Systematic Trading Research

AlphaHybrid

Systematic Research. Verified Edge. Autonomous Execution.

View Research ↓ Follow @Yunus_defi →

0 Best Sharpe

0% BTC 3Y Return

0+ Tests Run

0 Live Systems

ALPHAHYBRID — BACKTEST ENGINE v3.1

$ run_backtest --asset BTCUSDT --tf 15m

✓ Signal: momentum exhaustion signal + participation intensity filter

▸ Period: 2021-01-01 → 2024-01-01

▸ Trades: 1,895 | Win Rate: 49%

▸ Return: +312% | Sharpe: 3.14

$ run_backtest --asset ETH --tf 1h

✓ Signal: Structural Inflection Detection + Volatility-Normalized Execution

▸ Trades: 122 | Win Rate: 61.3%

▸ Return: +247% | Sharpe: 2.71

$ deploy --system all

ES VOLATILITY COMPRESSION — 6M EQUITY

Edge Stack

The Systems

Three independently validated strategies. Each survived walk-forward testing and multi-regime filtering. All systems validated across multiple market regimes using walk-forward testing.

BTC • 15m • 2021-2024

BTC Momentum System

Momentum Exhaustion + Participation Intensity + Structural Containment

BTCUSDT Perpetual 15-minute 3-Year Period

+312%

3Y Return

3.14

Sharpe Ratio

15.9%

Max Drawdown

49%

Win Rate

1,895

Total Trades

ETH • 1h • 2024-2025

ETH Pivot Strategy

Structural Inflection Detection + Volatility-Normalized Execution

ETH Spot / Perp 1-hour 1-Year Period

+247%

1Y Return

2.71

Sharpe Ratio

11.4%

Max Drawdown

61.3%

Win Rate

122

Total Trades

ES FUTURES • 1m • 2025

ES Volatility Compression Scalper

Volatility Compression Breakout on E-Mini S&P 500

E-Mini S&P 500 1-minute 6-Month Period

+1,050

Ticks PnL

2.43

Sharpe Ratio

69.6%

Win Rate

1.70

Profit Factor

NinjaTrader 8

Execution Platform

Quantitative Research

What the Data Shows

2,500+ configurations tested. Here is what survived and what did not — no filter on inconvenient results.

✓

What Works

✓

Momentum exhaustion signal — core signal in all BTC combinations

CONFIRMED

✓

Structural inflection structures on ETH with proper volatility-normalized sizing

61.3% WR

✓

Volatility compression breakouts on ES — compression → expansion

CONFIRMED

✓

Price inefficiency window signals

68% CONFIGS PROFITABLE

✓

Tuesday / Thursday session bias

SHARPE 1.43 vs MON 0.33

✓

2 PM EST execution window — peak performance

SHARPE 1.75

✗

What Fails

✗

Imbalance-only signals — eliminated entirely

0/50 CONFIGS PROFITABLE

✗

Week 4 of month — consistently negative

SHARPE −0.70

✗

Standalone ADX, Parabolic SAR in crypto

FILTERED OUT

✗

1 PM EST hour — destroys edge consistently

SHARPE −0.32

✗

Volatility signals without trend filter

NO EDGE

✗

Overfit parameter sets — look-back bias excluded

WALK-FWD REJECTED

Infrastructure

The Infrastructure

Purpose-built research and execution stack — no off-the-shelf black boxes.

Parameter combos tested

2,500+

Out-of-sample coverage

Walk-Fwd

Live exchanges connected

3 APIs

ML model features (LSTM)

System uptime (VPS)

24/7

🐍

CUSTOM

Python Backtesting Engine

Custom-built framework. 2,500+ parameter combinations tested across multiple market regimes with walk-forward validation.

📉

LIVE

NinjaTrader 8

Strategy execution and live backtest validation for ES futures. Real-time orderflow execution with pre-defined risk parameters.

📡

LIVE

Sierra Chart + DTC Protocol

Real-time orderflow data pipeline. High-fidelity tick data feed for ES futures analysis and signal generation.

🧠

PyTorch Neural Network

LSTM-based 6-feature ML model for ES futures. CUDA GPU acceleration via NVIDIA for fast training iterations.

⚡

INFRA

VPS Infrastructure

24/7 automated execution on dedicated server. Zero downtime deployment with redundant connectivity and failover.

🔗

LIVE

Multi-Exchange API

Binance, Bitget, and OKX connectivity. Unified order routing with real-time position monitoring and reconciliation.

✅

VALIDATION

Walk-Forward Validation Engine

Out-of-sample testing on all strategies. Rolling window validation to prevent curve-fitting and confirm genuine edge across unseen market data. No strategy deployed without passing this gate.

MACHINE LEARNING ENGINE

Neural Intelligence Layer

Trained on tick data. Sees what no indicator can.

MODEL 01 — TRADING TRANSFORMER

Transformer Model 8 attention heads scanning 100 ticks at once. Learns which moments in the tape matter. Outputs a directional signal with a confidence score. CUDA-trained.

MODEL 02 — LSTM SEQUENCE MODEL

LSTM Sequence Model 3 layers deep. Trained on 2 years of ES tick data. Dropout-regularized. GPU-accelerated. Deployed through Sierra Chart's DTC bridge in real time.

TRAINING INFRASTRUCTURE

PRODUCTION

Sierra Chart → DTC bridge → real-time signal JSON

Temporal train/val/test split — no data leakage

CUDA GPU training | Mixed precision | Early stopping

Process

How We Research

A repeatable three-stage pipeline. No discretion, no gut calls.

STEP 01

Signal Universe

“Define all possible entry/exit combinations”

Enumerate every viable technical signal combination across assets and timeframes. Momentum regime filter variants, structural inflection structures, volatility compression metrics, participation intensity filters, price inefficiency window patterns. Each combination is parameterized and queued for systematic testing. No manual curation at this stage.

›

STEP 02

Mass Validation

“Run 2,500+ backtests, apply walk-forward testing”

All configurations are run through the Python backtesting engine. In-sample optimization is followed immediately by out-of-sample walk-forward validation. Configurations that only profit in-sample are discarded. Sharpe, drawdown, and profit factor thresholds must all be met simultaneously.

›

STEP 03

Regime Filter

“Only deploy what survives multiple market regimes”

Surviving configurations are stress-tested across trending, ranging, and volatile regimes. Time-of-day and day-of-week filters are applied using the discovered session biases. Only strategies with consistent cross-regime performance reach live execution on VPS infrastructure.

Copy Trading