Latest revision as of 19:19, 11 January 2026

Multi-Timeframe Trading Strategy System

Version: 2.0 | Last Updated: 2026-01-12 | Status: Production

Executive Summary

This document describes a systematic, rule-based crude oil futures trading strategy implemented as a multi-component backtesting system. The strategy uses multi-timeframe alignment (5M to Daily), market state analysis, setup quality grading, and institutional-grade signal generation to identify high-probability trade setups.

**System Statistics (Latest Backtest)**
Metric	Value	Notes
Total Trades	1,829	7-month period (Jun 2025 - Dec 2025)
Win Rate	55.1%	Above random chance
Avg Holding Period	2.8 hours	Intraday focus
Primary Direction	85% Short	Trend-following bias
Exit Types	Stop: 80%, Target: 17%, Timeout: 4%	Stop-based risk management

System Architecture Overview

                          +------------------------------------------+
                          |          DATA LAYER                      |
                          |  +---------+ +---------+ +---------+     |
                          |  | 1-Min   | | 5-Min   | | 15-Min  |     |
                          |  | OHLCV   | | OHLCV   | | OHLCV   |     |
                          |  +---------+ +---------+ +---------+     |
                          |  +---------+ +---------+ +---------+     |
                          |  | 1-Hour  | | 4-Hour  | | Daily   |     |
                          |  | OHLCV   | | OHLCV   | | OHLCV   |     |
                          |  +---------+ +---------+ +---------+     |
                          +---------------------|--------------------+
                                                |
                                                v
+-------------------------------------------------------------------------------------------+
|                              PIPELINE ENGINE (Orchestrator)                                |
|                                                                                           |
|   +---------------+    +---------------+    +---------------+    +---------------+        |
|   | LAYER 1       |    | LAYER 2       |    | LAYER 3       |    | LAYER 4       |        |
|   | Trend         |--->| Market State  |--->| Setup Quality |--->| Signal        |        |
|   | Analysis      |    | Analysis      |    | Detection     |    | Generation    |        |
|   +---------------+    +---------------+    +---------------+    +---------------+        |
|          |                    |                    |                    |                 |
|          v                    v                    v                    v                 |
|   [MA21, MA200]        [range_bound]         [A+/A/B/C/D/F]      [Entry/Stop/TP]         |
|   [Trend Direction]    [creeper_move]        [Score 0-100]       [Direction]             |
|   [Trend Phase]        [narrow_volume]       [Risk %]            [Position Size]         |
|                                                                                           |
|                                          |                                                |
|                                          v                                                |
|                               +-------------------+                                       |
|                               | LAYER 5           |                                       |
|                               | Trade Execution   |                                       |
|                               | Engine            |                                       |
|                               +-------------------+                                       |
|                                          |                                                |
|                                          v                                                |
|                               [Open/Manage/Close Trades]                                  |
|                               [P&L Calculation]                                           |
|                               [Cost Accounting]                                           |
+-------------------------------------------------------------------------------------------+
                                          |
                                          v
                          +------------------------------------------+
                          |          OUTPUT LAYER                    |
                          |  +------------------+  +---------------+ |
                          |  | Trade-Level CSV  |  | Summary JSON  | |
                          |  | (All trades)     |  | (Metrics)     | |
                          |  +------------------+  +---------------+ |
                          +------------------------------------------+

Component Details

Layer 1: Multi-Timeframe Trend Analysis

Purpose: Analyze trend direction and strength across 5 timeframes

File: src/components/multi_timeframe_analysis.py, src/components/trend_analysis_core.py

Timeframes Analyzed:

Timeframe	Weight	Purpose
Daily (1D)	30%	Context (overall trend)
4-Hour (4H)	25%	Primary trend
1-Hour (1H)	20%	Confirmation
15-Min (15M)	15%	Fine-tuning
5-Min (5M)	10%	Entry timing

Key Calculations:

Alignment Score = SUM(Weight[tf] * Direction_Match[tf]) for all tf

Where:
- Direction_Match = 1 if tf direction matches dominant direction, else 0
- Dominant direction = direction with highest weighted votes

