⚡ FlashBack

High-performance backtesting engine for cryptocurrency strategies

Processes 210 million ticks per second using C# optimization techniques. Built to test trading strategies honestly - shows real results, not cherry-picked wins.

🎯 Real Results

I tested 58 variations of 10 popular strategies on 7.4M Bitcoin price points (2012-2026):

📊 Key Finding: 43% of strategies LOST money

Top Performers

• Z-Score Mean Reversion: +$480k, 77% win rate ✅
• Bollinger Bands: +$462k, 70% win rate ✅
• RSI (Oversold/Overbought): +$330k, 72% win rate ✅

Biggest Losers

• Fast EMAs (10-20 period): -$1M, 20% win rate ❌
• Fast SMAs: -$908k, 21% win rate ❌
• MACD variations: -$373k, 27% win rate ❌

Why? Bitcoin is choppy. Fast trend-following strategies get destroyed by whipsaw. Mean reversion strategies that fade extremes actually work.

⚡ Performance

Dataset: 7.4M minute-by-minute Bitcoin ticks (14 years)
Load Time: 1.65 seconds
Processing: 210M ticks/second
58 Strategy Tests: ~8 seconds total

How?

• readonly struct for cache locality
• Span<T> for zero-allocation iteration
• No LINQ in hot loops
• Value types over reference types

🚀 Quick Start

# Clone
git clone [https://github.com/iOptimizeThings/FlashBack.git](https://github.com/iOptimizeThings/FlashBack.git)
cd FlashBack

# Run
dotnet run -c Release

# Choose Option 1 to test all strategies
# Or Option 2 for quick single-strategy test

Requires: .NET 8.0

📊 Test Data

To reproduce the benchmark results (210M ticks/sec), this project was tested using Bitcoin 1-Minute Historical Data from Kaggle.

• Dataset Source: Bitcoin Historical Data (Kaggle)
• Specific File: btcusd_1-min_data.csv
• Timeframe: Jan 2012 - Present

Setup:

1. Download the .csv file from Kaggle.
2. Rename it to 1.csv (optional, for easier typing) or keep the original name.
3. Place it in the root directory or copy the path when prompted.

📈 Strategies Implemented

Trend Following:

• Simple Moving Average (SMA)
• Exponential Moving Average (EMA)
• Dual MA Crossover

Momentum:

• RSI (Relative Strength Index)
• MACD (Moving Average Convergence Divergence)
• Stochastic Oscillator

Volatility:

• Bollinger Bands
• ATR Breakout

Mean Reversion:

• Z-Score Statistical Arbitrage

Volume:

• VWAP (Volume Weighted Average Price)

Each strategy tested with multiple parameter combinations (fast/slow periods, thresholds, etc.)

📊 Output

The engine generates:

• Console summary with top/bottom performers
• Detailed text reports with every trade
• CSV export for Excel analysis
• Sharpe ratios, max drawdown, win rates

🎓 What I Learned

1. Most "textbook" strategies lose money on Bitcoin - The classic SMA(50) that everyone recommends? Lost $482k.

2. Mean reversion > Trend following for crypto - Bitcoin oscillates more than it trends. Strategies that buy dips and sell rips work better.

3. Parameter optimization matters - SMA(10) lost $908k but SMA(500) made $46k. Same strategy, different settings.

4. Win rate ≠ Profitability - Some strategies had 70%+ win rates but still lost money due to a few huge losing trades.

🛠️ Tech Stack

• Language: C# 10 / .NET 8.0
• CSV Parsing: nietras.SeparatedValues (fastest .NET CSV library)
• Architecture: Value types, Span, aggressive inlining
• Paradigm: Zero-allocation hot loops

🚀 Benchmark & Sharing

Want to see the engine in action? Follow these steps to generate the summary table:

1. Run the application: dotnet run -c Release
2. Select Option 1.
3. Enter the path to your dataset (e.g., 1.csv).
4. Wait for the analysis to complete (approx. 15-20 seconds for 14 years of data).

📢 Share the Results

If you find this project useful, feel free to use these templates to share your results!

🤝 Contributing

PRs welcome! Especially for:

• Additional strategies (Ichimoku, Fibonacci, etc.)
• Multiple timeframe analysis
• Portfolio optimization
• Walk-forward testing

📝 License

MIT

---

Note: This is educational software. Past performance ≠ future results. Don't trade with money you can't afford to lose.