Trend Following Strategy Development Project

    ___________                   __   ___________      .__  .__                .__
    \__    ___/______  ____   ____/  |_ \_   _____/____  |  | |  |   ______  _  _|__| ____    ____
      |    |  \_  __ \_/ __ \ /    \   __\|    __)/  _ \ |  | |  |  /  _ \ \/ \/ /  |/    \  / ___\
      |    |   |  | \/\  ___/|   |  \  |  |     \(  <_> )|  |_|  |_(  <_> )     /|  |   |  \/ /_/  >
      |____|   |__|    \___  >___|  /__|  \___  / \____/ |____/____/\____/ \/\_/ |__|___|  /\___  /
                           \/     \/          \/                                         \//_____/

Trend Following Strategy Development Project

A systematic exploration of 40+ trend-following strategies, documenting the journey from -93% to +40% through iterative learning and optimization.

New to this repository? Start with docs/START_HERE.md for a guided introduction, or dive directly into DISCOVERIES_AND_LEARNINGS.md for comprehensive results from 293 validated backtests. Basket traders should make a beeline for the new “Basket Trading Playbook (Options Baskets)” section, and analytic-minded readers can review the fresh “Segmented logistic drill-down” insights generated from the Python pipeline.

💡 KEY TAKEAWAYS (293 Validated Backtests Across 21 Securities)

The Critical Discovery: Strategy Performance is Asset-Dependent

What works on SPY often fails elsewhere. After testing 14 strategies across 21 securities (11 stocks + 10 ETFs), the data reveals that asset characteristics matter more than strategy sophistication. The same strategy can be the best performer on one security and catastrophic on another.

The Winners: Universal Strategies That Actually Work

Alt10 (Profit Targets) - 76.19% success rate (16/21 profitable)
- Explicit 3N-6N-9N profit taking = predictable 3-10 week holds
- Works on BOTH stocks AND ETFs (rare!)
- Best for: Healthcare (UNH +33.1%, XLV +27.7%), broad market (SPY +20.3%, QQQ +22.8%)
Alt45 (Dual-Momentum) - 66.67% success rate (14/21 profitable)
- 2nd most consistent strategy overall
- QQQ +29.23%, CAT +28.62%, UNH +27.07%
- Best for: Growth stocks with strong momentum
Alt26 (Fractional Pyramid) - 57.14% success rate (12/21 profitable)
- Best SPY performer (+33.50% validated)
- Lowest drawdowns (MSFT 4.98%, WMT 5.96%)
- Scale-out exits (100%→75%→50%→25%)

The Sector Truth: Some Markets Are Untradeable

Healthcare: 12/13 strategies profitable (92% success) - TRADE THIS
Technology: 8/10 strategies profitable (80% success) - TRADE THIS
Utilities: 0/14 strategies profitable (0% success) - AVOID COMPLETELY
Energy: 2/10 strategies profitable (20% success) - AVOID

Full sector-by-sector play-calling lives in the Basket Trading Playbook and the detailed sector recommendation matrix.

Latest Python Analytics & Basket Guidance

Segmentation confirms the story: The analysis-outputs/statistical_outputs/segmented_models/ logistic regressions reinforce healthcare/tech tailwinds, highlight the stock vs ETF edge, and flag Energy/Utilities as statistically hostile. See the breakdown in DISCOVERIES_AND_LEARNINGS.md → Segmented logistic drill-down.
Basket-ready roster: The newest Basket Trading Playbook maps every sector sleeve to its Pine Script, backup, and hard “do not trade” calls so options baskets stay aligned with the data.

Record-Breaking Performances (All Validated)

Best ETF Result: XLV Alt46 +34.80% (sector-adaptive parameters)
Lowest Drawdown Ever: CAT Alt47 3.96% max DD (across all 293 backtests!)
Highest Win Rate: UNH Alt28 72.22% (legendary quality)
Best Profit Factor: UNH Alt28 4.809 (exceptional risk/reward)

See DISCOVERIES_AND_LEARNINGS.md → Best by asset class and adjacent sections for the full leaderboards.

Data Quality: 99.74% Validation Rate

After systematic visual verification of all 388 screenshots:

1 exit bug found (SPY Alt39 - purple arrows stopped after 2017)
387 results validated as healthy with proper exit activity
Methodology: Visual inspection of exit markers throughout 2010-2025
See data-verification-plan.md for complete validation protocol

The Options Trading Insight

Not all profitable strategies suit options trading. Strategies need regular exits (3-12 weeks) to match options expiration. Visual cue: If the equity curve is zoomed (showing only 2010-2012 instead of full timeline), the strategy holds too long for options. The go/no-go list is maintained in DISCOVERIES_AND_LEARNINGS.md → Options trader final recommendations.

What Actually Matters (Validated by 293 Backtests)

Exit optimization > Entry optimization - Where you get out matters more than where you get in
Profit targets are essential - Multi-stage profit taking (3N, 6N, 9N) outperforms trailing-only
Pyramiding is mandatory - Single-position strategies catastrophically underperform
Asset selection trumps strategy selection - Alt10 on Healthcare beats best strategy on Utilities
Simplicity beats complexity - Top performers use 2-3 enhancements, not kitchen-sink approaches

The Bottom Line for Traders

Use Alt10, Alt45, or Alt26 on Healthcare/Technology stocks. Avoid Utilities and Energy entirely. Test thoroughly on YOUR instruments before live trading - what works on SPY may fail on your ticker.