MA Analysis per Timeframe:
- MA21 (fast) - Short-term trend
- MA200 (slow) - Long-term trend
- Price position relative to MAs
- MA slope (trending vs flat)

Output: TimeframeAnalysisResult

final_direction: LONG / SHORT / NEUTRAL
alignment_score: 0.0 - 1.0
timeframe_results: Per-timeframe analysis dictionary

Layer 2: Market State Analysis

Purpose: Classify current market conditions into actionable states

File: src/components/market_state_analysis.py

Market States:

+-------------------+--------------------------------+------------------+
| Market State      | Characteristics                | Trading Action   |
+-------------------+--------------------------------+------------------+
| range_bound       | Price oscillating, no clear    | Trade reversals  |
|                   | trend, MA21 relatively flat    | at range edges   |
+-------------------+--------------------------------+------------------+
| creeper_move      | Slow, grinding trend with      | Trade WITH trend |
|                   | small candles, steady MA21     | on pullbacks     |
+-------------------+--------------------------------+------------------+
| narrow_low_volume | Tight range, reduced volume,   | CAUTION - await  |
|                   | consolidation pattern          | breakout         |
+-------------------+--------------------------------+------------------+
| trending          | Strong directional move,       | Trade WITH trend |
|                   | clear MA separation            | aggressively     |
+-------------------+--------------------------------+------------------+

Detection Algorithm:

1. Calculate recent volatility (ATR or range)
2. Calculate MA21 slope
3. Calculate volume relative to average
4. Apply classification rules:

IF slope < threshold AND range tight THEN narrow_low_volume
ELIF slope small AND trend present THEN creeper_move
ELIF volatility high AND no trend THEN range_bound
ELIF strong directional move THEN trending

Output: MarketStateResult

market_state: Enum value
volatility_metrics: ATR, range data
trend_phase: early/middle/late/exhaustion

Layer 3: Setup Quality Detection

Purpose: Score and grade trade setups using 5-factor weighted system

File: src/components/setup_quality_detection.py

5-Factor Scoring System:

+-------------------------+--------+------------------------------------------+
| Factor                  | Weight | Description                              |
+-------------------------+--------+------------------------------------------+
| Timeframe Alignment     | 30%    | How many timeframes agree on direction   |
| Trend Strength          | 20%    | MA slope, price momentum                 |
| Entry Quality           | 15%    | Clean candle, near MA, not extended      |
| Key Level Proximity     | 20%    | Near support/resistance levels           |
| Risk/Reward Ratio       | 15%    | Potential reward vs risk                 |
+-------------------------+--------+------------------------------------------+
                           = 100%

Grade Classification:

Score >= 90  -->  A+ Grade  (Best setups, full position)
Score >= 80  -->  A Grade   (Excellent, 80% position)
Score >= 70  -->  B Grade   (Good, 60% position)
Score >= 60  -->  C Grade   (Average, 40% position)
Score >= 50  -->  D Grade   (Below average, 20% position)
Score <  50  -->  F Grade   (Poor, NO TRADE)

Penalty System:

Penalties Applied:
- Creeper Move detected: -50 points
- MA Struggle (price fighting MA): -30 points
- Missing Two-Day Trend: -30 points
- Phase Mismatch: -25 points
- Far from Key Level: -50 points
- Far from MA: -40 points

Bonuses Applied:
+ Railroad Trend (strong alignment): +15 points
+ At Key Level: +10 points
+ Clean Entry Candle: +10 points

Institutional Fight Detection:

IF 4H direction != Daily direction THEN
    Apply 0.7x multiplier to final score
    (Institutions may be fighting retail)

Output: SetupQualityResult

grade: A+ / A / B / C / D / F
score: 0 - 100
factor_scores: Individual factor breakdown
position_size: Recommended lots
risk_percent: Account risk percentage
can_auto_trade: Boolean

Layer 4: Signal Generation & Trade Management

Purpose: Generate precise entry/exit signals with risk parameters

File: src/components/signal_generation_trade_management.py

Signal Generation Flow:

                    +------------------+
                    | Setup Quality    |
                    | Grade >= B       |
                    +--------+---------+
                             |
                             v
                    +------------------+
                    | Direction Check  |
                    | MTF Aligned?     |
                    +--------+---------+
                             |
              +--------------+--------------+
              |                             |
              v                             v
     +--------+--------+           +--------+--------+
     | FILTERS         |           | GENERATE        |
     | Hour filter     |           | Entry price     |
     | Prob zone       |           | Stop loss       |
     | MA direction    |           | Take profit     |
     +--------+--------+           | Position size   |
              |                    +-----------------+
              v
     +------------------+
     | BLOCKED or       |
     | PASSED           |
     +------------------+

Entry Techniques (17 Types):

1. near_ma          - Entry near 21 MA
2. ma_bounce        - Bounce off 21 MA
3. break_of_structure (BOS) - Breakout entry
4. pullback_entry   - Trend pullback
5. key_level_entry  - At support/resistance
... (12 more specialized techniques)

Stop Loss Calculation:

LONG Stop = Entry Price - MAX(recent_low - buffer, min_stop_distance)
SHORT Stop = Entry Price + MAX(recent_high + buffer, min_stop_distance)

Where:
- min_stop_distance = 40 points (prevents too-tight stops)
- buffer = 5 points for BOS entries
- recent lookback = 5-10 bars depending on technique

Take Profit Methods:

1. Fixed R:R    - TP = Entry + (Stop_Distance * Risk_Reward_Ratio)
2. Fixed Points - TP = Entry +/- default_profit_points
3. ATR Multiple - TP = Entry +/- (ATR * atr_multiple)

Active Filters:

Filter	Purpose	Impact
Hour Filter	Block toxic hours (9, 22, 23)	Saves Rs 357k+
MA Direction	Trade only with MA21 slope	Improves trend alignment
Probability Zone	Block counter-trend in extremes	Reduces false signals

Output: Signal

signal_id: Unique identifier
direction: LONG / SHORT
entry_price: Precise entry level
stop_loss_price: Risk cutoff
take_profit_price: Target level
position_size: Lots to trade
setup_quality: Grade
estimated_costs: Brokerage + taxes

Layer 5: Trade Execution Engine

Purpose: Execute signals, manage open positions, calculate P&L

File: src/components/trade_execution_engine.py

Trade Lifecycle:

SIGNAL RECEIVED
      |
      v
+-----+-----+
| Validate  |
| Signal    |
+-----+-----+
      |
      v
+-----+-----+      +-----------+
| Check     |----->| REJECT    |
| Filters   |  NO  | Signal    |
+-----+-----+      +-----------+
      | YES
      v
+-----+-----+
| OPEN      |
| Position  |
+-----+-----+
      |
      v
+-----+-----+
| Monitor   |<----+
| Each Bar  |     |
+-----+-----+     |
      |           |
      v           |
+-----+-----+     |
| Check     |     |
| Exit      |     |
| Conditions|-----+
+-----+-----+  NO HIT
      | HIT
      v
+-----+-----+
| CLOSE     |
| Trade     |
+-----+-----+
      |
      v
+-----+-----+
| Calculate |
| P&L       |
+-----+-----+

Exit Conditions:

1. STOP HIT    - Price touches stop_loss_price
2. TARGET HIT  - Price touches take_profit_price
3. TIMEOUT     - Max holding period reached
4. EOD         - End of day force close

P&L Calculation:

For LONG:
  Gross P&L = (Exit_Price - Entry_Price) * Position_Size * Lot_Multiplier

For SHORT:
  Gross P&L = (Entry_Price - Exit_Price) * Position_Size * Lot_Multiplier

Net P&L = Gross P&L - Commission - Taxes

Where:
  Commission = Position_Size * Rs 20 per leg * 2 (entry + exit)
  Taxes (CTT) = Contract_Value * 0.01% (sell side)
  Lot_Multiplier = 100 (MCX Crude Oil)

