Skip to content

Optimal Execution of Portfolio Transactions

Authors: Robert Almgren, Neil Chriss | Year: 2000 | Journal: Journal of Risk, 3(2), 5-39

Thesis

There is a fundamental trade-off between execution speed and market impact when liquidating a portfolio position. Trading too fast incurs large temporary and permanent market impact costs; trading too slowly incurs inventory risk (the price may move against you). The Almgren-Chriss framework provides the optimal execution trajectory that minimizes the expected cost plus a risk penalty, yielding the efficient frontier of execution. The optimal strategy is a deterministic schedule that depends on the trader's risk aversion, the asset's volatility, and the market impact parameters. For a risk-neutral trader, the optimal strategy is linear (TWAP). For a risk-averse trader, the optimal strategy is front-loaded (trade faster initially).

Key Math

The trader must liquidate \(X\) shares over \(T\) time steps. The trajectory \(x_t\) (remaining shares) and trade list \(n_t = x_{t-1} - x_t\) minimize:

\[\min_{\{n_t\}} \; E[\text{Cost}] + \lambda \cdot \text{Var}[\text{Cost}]\]

With temporary and permanent impact:

\[\text{Temporary impact: } \quad h(v) = \eta \cdot v \quad \text{(linear in trade rate } v = n_t / \tau\text{)}\]
\[\text{Permanent impact: } \quad g(v) = \gamma \cdot v \quad \text{(linear in trade rate)}\]

The optimal trajectory (continuous-time limit):

\[x(t) = X \cdot \frac{\sinh\!\left(\kappa(T-t)\right)}{\sinh(\kappa T)}\]

where the urgency parameter \(\kappa = \sqrt{\frac{\lambda \sigma^2}{\eta}}\) controls front-loading. As \(\lambda \to 0\) (risk-neutral), \(x(t)\) becomes linear (TWAP). As \(\lambda \to \infty\) (infinitely risk-averse), all trading happens at \(t=0\).

Data & Method

  • Theoretical framework, no empirical data in the original paper.
  • Market impact parameters (\(\eta\), \(\gamma\)) are estimated from empirical transaction data in practice. For COMEX gold futures, typical estimates: \(\eta \approx 0.01\text{-}0.05\) bps per contract per unit trade rate, \(\gamma \approx 0.001\text{-}0.01\) bps permanent.
  • The model assumes arithmetic Brownian motion for unaffected prices and linear impact functions.
  • Extensions by Almgren (2003) add nonlinear temporary impact: \(h(v) = \eta \cdot |v|^{3/5} \cdot \text{sign}(v)\) (the "square root law").

Our Replication Verdict

CONFIRMED -- The framework is the industry standard for execution optimization. For gold and silver futures specifically: (1) Gold (GC) is highly liquid; temporary impact is minimal for positions under 500 contracts. The model is most relevant for large position changes (>1000 contracts) or during low-liquidity windows (Asian session, holidays). (2) Silver (SI) is less liquid; the impact parameters are 3-5x larger than gold's on a per-notional basis. Execution optimization matters more for silver. (3) The linear impact assumption holds well for gold but breaks down for silver during volatile periods (impact is convex). We use the Almgren (2003) square-root extension for silver. (4) The permanent impact component is debatable -- some evidence suggests it's smaller than modeled for precious metals futures due to deep order books.

Signal Mapping

  • Execution engine (SS5.7). The Almgren-Chriss trajectory is the core algorithm for all order execution.
  • Implementation: For each trade, the system computes the optimal trajectory given current volatility (\(\sigma\) from real-time data), estimated impact parameters (calibrated monthly from our execution data), and a risk-aversion parameter (\(\lambda\)) set per strategy.
  • Trade scheduling: Trend-following signals use lower \(\lambda\) (more willing to trade slowly) while mean-reversion signals use higher \(\lambda\) (need faster execution to capture short-lived dislocations).
  • Cost attribution: Post-trade, we decompose execution cost into permanent impact, temporary impact, and timing (market drift during execution) for performance attribution.

References

  • Almgren, R. & Chriss, N. (2000). "Optimal Execution of Portfolio Transactions." Journal of Risk, 3(2), 5-39.
  • Almgren, R. (2003). "Optimal Execution with Nonlinear Impact Functions and Trading-Enhanced Risk." Applied Mathematical Finance, 10(1), 1-18.
  • Gatheral, J. (2010). "No-Dynamic-Arbitrage and Market Impact." Quantitative Finance, 10(7), 749-759.
  • Bouchaud, J.-P., Farmer, J.D. & Lillo, F. (2009). "How Markets Slowly Digest Changes in Supply and Demand." In Handbook of Financial Markets: Dynamics and Evolution, 57-160.