✅ Project Status: COMPLETED

293 Valid Backtests across 14 strategies and 21 securities (1 removed due to exit bug - see data-verification-plan.md)

Key Findings:

Best Overall Strategy: Alt10 (Profit Targets) - 76.19% success rate (16/21 profitable)
Best SPY Strategy: Alt26 (Fractional Pyramid) - +33.50% validated result
Best ETF Strategy: Alt46 (Sector Adaptive) - XLV +34.80% return
Lowest Drawdown Ever: Alt47 CAT - 3.96% max DD (across all 293 backtests)
Sector Insight: Healthcare 92% success (12/13), Utilities 0% success (0/14)
Data Quality: 388 screenshots validated, 387 healthy (99.74%), 1 exit bug found (SPY Alt39)

See DISCOVERIES_AND_LEARNINGS.md for complete analysis.

🔬 Methodology: How This Research Was Conducted

This project represents a systematic human-AI collaboration for discovering robust trend-following strategies. Here's the exact process used, which anyone can replicate:

1. Strategy Generation via AI

Process: I used AI (primarily ChatGPT and Claude) to generate individual Pine Script strategies that embody Ed Seykota's trend-following principles while testing unique underlying theories.

Key Prompt Pattern:

"Create a TradingView Pine Script strategy that follows Ed Seykota's philosophy of cutting losses and letting winners run. The strategy should test [SPECIFIC THEORY] such as [fractional pyramiding / momentum-gated entries / profit targets / etc.]. Include proper position sizing, ATR-based stops, and Donchian channel entries."

Why This Worked:

AI excels at generating syntactically correct Pine Script code
Each strategy tested a distinct hypothesis (not just parameter tweaks)
40+ variations created, each exploring different exit logic, sizing methods, or filters
AI provided consistent code structure while varying the experimental components

2. Manual Testing in TradingView

Process Overview: For each AI-generated strategy, I tested it across all 21 securities using an optimized workflow that dramatically reduced testing time.

Code Transfer Workflow (Cursor + Vim)

I used Cursor (a VS Code fork) with Vim keybindings for lightning-fast code copying:

Open Pine Script file in Cursor
Click anywhere in the file
Press gg to jump to the top
Press Shift+V to enter Visual Line mode (selects entire line)
Press G (uppercase) to extend selection to the bottom of the file
Press Ctrl+C to copy (standard Windows clipboard)
Alt+Tab to TradingView
Paste into Pine Editor

Why This Matters: The entire code selection process takes <2 seconds. No mouse dragging, no scroll-and-shift-click, just muscle memory keystrokes.

Date Range Selection (One-Time Setup Per Strategy)

When first loading a new Pine Script strategy in TradingView:

TradingView prompts for backtest date range
I selected dates with my mouse:
- Start Date: Beginning of 2010 (roughly January 2010)
- End Date: November 1, 2025 (current date at time of testing)
This range was locked in for that strategy session

Critical Optimization: Once the strategy and date range were configured, TradingView remembered these settings. This meant I could systematically test all 21 securities without having to:

Re-select the Pine Script strategy for each ticker
Re-enter the date range for each ticker
Manually configure any other strategy settings

Ticker Switching Workflow (The Speed Secret)

After setting up one strategy with its date range, testing all 21 securities became extremely fast:

Alt+Tab to TradingView (from documentation/screenshot workflow)
Start typing the next ticker symbol (e.g., "M", "S", "F", "T")
TradingView automatically opens the ticker search bar
TradingView automatically recognizes whether I'm searching for:
- Individual stock (MSFT, GOOGL, AMZN, etc.)
- ETF (SPY, QQQ, XLV, XLE, etc.)
Press Enter to load the ticker
Strategy immediately runs on the new security with same date range
Review results, capture screenshot (see below)
Repeat for next ticker

Time Savings: This workflow reduced per-ticker testing from ~2-3 minutes (if reconfiguring each time) to ~15-30 seconds. Testing all 21 securities for one strategy went from 40-60 minutes to approximately 5-10 minutes.

Example Workflow for Alt47:

Load Alt47 Pine Script, set 2010-2025 date range
Type "FCX" + Enter → Test → Screenshot
Type "GOOGL" + Enter → Test → Screenshot
Type "AMZN" + Enter → Test → Screenshot
... (18 more securities)
Complete all 21 tests in <10 minutes

Screenshot Capture Workflow

For every backtest result:

Strategy runs and displays results in TradingView
Press Win + Shift + S (Windows Snipping Tool)
Draw selection box around Strategy Tester panel
Screenshot auto-saves to Windows Photos folder with timestamp
Later: Navigate to Photos folder, move screenshots to sector folders
Rename to descriptive format: alt[X]-[result]_reduced.png

PowerShell Size Reduction: After capturing batches of screenshots, I ran resize_images.ps1 to reduce them by 35% (to 65% of original size). This was necessary because TradingView's high-resolution screenshots often exceeded Claude Code's image processing limits.