Data Flow Diagram

                          RAW 1-MIN DATA
                               |
                               v
                    +--------------------+
                    | Timeframe          |
                    | Conversion         |
                    | (1m->5m,15m,1h,4h,1d)
                    +--------------------+
                               |
                               v
+------------------+   +------------------+   +------------------+
| 5-Min Candles    |   | 15-Min Candles   |   | Higher TFs       |
| with MA21, MA200 |   | with MAs         |   | (1H, 4H, Daily)  |
+--------+---------+   +--------+---------+   +--------+---------+
         |                      |                      |
         +----------------------+----------------------+
                               |
                               v
                    +--------------------+
                    | Bar-by-Bar         |
                    | Processing Loop    |
                    +--------------------+
                               |
         +---------------------+---------------------+
         |                     |                     |
         v                     v                     v
+--------+--------+  +--------+--------+  +--------+--------+
| Trend Analysis  |  | MTF Analysis    |  | Market State    |
| (5M focus)      |  | (All TFs)       |  | Analysis        |
+-----------------+  +-----------------+  +-----------------+
         |                     |                     |
         +---------------------+---------------------+
                               |
                               v
                    +--------------------+
                    | Setup Quality      |
                    | Detection          |
                    | (5-Factor Score)   |
                    +--------------------+
                               |
                               v
                    +--------------------+
                    | Signal Generation  |
                    | (Entry/Stop/TP)    |
                    +--------------------+
                               |
                               v
                    +--------------------+
                    | Trade Execution    |
                    | Engine             |
                    +--------------------+
                               |
                               v
                    +--------------------+
                    | Trade Record       |
                    | (CSV + JSON)       |
                    +--------------------+

Key Decision Points

When to Generate Signal

+------------------------------------------------------------------+
|                    SIGNAL GENERATION DECISION TREE                |
+------------------------------------------------------------------+

START
  |
  v
[MTF Alignment Score >= 0.6?] --NO--> NO SIGNAL
  |
  YES
  v
[Setup Quality Grade >= B?] --NO--> NO SIGNAL
  |
  YES
  v
[Current Hour in Blocked Hours?] --YES--> BLOCKED (NO SIGNAL)
  |
  NO
  v
[MA Direction Matches Trade Direction?] --NO--> BLOCKED
  |
  YES
  v
[In Probability Zone Against Trade?] --YES--> BLOCKED
  |
  NO
  v
[Position Already Open?] --YES--> NO NEW SIGNAL
  |
  NO
  v
+-------------------+
| GENERATE SIGNAL!  |
+-------------------+

Trade Direction Logic

LONG CONDITIONS (all must be true):
  - Daily trend = UP or NEUTRAL
  - 4H trend = UP
  - 1H trend = UP
  - MA21 slope > 0 (rising)
  - Price above MA21 or bouncing from it
  - Not in upper probability zone (overbought)

SHORT CONDITIONS (all must be true):
  - Daily trend = DOWN or NEUTRAL
  - 4H trend = DOWN
  - 1H trend = DOWN
  - MA21 slope < 0 (falling)
  - Price below MA21 or rejected from it
  - Not in lower probability zone (oversold)

Configuration Reference

Strategy Configuration

StrategyConfig:
  primary_timeframe: 5M
  ma21_period: 21
  ma200_period: 200
  min_bars_for_analysis: 21
  max_position_size: 10 lots
  max_concurrent_trades: 3
  default_risk_percent: 1.0%
  base_slippage_points: 2.0

Signal Generator Configuration

SignalGeneratorConfig:
  commission_per_lot: Rs 20.00
  transaction_tax_rate: 0.005% per leg
  lot_size_multiplier: 100 (MCX)
  min_stop_distance: 40 points
  default_risk_reward: 2.0

  # Filters
  enable_hour_filter: true
  blocked_hours: [9, 22, 23]
  enable_ma_direction_filter: true
  enable_probability_zone_filter: true

Performance Analysis Framework (ISOLATION)

