Kelly Criterion

The mathematical foundation of every position we take. No hardcoded lot sizes, no intuition — just optimal growth.

f* = (bp - q) / b

f* = fraction to bet · b = net odds · p = win probability · q = loss probability

The Kelly Criterion maximizes the expected logarithm of wealth. In trading, it prevents overbetting (risk of ruin) and underbetting (suboptimal growth). Every strategy's Kelly fraction is computed dynamically based on historical win rates, current volatility, and regime detection.

Why Kelly?

  • Mathematically optimal — Maximizes long-term growth rate
  • Risk-aware — Naturally sizes down in high-volatility regimes
  • No parameters — Only requires win probability and odds
  • Bankroll preservation — Cannot bet more than edge allows
  • Fraud detection — Sudden Kelly deviations trigger circuit breakers

RL-Based Trading

We treat trading as a sequential decision problem. The agent learns optimal policies through continuous interaction with market data.

DQN

Deep Q-Networks

State: market indicators, order book, portfolio. Action: buy/sell/hold quantities. Reward: PnL with risk penalty. Experience replay breaks temporal correlations.

PP

Policy Gradients

Direct policy optimization for continuous action spaces. PPO and A2C implementations with generalized advantage estimation (GAE).

MV

Multi-Agent Systems

Multiple RL agents operating on different timeframes and instruments. Centralized critic, decentralized execution (CTDE) architecture.

Market Signal Processing

Every indicator is computed in zero-copy Go pipelines. No unnecessary allocations — every nanosecond counts.

RSI

RSI

Relative Strength Index with customizable periods and smoothing. Wilder's and exponential variants. Oversold/overbought regime detection.

MACD

MACD

Moving Average Convergence Divergence. Signal line crossovers, histogram divergence, and momentum regime classification.

BB

Bollinger Bands

Volatility-based envelopes. Band width indicates market regime. Squeeze detection for breakout prediction.

ATR

ATR

Average True Range for volatility-adjusted position sizing. Used as direct input to Kelly Criterion calculation.

OB

Order Book

Real-time order book imbalance, depth pressure, and bid-ask spread analysis. Microstructure signals for short-term prediction.

VF

Volume Profile

Volume-weighted price levels. High-volume nodes as support/resistance. Delta analysis for directional conviction.

Meta Learning & Transfer Learning

Models don't stay static. They learn across trading sessions, adapt to new regimes, and continuously improve.

ML

Meta Learning

Model-agnostic meta-learning (MAML) enables rapid adaptation to new market conditions. After just K=10 episodes in a new regime, the model achieves near-optimal performance. The outer loop runs continuously on the Worker service.

TL

Transfer Learning

Knowledge from liquid markets transfers to less liquid ones. Feature extractor layers are frozen, only the policy head is fine-tuned. Reduces cold-start time by 80%. Cross-exchange and cross-instrument transfer supported.