Standardized Backtest Settings

Every test used identical parameters:

Initial Capital: $100,000
Commission: 0.005% (0.5 basis points)
Slippage: 2 ticks
Backtest Period: ~2010-2025 (approximately 15 years of data)

Why Manual Testing Over Automation

Despite the manual workflow, this approach provided advantages automated testing couldn't:

Visual equity curve analysis: Smooth vs choppy growth patterns immediately visible
Zoom level indicators: TradingView auto-zooms when exits are rare (critical warning sign for options traders)
Trade distribution patterns: Clustering of trades reveals market regime sensitivity
Drawdown visualization: Shape and frequency of drawdowns matter, not just max DD number
Quick pattern recognition: Human brain excels at spotting visual anomalies across hundreds of charts

Result: 294 backtests completed across 14 strategies and 21 securities with high-quality visual documentation of every result.

3. Project Infrastructure via Claude Code

Process: I used Claude Code (AI coding assistant) to build the entire project skeleton:

Folder Structure Created:

sectors/
├── technology-MSFT/
│   ├── screenshots/
│   └── strategies/
├── healthcare-UNH/
│   ├── screenshots/
│   └── strategies/
├── energy-XOM/
├── utilities-DUK/
├── etfs-SPY/
├── etfs-QQQ/
├── etfs-XLV/
... (11 sectors total, 21 securities)

What Claude Code Automated:

Created 21 sector folders (11 individual stocks + 10 sector ETFs)
Generated standardized subfolder structure (screenshots/, strategies/)
Built utility scripts (resize_images.ps1 for managing screenshot sizes)
Created documentation templates (DISCOVERIES_AND_LEARNINGS.md, analysis.csv)
Set up naming conventions and file organization systems

Why This Mattered:

Systematic organization prevents chaos when testing 210+ strategy-security combinations (10 strategies × 21 securities)
Automated folder creation ensured consistency across all sectors
Templates enforced disciplined documentation from day one

4. Screenshot Capture Workflow

Process: For every strategy backtest, I captured evidence using Windows' built-in screenshot tool:

Steps:

Run backtest in TradingView
Press Win + Shift + S (Windows Snipping Tool)
Draw selection box around TradingView's Strategy Tester panel
Screenshot automatically saved to Windows Photos folder with timestamp
Navigate to Photos folder, locate the screenshot
Move screenshot from Photos to appropriate sectors/[ticker]/screenshots/ folder
Rename file to descriptive format: alt[X]-[result]_reduced.png
- Example: alt26-very-profitable_reduced.png

Filename Conventions:

alt[X] = Strategy number being tested
[result] = Performance category (very-profitable, profitable, scratch, unprofitable, catastrophic)
_reduced = Indicates image has been resized (see next section)

Why Screenshots:

Visual proof of every backtest result
Equity curve reveals characteristics automated metrics miss
Screenshots show TradingView's complete performance panel (trades, win rate, PF, drawdown, equity curve)
Creates audit trail - anyone can verify results

5. Image Size Management

The Problem: TradingView's high-resolution screenshots often exceeded 3000 pixels wide, which Claude Code couldn't process for analysis.

The Solution: Created resize_images.ps1 PowerShell script:

What It Does:

# Reduces all images by 35% (to 65% of original size)
# Recursively finds PNG/JPG files in all sector folders
# Skips files already containing "_reduced" in filename
# Uses high-quality bicubic interpolation
# Deletes original large files after successful resize
# Adds "_reduced" suffix to processed files

Why 35% Reduction:

Brings 4000px+ screenshots down to ~2600px (within Claude Code's limits)
Preserves visual quality for analysis (bicubic interpolation)
Reduces repository size significantly (hundreds of screenshots)
Prevents recursive resizing (skips files with _reduced suffix)

Usage:

.\resize_images.ps1
# Processes all new screenshots in one batch
# Safe to run multiple times (won't re-process reduced images)

6. Screenshot Validation and Bug Detection

The Problem: After capturing 388 backtest screenshots, a critical quality issue emerged: some strategies appeared profitable but were actually defective. The "exit bug" caused strategies to stop executing exits partway through the test period, accidentally becoming buy-and-hold instead of actively trading.

The Discovery: During analysis, one result looked suspicious - SPY Alt39 showed +121.89% returns but with visual anomalies:

Purple exit arrows (marking position closures) stopped appearing after ~2017
Equity curve began tracking the price line closely after 2017
High returns with relatively few recent trades

This revealed a coding bug where the exit logic failed partway through the backtest, turning an active strategy into passive buy-and-hold. The inflated returns were misleading.

The Solution: Systematic Visual Validation

Created bug-log.txt and data-verification-plan.md to systematically audit all 388 screenshots:

What We Check For:

Purple exit arrows - Should appear consistently from 2010 through 2024-2025
- ❌ RED FLAG: Arrows stop partway through (especially after ~2017)
Equity curve behavior - Should diverge from price line (shows strategy activity)
- ❌ RED FLAG: Green line starts tracking red price closely after arrows stop
Performance stats - Bottom left P&L vs right benchmark
- ❌ RED FLAG: Numbers within ~10% of each other AND arrows have stopped
Trade frequency - Should see consistent activity throughout timeframe
- ❌ RED FLAG: High returns with no recent exit activity

Key Insight: Left P&L ≈ Right benchmark alone is NOT a bug! Some strategies legitimately perform similar to buy-and-hold while still actively managing exits. Only the combination of similar numbers + stopped arrows indicates a defect.

