Backtesting Engine

in-development

Desktop-first, Rust + CUDA backtesting that can sweep thousands of strategies in minutes, run offline on your own data, and surface results you can ship to production.

GitHub Coming Soon

View Benchmarks

66x

Faster with GPU

100+

Built-in indicators

46k

Strategies per second

Desktop vs WASM Performance

Benchmark context GPU: RTX 4090 • CPU: Ryzen 9950X • RAM: 64GB Data: 60k hourly candles (7 yrs) • Grid: 1,936 params Run: Nov (benchmark run)

Backtester vs WASM benchmark timings and relative speedups.
Implementation	Time	Strategies/sec	vs WASM
WASM (Browser)	2.8s	691	1.0x
Desktop (CPU only)	580ms	3,338	4.8x
Desktop (CUDA)	42ms	46,095	66.7x

Wall-clock times for the full 1,936-scenario grid; lower is better. Recorded on RTX 4090 + Ryzen 9950X.

CPU vs CUDA kernel (ALMA×ALMA grid sweep)

Each “pair” is a unique parameter-combination backtest over 200k candles. The CUDA kernel path keeps the full pipeline VRAM-resident for maximum throughput.

CPU backtestCUDA kernel

Small · 1,590 pairs

1.04× faster on CUDA

105ms → 102ms

Medium · 14,150 pairs

6.64× faster on CUDA

728ms → 110ms

Large · 58,300 pairs

16.3× faster on CUDA

2.71s → 166ms

Error bars show the min–max range from the benchmark run; markers show the point estimate (log-scaled time axis).

Native Performance Benefits

Direct hardware access without browser limits.

Optimized memory management tuned for Rust.

Full utilization of CPU cores.

Zero-overhead abstractions.

GPU Acceleration Impact

Test thousands of strategies simultaneously.

Massive parallel processing for sweeps.

Optimized data pipelines to keep GPUs fed.

Real-time strategy evaluation feedback.

Interactive Optimization Demo

VectorAlpha Backtester - Strategy Optimization

CUDA EnabledIteration 0/177,450

ALMA Parameter Optimization - Sharpe Ratio

Period [2-22] × Offset [0.10-0.19] × Sigma [0.2-0.34]

Loading 3D visualization...

Sharpe Ratio Distribution

< -0.5

2.1%

-0.5 – 0

17.4%

0 – 0.5

30.2%

0.5 – 1.0

25.1%

1.0 – 1.5

15.0%

1.5 – 2.0

7.5%

> 2.0

2.6%

Tested:0

Profitable:50%

Mean:0.42

Std Dev:±0.68

Best Strategy Performance

ParametersALMA(50, 0.85, 0.45)

Sharpe Ratio0

Total Return0%

Win Rate0%

Max Drawdown0%

Processing Performance

Throughput0/sec

GPU Utilization

94%

Est. Time Remaining3s

Recent Batch Results

P[2-22] O[0.10-0.19] S[0.20-0.34]2048 tests

P[22-42] O[0.19-0.28] S[0.34-0.48]2048 tests

P[42-62] O[0.82-0.91] S[0.41-0.55]✓ Best found

P[2-22] O[0.10-0.19] S[0.20-0.34]Processing...

Optimization Configuration

AlgorithmGenetic Grid

Search Space177,450 combos

Batch Size2,048 parallel

ConvergenceAdaptive ε=0.001

Walk Forward70/30 split

Risk MetricSharpe Ratio

Instant Feedback

See your strategy performance update in real-time as optimization progresses.

Hours to Seconds

Complete backtests that would take hours in minutes with GPU acceleration.

Find Better Strategies

Intelligent optimization explores more possibilities to find optimal parameters.

Monte Carlo Simulation

1,000 simulations showing possible portfolio outcomes over one trading year

Portfolio Value

Median Return

6.5%

50% of outcomes above this

95th Percentile Return

43.2%

Best 5% of outcomes

5th Percentile Return

-19.5%

Worst 5% of outcomes

Median Max Drawdown

-14.5%

Typical worst decline

95th Percentile Drawdown

-28.2%

Worst 5% drawdowns

Probability of Loss

33.0%

Chance of negative return

95th Percentile

75th Percentile

Median

25th Percentile

5th Percentile

Cross-Validation Analysis

Validating strategy robustness across different time periods

Walk-Forward Analysis Windows

In-Sample vs Out-of-Sample Sharpe

Overfitting Detection

Avg IS Sharpe

1.20

Avg OOS Sharpe

1.03

Avg Degradation

14.2%

Consistency

85.8%

Development Status

Working

Core backtesting engine
Basic indicators (SMA, EMA, RSI)
CSV data import
Basic Tauri UI
Multi-threading

In Progress

CUDA integration
Advanced risk metrics
Real-time data feeds
Strategy marketplace

Planned

Portfolio backtesting
Options strategies
ML integration
Cloud deployment

Detailed roadmap coming soon with public release