What it means
A large stock order need not be executed all at once. Trading it quickly can increase market impact; trading it slowly leaves more time for the price to move before the order is complete. The Almgren-Chriss model, associated with Almgren and Chriss (2000), formalizes that trade-off as an execution-scheduling problem.
The model is about how to execute a given order, not whether a stock is worth owning or what its price will do next. Its answer depends on assumptions about liquidity, volatility, impact and the trader’s tolerance for execution risk. A schedule optimized under those assumptions is not a guaranteed outcome.
How it is calculated
In the standard discrete-time setup, an order is divided among trading intervals. A common simplified objective is min E[C] + λ Var(C), where:
Cis execution cost measured against a specified benchmark, such as the price when the decision to trade was made, in US$.E[C]is expected execution cost, in US$.Var(C)is the variance of that cost, in US$².λis the trader’s risk-aversion parameter, with units of 1/US$ in this formulation, so both terms have units of US$.
The schedule must also satisfy Σ n_k = X: n_k is the number of shares traded in interval k, and X is the total order in shares. Impact assumptions determine expected cost; a price-volatility assumption helps determine cost variance. The full model distinguishes temporary impact from permanent impact and can produce an optimal schedule under its stated assumptions. The objective above is a framework, not a standalone calculator: numerical results require specified impact functions, time intervals and a value of λ.
Worked example
Consider a hypothetical purchase of 1,000 shares at an arrival price of US$50 per share. The benchmark order value is 1,000 × US$50 = US$50,000. Suppose a trader compares two possible schedules using illustrative cost estimates:
| Schedule | Estimated expected cost | Estimated cost standard deviation |
|---|---|---|
| Complete quickly | US$40 | US$10 |
| Spread across several intervals | US$20 | US$30 |
- Convert each standard deviation to variance: the quick schedule has
(US$10)² = 100 US$²; the slower one has(US$30)² = 900 US$². - For an illustrative
λ = 0.05 per US$, the quick schedule’s objective isUS$40 + (0.05 per US$ × 100 US$²) = US$45. - The slower schedule’s objective is
US$20 + (0.05 per US$ × 900 US$²) = US$65.
Under these hypothetical inputs and this risk preference, the quick schedule scores lower. With λ = 0, the slower schedule instead scores lower because its estimated expected cost is US$20 rather than US$40. These figures illustrate the trade-off; they are not model-calibrated estimates or predictions of actual fills.
Why it matters before you trade
Order size alone does not describe execution difficulty. The same number of shares can represent a very different fraction of daily volume across stocks. A pre-trade analysis can examine size relative to liquidity, estimated impact, volatility and execution conditions before any order is placed.
The model also clarifies why minimizing expected cost is not the only possible objective. A slower schedule may lower estimated impact while increasing exposure to unpredictable price movements. Neither approach eliminates the bid-ask spread, fees, changing liquidity or differences between modeled and actual fills.
How PreTrAIde uses it
PreTrAIde’s deterministic quantitative layer estimates market impact with an Almgren-Chriss-style model. Its pre-trade analysis also considers estimated spread, participation in daily volume and daily volatility. For a user-built stock basket, it returns basket-level and line-level GO, NO-GO or MODIFY verdicts, with risk assessments, suggested execution strategies and plain-language reasoning.
“Almgren-Chriss-style” does not mean PreTrAIde claims to implement every feature of the original optimization model or to deliver its theoretical optimal schedule. PreTrAIde is decision-support software: it does not receive, transmit or execute orders, and users trade through their own brokers.
Common mistakes
- Treating expected cost as a quoted fee: Market impact is an estimate, not a broker charge or guaranteed fill price.
- Ignoring units: Cost variance is in US$²;
λmust be scaled consistently before adding its weighted value to expected cost. - Assuming slower always means cheaper: Lower estimated impact can come with greater exposure to price moves.
- Confusing execution risk with investment risk: The model addresses costs and uncertainty while completing an order, not the stock’s long-term prospects.
- Assuming estimates apply unchanged in every market: Liquidity and volatility can change while an order is being executed.