Results:

Total images analyzed: 143+ (ongoing)
Bugs found: 1 (SPY Alt39 only - isolated incident)
Status: All other verified images show healthy exit activity throughout 2010-2025

Why This Matters:

Data integrity: Prevents counting false positives as strategy successes
Transparency: Every result is visually validated, not just trusted
Reproducibility: Anyone can audit the screenshots using the same criteria
Research quality: Only legitimate results inform strategy development decisions

The Bug Pattern (INVALID results):

❌ Purple arrows STOP partway through chart
❌ Equity curve follows price closely after arrows stop
❌ Suspiciously high returns with no recent exits
❌ Strategy accidentally became buy-and-hold

Healthy Pattern (VALID results):

✅ Purple arrows continue throughout 2010-2025
✅ Equity curve diverges from price (shows strategy activity)
✅ Trade markers present through recent periods
✅ Strategy is actively managing positions

Documentation:

bug-log.txt - Master list of confirmed bugs and verified images
data-verification-plan.md - Complete validation protocol with visual examples
Each sector folder contains only validated screenshots

Ongoing Process: New screenshots are validated before being included in analysis, ensuring research findings are based on legitimate strategy performance, not coding defects.

7. Analysis and Documentation

Process: As results came in from each validated strategy-security combination, I:

Recorded metrics in analysis.csv:
- Return %, Profit Factor, Win Rate %, Max Drawdown %
- Trade count, performance category, options suitability
Documented insights in DISCOVERIES_AND_LEARNINGS.md:
- What worked and why
- What failed and why
- Sector-specific patterns
- Strategy suitability by asset type
Used Claude Code for pattern recognition:
- Bulk analysis of screenshot results
- Aggregating performance across strategies
- Identifying strategy rankings and tier classifications
- Generating comparison tables and statistical summaries

Key Insight from This Process:

After testing 210+ strategy-security combinations, clear patterns emerged: Alt10 (Profit Targets) and Alt26 (Fractional Pyramid) both achieved 76% success rates, while utilities failed 100% of all strategies tested. These discoveries only emerged through systematic, comprehensive testing.

Why This Methodology Works

Reproducibility: Every step is documented and can be replicated by anyone with:

TradingView account (free version works)
AI access for Pine Script generation (ChatGPT, Claude, etc.)
Windows computer for screenshot workflow
PowerShell for image processing

Transparency: Nothing is hidden:

All 40+ strategy codes are in pine-scripts/
All screenshots are in sectors/ folders
All analysis is in analysis.csv and DISCOVERIES_AND_LEARNINGS.md
Failures documented alongside successes
Bug detection: Defective results openly identified and invalidated (bug-log.txt)

Systematic Discovery: Not cherry-picking:

Every strategy tested on same set of securities
Standardized backtest settings eliminate variables
Failed strategies kept for learning (not deleted)
Honest documentation of what doesn't work
Visual validation: Every screenshot checked for exit bugs before analysis

Human-AI Synergy: Leveraging strengths of both:

AI: Generate code variations, build infrastructure, analyze patterns
Human: Visual pattern recognition, hypothesis formation, judgment calls
Together: Systematic testing at scale with qualitative insights

How You Can Replicate This

Generate strategies: Ask AI to create Pine Scripts testing specific theories
Setup infrastructure: Use Claude Code to create folder structure
Test systematically: Run each strategy on each security, capture screenshots
Manage images: Use resize_images.ps1 to handle screenshot sizes
Validate screenshots: Check for exit bugs using data-verification-plan.md
Document findings: Use templates provided to record insights
Iterate: Learn from failures, refine theories, generate new strategies

The beauty of this approach: It's accessible to anyone. No expensive software, no advanced coding skills, no proprietary data. Just systematic application of free tools and disciplined documentation. Plus quality assurance: Visual validation ensures only legitimate results drive strategy decisions.

📂 Repository Structure

trend-following-backtesting-strategies/
├── START_HERE.md                    ← Begin here for sector testing guide
├── README.md                        ← You are here (project overview)
├── pine-scripts/                    ← All 40+ TradingView strategy files
├── sectors/                         ← Sector-by-sector testing results
│   ├── technology/
│   ├── healthcare/
│   ├── energy/
│   └── ... (11 sectors total)
├── resize_images.ps1                ← Utility to manage screenshot sizes
├── bug-log.txt                      ← Bug detection log (quality assurance)
├── data-verification-plan.md        ← Screenshot validation protocol
└── Documentation files:
    ├── DISCOVERIES_AND_LEARNINGS.md    ← Document your findings
    ├── SECTOR_TESTING_WORKFLOW.md      ← Detailed testing methodology
    ├── FINAL_10_STRATEGIES.md          ← Why these 10 for sector testing
    ├── STREAMLINED_TESTING_GUIDE.md    ← Complete testing protocol
    └── ... (additional guides)

🏆 The Champions

After testing 40 different strategy variations over 3 iterations, three strategies emerged as clear winners:

🥇 Alt 31: Fractional + Breakeven Lock (+40.1%)

THE UNDISPUTED KING

Profit Factor: 1.474
Max Drawdown: 12.46%
Win Rate: 51.82%
Why it dominates: The perfect combination of intelligent position sizing (fractional pyramiding) and ironclad protection (breakeven locks). This strategy has the smoothest equity curve of all tested strategies.

