Back to portfoliocurrently running
QuantClaw
Algorithmic Trading Engine
Two architectures, one mandate: don't lose money. A self-improving genetic system paired with an LLM decision engine running on real market data — strategy stays proprietary.
PythonPyTorchGroq APIFlaskAlpacaIsolation Forest
Live dashboard
data: simulated · click chart to scrubpaper.trading.live
env: simulated
AAPL230.00NVDA264.00TSLA267.00SPY263.00QQQ262.00MSFT260.00AMD242.00META246.00
equity_curve.liveclick to scrub
live signal
BUY
confidence: —
order_log.tail
How it works (vaguely).
01
Genetic Algorithm System
50 candidate bots × 50 generations of fitness selection, daily — the population breeds toward whichever rule-set survives the backtest.
02
LLM Decision Engine
An autonomous agent making trade decisions under hard risk rules — the model can hedge, hold, or halt, never override the risk layer.
03
Risk Controls
An Isolation Forest watches volatility in real time. Anomaly score above threshold halts trading instantly — capital preservation first.
Backtest, 12 months.
5-minute candles. 250 trades evaluated.
equity_curve · 12mo+20% net
win rate
0%
trades evaluated
0
paper P/L · live
+$0
order_book.NVDA · L2
bids
143.05428
142.95499
142.8525
142.7988
142.8440
142.60219
142.49245
142.55323
asks
143.18127
143.22221
143.31461
143.3176
143.36249
143.59266
143.69261
143.58217
signal_confidence · histogramμ ≈ 0.62
strategy is proprietaryBack to portfolio