Overview Backtesting Paper Trading History

Meet ScalarQ

Autonomous TAO/USD perpetual futures trader. Powered by DeepSeek R1 671B running on a dedicated NVIDIA H100 GPU rented from Targon (Bittensor Subnet 4). Zero API costs. 13 proprietary signals. Hyperliquid execution.

LONG SHORT TIME PRICE
Paper
Mode
$10,000
Equity
Total PnL
0
Trades
R1 671B
Model

How ScalarQ Works

A dedicated trading loop running every 15 minutes. Fetches live market data, computes signals, asks DeepSeek R1 for analysis, and executes.

Market Data

Real-time TAO/USD price, 1h candles, L2 orderbook depth, funding rate, open interest, and 24h volume from Hyperliquid.

13 Signals

Multi-timeframe momentum (1h/4h/24h/7d), mean reversion, funding rate, OI change, volume spike, orderbook imbalance, BTC correlation.

DeepSeek R1 Analysis

Full 671B parameter model (MoE, 37B active) running on a dedicated Targon H100 GPU. Reasons through signals, generates trade decisions with confidence scores.

Infrastructure

Self-hosted inference on Targon (Bittensor Subnet 4) — Bittensor's own decentralized GPU network by Manifold Labs. No third-party API dependencies. The model runs 24/7 on dedicated hardware, keeping everything within the Bittensor ecosystem.

Model
DeepSeek R1 671B
GPU
NVIDIA H100 80GB
Quantization
FP8 (native)
Serving
vLLM 0.18
Provider
Targon (SN4)
Exchange
Hyperliquid
Inference
~10-15s
Cycle
15 min
Hyperliquid API price, candles, OI, funding Signal Engine 13 sigmoid-normalized DeepSeek R1 671B params (MoE) H100 80GB + FP8 Targon SN4 Trade Decision long / short / hold Execute on Hyperliquid

Signal System

13 TAO-specific signals normalized with sigmoid functions. Each signal has absolute meaning — 0.5 is neutral, above is bullish, below is bearish.

Momentum 1h
Momentum 4h
Momentum 24h
Momentum 7d
Mean Reversion
Funding Rate
OI Change
Volume Spike
Book Imbalance
BTC Correlation
BTC Momentum
TAO vs BTC

Phased Rollout

ScalarQ is being built in public. Paper trading first, then backtesting validation, then live execution with real capital.

ACTIVE

Phase 1: Backtesting

Walk-forward blind testing against historical TAO price data. Monte Carlo validation with 1000 shuffles. Overfit detection. Strategy must prove itself on unseen data first.

Phase 2: Paper Trading

Once backtests pass, virtual $10K portfolio trading against real Hyperliquid prices. DeepSeek R1 analysis every 15 minutes. Full position tracking and PnL.

Phase 3: Hyperliquid Live

Automated TAO-PERP trading with real USDC. Start small ($100-500), scale with proven performance. 1-5x leverage.

Agent Log

Connecting...

Live stream of ScalarQ's thinking — market analysis, signal readings, trade decisions, backtest results, and learning insights.

Waiting for agent log...

Full technical documentation — signals, risk management, model architecture, infrastructure.

Read the Docs