🥈 Alt 38: Triple Combo (+38.9%)

PHENOMENAL PERFORMANCE

Profit Factor: 1.466
Max Drawdown: 12.45%
Win Rate: 51.65%
The trifecta: Combines fractional pyramiding, breakeven locks, AND momentum-gated entries. Three proven winners working in harmony.

🥉 Alt 34: Fractional + Momentum (+33.8%)

BRONZE MEDAL

Profit Factor: 1.396
Max Drawdown: 12.45%
Win Rate: 50.73%
Smart additions: Fractional pyramiding enhanced with RSI-gated position adds for improved trade quality.

📊 Performance Overview

Backtest Period: January 2022 - October 2025 (SPY Daily) Initial Capital: $100,000 Strategies Tested: 40 variations across 3 iterations

Success Rate Evolution

Iteration 1 (Alts 1-20): ~30% profitable
Iteration 2 (Alts 21-30): 70% profitable
Iteration 3 (Alts 31-40): 90% profitable! 🎯

Performance Journey

Iteration	Best Strategy	Return	Improvement
1	Alt 10 (Profit Targets)	+20.3%	Baseline
2	Alt 26 (Fractional Pyramid)	+33.5%	+65%
3	Alt 31 (Fract + BE Lock)	+40.1%	+97% total

Nearly doubled performance over 3 iterations!

🎯 Key Discoveries

The Golden Combo

Fractional Pyramiding + Breakeven Lock = Maximum Performance

This combination proved superior to any other two-element pairing. It delivers:

Best sizing strategy (100% → 75% → 50% → 25%)
Best protection (move stop to breakeven after 2N profit)
Lowest drawdown among top performers
Smoothest equity curve ("almost always up")

Critical Lessons Learned

✅ What Works

Profit Targets Are Essential - Multi-stage profit-taking (3N, 6N, 9N) dramatically outperforms all-or-nothing exits
Pyramiding Is Mandatory - Single-position strategies catastrophically underperform (-21.8%)
Slower Entries Filter Noise - 55-bar Donchian >> 20-bar breakouts (40% vs -93%)
Exit Optimization > Entry Optimization - Where you get out matters more than where you get in
Simplicity Beats Complexity - Two well-chosen elements outperform kitchen-sink approaches
Fractional Sizing Wins - Descending position sizes (100%→75%→50%→25%) beat equal sizing
Breakeven Locks Work - Moving stops to breakeven after 2N profit reduces risk without sacrificing returns

❌ What Fails

Fast Breakouts - 20-bar entries = -93% (death by whipsaws)
No Pyramiding - Single positions = -21.8% (missed compounding)
Adaptive Entry Levels - Keltner channels = -58.5% (complexity doesn't help)
Alternative Trailing Methods - Parabolic SAR = -16.8% (Chandelier superior)
Excessive Filtering - ADX, dual timeframe, anti-chop filters all reduce performance
Time-Based Exits - Markets don't care about your calendar = -21.7%
Too Many Features - "ULTIMATE" strategy with all features only placed 4th

📁 File Organization

All strategy files are ranked by performance using this naming convention:

RANK_PF-X.XXX_strategy-name.pine

Example:

01_PF-1.474_seykota_alt31_fractional_breakeven.pine  👑 KING
02_PF-1.466_seykota_alt38_triple_combo.pine          🥈 RUNNER-UP
03_PF-1.396_seykota_alt34_fractional_momentum.pine   🥉 BRONZE
...
24_PF-0.131_seykota_alt1_fast_breakout.pine          💀 DISASTER

Reading the Filename

RANK: Performance ranking (1 = best)
PF-X.XXX: Profit Factor (>1.0 = profitable, >1.3 = excellent)
strategy-name: Descriptive name of the approach

Quick Reference Tiers

PF > 1.40: Elite performers (top 2)
PF 1.30-1.40: Excellent strategies (ranks 3-6)
PF 1.20-1.30: Solid performers (ranks 7-14)
PF 1.00-1.20: Profitable but mediocre (ranks 15-19)
PF < 1.00: Net losers (ranks 20-24)

🔬 Strategy Components

Core Elements (Present in All Winners)

Entry: 55-bar Donchian breakout (slow and steady)
Position Sizing: 1% risk per unit, 4 units maximum
Initial Stop: 2N (2× Average True Range)
Profit Targets: Multi-stage at 3N, 6N, 9N
Trailing Stop: Chandelier (ATR-based) for remainder

Enhancement Elements (Top Performers)

Fractional Pyramiding (Alts 31, 38, 34, 26, 35, 40, 36)
- Position sizes: 100% → 75% → 50% → 25%
- Adds every 0.5N in profit
Breakeven Lock (Alts 31, 38, 21, 36, 37, 40)
- Move stop to entry after 2N profit
- Guarantees no loss on mature positions
Momentum Gating (Alts 38, 34, 30, 36, 32, 40)
- Only add to positions when RSI confirms strength
- Improves pyramid quality
Time-Based Tightening (Alts 35, 37, 32, 25, 40)
- After 30 bars, tighten trail from 3N to 1.5N
- Protects aging positions

💀 Notable Failures

Understanding what doesn't work is as important as knowing what does:

