis.py	MTF alignment across 5 timeframes	View Code
(1a)	trend_analysis_core.py	Per-timeframe trend direction, MA slopes	View Code
2	market_state_analysis.py	Railroad, creeper, volatility, trend phase	View Code
3	setup_quality_detection.py	5-factor scoring, A+ to F grades	View Code
4	signal_generation_trade_management.py	Direction, filters, entry/stop/target	View Code
5	trade_execution_engine.py	Execution, position mgmt, P&L	View Code

Support Components:

Component	What It Does	Used By	Code Link
probability_zone_analysis.py	Probability zones (80%/65%/35%/15%), crash bar, color change	signal_generation (TICKET-20)	View Code
data_manager.py	Data loading, timeframe aggregation	main.py	View Code
backtesting_analytics.py	Performance metrics, reporting	main.py	View Code

---

PART 3: Trading Concept → Code Mapping

PDF Concept	Implementation Status	Code Location	Notes
Halves/Thirds (80%/65%/35%/15%)	✅ Implemented	probability_zone_analysis.py	Used via TICKET-20 filter in signal_generation
Crash Bar	✅ Implemented	probability_zone_analysis.py	Bar 2x average = structural break
Color Change	⚠️ Disabled	probability_zone_analysis.py	TICKET-21: Too restrictive, set enable_color_change_filter=False
Three-Finger Spread	✅ Implemented	probability_zone_analysis.py	Price/21MA/200MA separation detection (PDF calls it large MA spread)
45-degree angle	✅ Implemented	trend_analysis_core.py	MA slope analysis, "flattish" detection
Railroad Trend	✅ Implemented	market_state_analysis.py	Consistency > 80%, 3+ strong bars
Creeper Move	✅ Implemented	market_state_analysis.py	Avg range < 0.5% over 7 bars
Two-Day Trend	✅ Implemented	market_state_analysis.py	Both daily bars close same direction
Fab Four (21MA + 200MA zone)	❌ Not Implemented	-	PDF concept not directly coded; closest is Three-Finger Spread
Traffic Jam	❌ Not Implemented	-	Prior day congestion as S/R not coded
Big Bar by Big Bar	⚠️ Partial	trend_analysis_core.py	Slope calculation uses significant moves; no explicit "noise filter"

---

PART 4: Key Configuration Parameters

Setup Quality Weights (setup_quality_detection.py)

timeframe_alignment_weight = 0.30   # 30%
trend_strength_weight      = 0.20   # 20%
entry_technique_weight     = 0.15   # 15%
key_level_proximity_weight = 0.20   # 20%
risk_reward_weight         = 0.15   # 15%
                           -------
                           = 1.00   # Must sum to 1.0

Penalty Constants

creeper_move_penalty       = -50    # Major penalty for slow grinding
ma_struggle_penalty        = -30    # Price repeatedly testing MA
two_day_trend_penalty      = -30    # Missing 2-day directional trend
phase_mismatch_penalty     = -25    # Not in MIDDLE phase
railroad_trend_bonus       = +15    # Strong one-sided move

Grade Thresholds

A+ = score >= 90    →  2 lots
A  = score >= 80    →  1 lot
B  = score >= 70    →  1 lot
C  = score >= 60    →  1 lot
D  = score >= 50    →  1 lot
F  = score < 50     →  1 lot (signal may be rejected by filters)

---

PART 5: Component Details

Multi-Timeframe Analysis

See: Component:Multi_Timeframe_Analysis

Trend Analysis Core

See: Component:Trend_Analysis_Core

Market State Analysis

See: Component:Market_State_Analysis

Setup Quality Detection

See: Component:Setup_Quality_Detection

Signal Generation

See: Component:Signal_Generation

Trade Execution

See: Component:Trade_Execution

---

PART 6: JIRA Ticket Reference

Ticket	Component	What It Fixed
TICKET-5	setup_quality	Realistic position sizing (was 100-200 lots, now 1-2 lots)
TICKET-8	signal_generation	CTT rate correction (was 0.05%, now 0.01%)
TICKET-11	signal_generation	BOS stop distance (extended lookback from 5 to 10 bars)
TICKET-12	signal_generation	Hour filter (block 9AM, 10PM, 11PM)
TICKET-13	trade_execution	Breakeven stop (25pt activation)
TICKET-15/17/20	trade_execution	Trailing stop (ATR-based, 2x ATR)
TICKET-19	signal_generation	MA direction alignment filter
TICKET-20	signal_generation	Probability zone filter (integrates probability_zone_analysis)
TICKET-21	signal_generation	Color change filter (DISABLED - too restrictive)
TICKET-25	market_state	Fixed creeper detection to use 1H data, not 5M

---