Edge Discovery Process

STEP 1: BASELINE
  - Run full backtest
  - Calculate: Trades, Win Rate, P&L, Expectancy

STEP 2: SLICE BY DIMENSIONS
  Dimension 1: Market State (range_bound, creeper_move, etc.)
  Dimension 2: Setup Grade (A+, A, B, C, D)
  Dimension 3: Hour of Day (9-23)
  Dimension 4: Month
  Dimension 5: Direction (Long/Short)

STEP 3: IDENTIFY EDGES
  Per slice calculate:
  - Trade count
  - Win rate
  - Net P&L
  - Expectancy = (WR% * Avg Win) + ((1-WR%) * Avg Loss)
  - Profit Factor = Gross Profit / Gross Loss

STEP 4: GATE 1 CRITERIA
  - Win Rate > 30%
  - Expectancy > 0
  - Sample Size >= 30 trades

STEP 5: FILTER/FOCUS
  - Remove toxic slices
  - Trade only edge slices

Latest Analysis Results

EDGE Combinations (Profitable):

Combination	Trades	Win%	Net P&L	Expectancy
range_bound + H11	100	82.0%	Rs +358,640	Rs +3,586/trade
creeper_move + H19	42	78.6%	Rs +123,870	Rs +2,949/trade
creeper_move + H21	36	77.8%	Rs +105,293	Rs +2,925/trade
creeper_move + H17	30	86.7%	Rs +37,426	Rs +1,248/trade

TOXIC Combinations (Loss-Making):

Combination	Trades	Win%	Net P&L	Expectancy
narrow_low_volume + H14	38	7.9%	Rs -157,167	Rs -4,136/trade
creeper_move + H9	33	21.2%	Rs -107,488	Rs -3,257/trade
range_bound + H9	110	45.5%	Rs -109,227	Rs -993/trade

File Structure

services/backtest/
|
+-- src/
|   +-- core/
|   |   +-- pipeline_engine.py      # Main orchestrator
|   |   +-- data_structures.py      # Enums, dataclasses
|   |   +-- base_component.py       # Component base classes
|   |   +-- timeframe_conversion.py # 1min -> all TFs
|   |
|   +-- components/
|   |   +-- trend_analysis_core.py           # Layer 1
|   |   +-- multi_timeframe_analysis.py      # Layer 1a
|   |   +-- market_state_analysis.py         # Layer 2
|   |   +-- setup_quality_detection.py       # Layer 3
|   |   +-- signal_generation_trade_management.py  # Layer 4
|   |   +-- trade_execution_engine.py        # Layer 5
|   |   +-- probability_zone_analysis.py     # Support component
|   |
|   +-- main.py                     # Entry point
|
+-- data/
|   +-- input/
|   |   +-- Crude/
|   |       +-- 1min/crude_1m.csv   # Raw data
|   |       +-- 5min/crude_5m.csv   # Converted
|   |       +-- ... (other TFs)
|   |
|   +-- output/
|       +-- trade_level_data_for_frontend.csv
|       +-- trade_level_data_for_frontend_summary.json
|
+-- docs/
    +-- STRATEGY_SYSTEM_ARCHITECTURE.mediawiki  # This file

Glossary

Term	Definition
MTF	Multi-Timeframe - analyzing multiple chart timeframes simultaneously
MA21	21-period Moving Average - primary trend indicator
MA200	200-period Moving Average - long-term trend context
Setup Grade	A+ to F quality score for trade setups
Expectancy	Average profit/loss per trade (Rs/trade)
Profit Factor	Gross Profits / Gross Losses ratio
BOS	Break of Structure - price breaking previous swing high/low
CTT	Commodities Transaction Tax (India)
MCX	Multi Commodity Exchange (India)
R:R	Risk to Reward Ratio
Lot Size	100 barrels for MCX Crude Oil futures

Version History

Version	Date	Changes
2.0	2026-01-12	ISOLATION analysis integration, filter documentation
1.5	2026-01-10	Added probability zone filtering, hour filters
1.0	2025-12-01	Initial 5-layer architecture documentation