Worst Performers

Alt 1: Fast Breakout (-93.2%, PF 0.131)
- 20-bar entries = constant whipsaws
- 280 trades with 18.6% win rate
- Lesson: Speed kills
Alt 23: Keltner Channel (-58.5%, PF 0.744)
- Adaptive EMA±ATR entry levels completely failed
- Complexity doesn't improve signal quality
- Lesson: Stick to simple Donchian breakouts
Alt 15: Single Position (-21.8%, PF 0.177)
- Only 12 trades in 3+ years
- 8.3% win rate (catastrophic)
- Lesson: Pyramiding is essential
Turtle Core v2.1 (-20.9%, PF 0.715)
- 10-bar exit too tight, gave back all gains
- Good entries, terrible exits
- Lesson: Exits matter more than entries

🚀 Getting Started

For Traders New to This Repository

Start here if you want to:

Use These Strategies for Trading → Read START_HERE.md then test on YOUR instruments
- Don't assume Alt 31 works on everything (it doesn't!)
- Test the 10 representative strategies on your target tickers
- Follow the 4-phase protocol to find what works for YOU
Learn From the Research → Read the complete strategy analysis:
- This README for overview and top performers
- DISCOVERIES_AND_LEARNINGS.md for detailed findings
- Individual .pine files to see implementation details
Replicate the Study → All strategies and methodology are documented:
- 40+ TradingView Pine Script files in pine-scripts/
- Testing methodology in SECTOR_TESTING_WORKFLOW.md
- Screenshot evidence in sectors/ folder
Contribute Your Findings → Test additional sectors/instruments:
- Follow the testing protocol
- Document your results
- Share discoveries with the community

For Live Trading (SPY Only - Validated)

Recommended strategies for SPY (in order):

Alt 31 - Best overall performance, smoothest equity curve (PF: 1.474)
Alt 38 - Highest profit factor, slightly more aggressive (PF: 1.466)
Alt 34 - Excellent risk-adjusted returns (PF: 1.396)

All three strategies have:

Similar low drawdowns (~12.5%)
Win rates >50%
Profit factors >1.39
Proven performance across bull/bear markets (2022-2025)

⚠️ Important: These results are specific to SPY. Test thoroughly on other instruments before live trading.

For Further Development

Promising areas to explore:

Complete sector testing across all market segments
Identify sector-specific strategy patterns
Parameter optimization for top strategies per sector
Walk-forward analysis for robustness testing
Different timeframes (4-hour, weekly)
Portfolio approach with sector-appropriate strategies

Avoid these dead ends (learned the hard way):

Fast breakout periods (<40 bars)
Single-position approaches
Complex adaptive entry systems
Alternative trailing stop methods (Parabolic SAR, etc.)
Excessive filtering (ADX, regime, anti-chop)

📈 Complete Strategy Rankings

Top 10 Performers

Rank	Strategy	Return	PF	Win Rate	Drawdown
1	Alt 31: Fract + BE Lock	+40.1%	1.474	51.8%	12.46%
2	Alt 38: Triple Combo	+38.9%	1.466	51.7%	12.45%
3	Alt 34: Fract + Momentum	+33.8%	1.396	50.7%	12.45%
4	Alt 26: Fractional Pyramid	+33.5%	1.388	50.7%	15.49%
5	Alt 40: ULTIMATE	+31.3%	1.348	51.9%	12.47%
6	Alt 36: BE + Momentum	+28.4%	1.338	50.0%	15.75%
7	Alt 21: Breakeven Lock	+28.2%	1.335	49.8%	15.59%
8	Alt 35: Fract + Time	+28.1%	1.306	50.5%	12.49%
9	Alt 37: BE + Time	+26.4%	1.301	50.2%	15.67%
10	Alt 30: Momentum Pyramid	+23.6%	1.277	48.1%	16.04%

Solid Performers (11-20)

All profitable strategies above breakeven, ranging from +5.6% to +24.4%. See LESSONS_LEARNED.md for complete details.

Failures (21-24)

Net negative strategies kept for educational purposes. Study these to understand what NOT to do.

🧪 Sector Testing Methodology

The Discovery

After extensive testing on SPY, a critical insight emerged: strategy performance is highly dependent on the underlying asset's characteristics. What works brilliantly on SPY may fail on individual stocks or sector ETFs.

The Solution

A systematic sector testing framework to discover which strategies work for which market segments:

10 Representative Strategies Selected - Chosen to test different hypotheses (not just performance tiers):
- Baseline (Classic Turtle)
- Alt 2 (EMA Crossover - different entry)
- Alt 9 (Time Exit - different exit philosophy)
- Alt 10 (Profit Targets - early profit taking)
- Alt 15 (Single Position - no pyramiding)
- Alt 17 (Dual Timeframe - confluence)
- Alt 20 (Asymmetric - long/short bias)
- Alt 22 (Parabolic SAR - adaptive exits)
- Alt 26 (Fractional Pyramid - smart sizing)
- Alt 28 (ADX Filter - trend strength)
11 Market Sectors Covered:
- ETFs (QQQ, XLE, XLU)
- Technology, Healthcare, Energy
- Financials, Industrials, Utilities
- Consumer Cyclical/Defensive
- Communication Services, Real Estate, Basic Materials
4-Phase Testing Protocol (~15 min per ticker):
- Phase 1: Foundation (Baseline, Alt 15, Alt 26)
- Phase 2: Entry Filters (Alt 2, Alt 28, Alt 17)
- Phase 3: Exit Optimizations (Alt 10, Alt 22, Alt 9)
- Phase 4: Special Cases (Alt 20)

Expected Outcomes

Identify universal strategies (work across many tickers)
Identify sector-specific strategies (niche edges)
Discover trend-friendly vs difficult sectors
Build strategy-sector matching decision tree

🛠️ Repository Utilities

Image Resize Script

To manage screenshot file sizes efficiently, a PowerShell utility is included:

resize_images.ps1 - Automatically resize screenshots by 35%

Recursively finds all PNG/JPG images
Reduces to 65% of original dimensions
Adds _reduced suffix to processed files
Skips already-reduced images (prevents recursive resizing)
Deletes original large files
High-quality bicubic interpolation

Usage:

.\resize_images.ps1

This helps keep the repository size manageable when storing hundreds of backtest screenshots across multiple sectors.

