Capital management — how much to allocate to each strategy, position, or asset class, and how to adjust those allocations dynamically as conditions change — is one of the highest-leverage decisions in trading and investment management. Poor capital management turns a portfolio of profitable strategies into an overall loss by over-allocating to strategies in bad regimes and under-allocating to those in good ones. Optimal capital management is equally important to signal quality as a determinant of live performance.
We have built capital management systems as components of broader trading platform engagements — both as standalone tools for portfolio managers and as automated components of systematic trading platforms. Our approach combines classical capital allocation theory (Kelly criterion, risk parity, mean-variance optimisation) with ML-based regime detection and dynamic adjustment.
Capital management AI capabilities
Dynamic position sizing
Position sizing models that adjust allocation to each trade or strategy based on current signal confidence, volatility regime, correlation with existing positions, and available risk budget. At its simplest, this means scaling position sizes with signal strength and inversely with volatility — higher conviction signals in low-volatility environments receive larger allocations; weak signals in volatile regimes receive minimal allocation. More sophisticated implementations use Kelly-based sizing with shrinkage (to account for estimation error in expected returns and volatility), correlation-adjusted sizing (to prevent over-concentration in correlated positions), and regime-conditional sizing (to reduce overall risk during historically adverse market regimes).
We build position sizing as a configurable service that receives signal outputs and returns position size recommendations with the calculation basis — so risk managers can understand and audit the sizing logic, and override it when they have information the model does not.
Portfolio-level risk management
Real-time monitoring of portfolio-level risk metrics with automatic alerts and configurable automated responses. We build risk monitoring systems that track: total portfolio VaR and CVaR with scenario decomposition; factor exposure concentration (net and gross exposure to market beta, sector factors, style factors); correlation risk within the portfolio; leverage and margin utilisation; and currency and geographic concentration. Alert thresholds are configurable by risk managers; responses range from notification-only through to automated position reduction for hard limits.
Drawdown protection and capital preservation
Systematic drawdown management is one of the most valuable features of automated capital management — but also one of the most difficult to calibrate correctly. Drawdown controls that trigger too easily cause strategies to sit out recoveries. Controls that trigger too late allow losses to compound to levels that are psychologically and economically difficult to recover from. We build drawdown management systems with configurable tiers: soft limits that trigger de-risking (reducing position sizes proportionally), medium limits that trigger defensive positioning (shifting to lower-volatility assets or cash), and hard limits that trigger full halt for human review.
Multi-strategy capital allocation
For platforms running multiple simultaneous strategies, dynamic capital allocation distributes available capital across strategies based on recent performance, current capacity (strategies have diminishing returns as capital allocated increases due to market impact), and portfolio-level risk contribution. We build multi-strategy allocation using optimisation frameworks that maximise risk-adjusted portfolio-level returns subject to minimum allocation constraints (to maintain strategy diversification) and maximum allocation constraints (to prevent over-concentration in any single strategy).
Transparency and human oversight
Every capital management decision our systems make is logged with its input data and calculation basis — so risk managers can review, audit, and understand every allocation and sizing decision. Automated actions are limited to pre-defined parameter ranges; decisions outside those ranges require human approval. We design capital management systems to augment the judgment of risk managers and portfolio managers, not to operate outside of their oversight.