📚 Documentation

Getting Started

START_HERE.md - Quick start guide for sector testing (read this first!)
STREAMLINED_TESTING_GUIDE.md - Complete testing methodology
QUICK_REFERENCE_10_STRATEGIES.md - Print-friendly strategy reference card

Strategy Analysis

FINAL_10_STRATEGIES.md - Why these 10 strategies were selected for sector testing
DISCOVERIES_AND_LEARNINGS.md - Document findings as you test

Sector Testing

SECTOR_TESTING_WORKFLOW.md - Detailed sector testing workflow
SECTOR_TICKERS.md - Recommended tickers for each sector (2-3 per sector)
QUICK_TEST_CHECKLIST.md - Testing checklist

Original SPY Research

Individual .pine files - TradingView Pine Script implementations (40+ strategies)
sectors/etfs/ - Original SPY backtest results and screenshots

🎓 Philosophy

This project embodies several key principles:

Ed Seykota's Wisdom

"The elements of good trading are: (1) cutting losses, (2) cutting losses, and (3) cutting losses."

Our implementation: Initial stops, profit targets, trailing stops, and breakeven locks.

Lessons from the Market

"Win or lose, everybody gets what they want out of the market."

Systems focused on profit-taking win. Systems that fear missing out (no profit targets) lose. The market gives you what your system asks for.

Our Corollary

"The goal of successful strategy development is to build the best system. Money is secondary. Complexity is secondary."

Simplicity + Profit Targets + Pyramiding = Winner

📊 Market Conditions Tested

The backtest period (2022-2025) includes diverse market conditions:

2022: Severe bear market (-18% SPY)
2023-2024: Recovery and new all-time highs
2024-2025: Continued volatility and uncertainty

Top strategies proved robust across all these environments, demonstrating true trend-following resilience.

🖼️ Reading Backtest Screenshots

Critical Warning: Zoom-Level Indicates Exit Frequency

When reviewing backtest screenshots, the equity curve's X-axis time range is extremely revealing:

Full Timeline (1993-2025):

Strategy has regular exits throughout the backtest period
Equity curve shows complete P&L history
Total P&L (left) matches equity curve endpoint (right)
✅ Suitable for options traders who need frequent exits

Zoomed Timeline (e.g., 2010-2012 only):

Strategy has very few exits - TradingView auto-zooms to where trades occurred
Equity curve only shows small portion of backtest
Total P&L (left) includes unrealized gains from long-held positions
Equity curve endpoint (right) shows value at zoom endpoint, NOT final value
❌ NOT suitable for options traders - positions held too long

Example: Alt 6 Pure ATR Stop shows +$282K Total P&L but -$15K equity at chart end. The equity curve is zoomed to 2010-2012 (early losing period) because the strategy holds positions without exiting from 2013-2025. Those unrealized gains inflate the Total P&L but aren't visible on the zoomed chart.

For Options Traders: Immediately discard any strategy screenshot with a zoomed equity curve. You need strategies that close positions regularly, which will show full timeline equity curves.

⚠️ Disclaimer

This is educational research, not investment advice.

Past performance does not guarantee future results
All backtests are subject to curve-fitting risk
Walk-forward testing recommended before live trading
Use appropriate position sizing for your risk tolerance
Consider paper trading before risking real capital

🔄 Project Status

Current Phase: Multi-Sector Testing & Analysis

Completed:

✅ 40 strategy variations tested on SPY
✅ 3 complete iterations
✅ Comprehensive documentation
✅ Clear performance rankings
✅ Golden Combo identified (Alt 31)
✅ Sector testing framework developed
✅ Image management utilities created

Recent Discovery: Alt 31 works great on SPY but may fail on other tickers. This critical finding led to the development of a systematic sector testing workflow to discover which strategies work for which market segments.

Current Focus:

Testing 10 representative strategies across all market sectors
Identifying sector-specific strategy performance patterns
Building strategy-sector matching system
Documenting findings for the trading community

Next Steps:

Complete sector testing (Technology, Healthcare, Energy, Utilities, etc.)
Develop sector-optimized strategy variants
Create strategy selector based on ticker characteristics
Begin paper trading with sector-appropriate strategies
Walk-forward analysis for robustness validation

📞 Contact & Contributions

This is an open research project documenting a systematic approach to trend-following strategy development. The methodology and results are shared for educational purposes.

How Other Traders Can Use This Research

1. Direct Application:

Clone/download the repository
Test strategies on your instruments using TradingView
Use the SECTOR_TESTING_WORKFLOW.md as your guide
Document your findings in the appropriate sector folders

2. Learn From the Process:

Study the iteration-by-iteration improvement (from -93% to +40%)
Understand what works and what fails (and why)
Apply the systematic testing methodology to your own strategy ideas
Avoid the documented "dead ends"

3. Contribute Your Findings: If you test these strategies on additional instruments/sectors:

Document your results using the provided templates
Share screenshots and performance metrics
Add insights to DISCOVERIES_AND_LEARNINGS.md
Help build the community knowledge base

4. Build Upon This Work:

Optimize parameters for specific sectors
Test on different timeframes
Develop sector-specific variants
Create automated strategy selectors
Share improvements back to the community

What Makes This Research Valuable

Transparency: Every strategy is documented with:

Complete Pine Script source code
Backtest screenshots
Performance metrics
Analysis of why it works (or fails)

Systematic Approach: Not cherry-picked results:

All 40 strategies tested on same timeframe
Failures documented alongside successes
Clear methodology anyone can replicate
Honest evaluation of what doesn't work

Actionable Insights: Beyond just "here's a profitable strategy":

Understanding of WHY strategies work
Knowledge of WHEN they fail
Framework for testing YOUR instruments
Methodology for continuous improvement

Key Takeaway

Through systematic testing and learning from failures, strategy performance improved by 97% over 3 iterations (-93% → +40%), proving that:

Disciplined experimentation beats guru predictions
Systematic testing reveals hidden patterns
Understanding failures is as valuable as celebrating successes
Strategy performance is asset-dependent (Alt 31 on SPY ≠ Alt 31 on everything)

The journey from failure to success is the real value, not just the final strategy.

"The goal of a successful trader is to make the best trades. Money is secondary." - Alexander Elder

Last Updated: November 2, 2025 Strategies Tested: 40 (SPY); 10 selected for multi-sector testing SPY Success Rate: 90% (Iteration 3) Best SPY Strategy: Alt 31 (+40.1%, PF 1.474) Current Phase: Multi-sector testing to identify strategy-asset matching patterns

Name		Name	Last commit message	Last commit date
Latest commit History 11 Commits
.claude		.claude
analysis-outputs		analysis-outputs
archive		archive
docs		docs
pine-scripts		pine-scripts
screeners		screeners
scripts		scripts
sectors		sectors
CLAUDE.md		CLAUDE.md
DISCOVERIES_AND_LEARNINGS.md		DISCOVERIES_AND_LEARNINGS.md
README.md		README.md
analysis.csv		analysis.csv
bug-log.txt		bug-log.txt
cleaned-data.csv		cleaned-data.csv
data-analysis-plan.md		data-analysis-plan.md
data-verification-plan.md		data-verification-plan.md

trustdan/trend-following-backtesting-strategies

Folders and files

Latest commit

History

Repository files navigation

Trend Following Strategy Development Project

💡 KEY TAKEAWAYS (293 Validated Backtests Across 21 Securities)

The Critical Discovery: Strategy Performance is Asset-Dependent

The Winners: Universal Strategies That Actually Work

The Sector Truth: Some Markets Are Untradeable

Latest Python Analytics & Basket Guidance

Record-Breaking Performances (All Validated)

Data Quality: 99.74% Validation Rate

The Options Trading Insight

What Actually Matters (Validated by 293 Backtests)

The Bottom Line for Traders

✅ Project Status: COMPLETED

🔬 Methodology: How This Research Was Conducted

1. Strategy Generation via AI

2. Manual Testing in TradingView

Code Transfer Workflow (Cursor + Vim)

Date Range Selection (One-Time Setup Per Strategy)

Ticker Switching Workflow (The Speed Secret)

Screenshot Capture Workflow

Standardized Backtest Settings

Why Manual Testing Over Automation

3. Project Infrastructure via Claude Code

4. Screenshot Capture Workflow

5. Image Size Management

6. Screenshot Validation and Bug Detection

7. Analysis and Documentation

Why This Methodology Works

How You Can Replicate This

📂 Repository Structure

🏆 The Champions

🥇 Alt 31: Fractional + Breakeven Lock (+40.1%)

🥈 Alt 38: Triple Combo (+38.9%)

🥉 Alt 34: Fractional + Momentum (+33.8%)

📊 Performance Overview

Success Rate Evolution

Performance Journey

🎯 Key Discoveries

The Golden Combo

Critical Lessons Learned

✅ What Works

❌ What Fails

📁 File Organization

Reading the Filename

Quick Reference Tiers

🔬 Strategy Components

Core Elements (Present in All Winners)

Enhancement Elements (Top Performers)

💀 Notable Failures

Worst Performers

🚀 Getting Started

For Traders New to This Repository

For Live Trading (SPY Only - Validated)

For Further Development

📈 Complete Strategy Rankings

Top 10 Performers

Solid Performers (11-20)

Failures (21-24)

🧪 Sector Testing Methodology

The Discovery

The Solution

Expected Outcomes

🛠️ Repository Utilities

Image Resize Script

📚 Documentation

Getting Started

Strategy Analysis

Sector Testing

Original SPY Research

🎓 Philosophy

Ed Seykota's Wisdom

Lessons from the Market

Our Corollary

📊 Market Conditions Tested

🖼️ Reading Backtest Screenshots

Critical Warning: Zoom-Level Indicates Exit Frequency

Packages