Capstone 6 - Algorithmic Execution
Split a large order to balance market impact against timing risk - the Almgren–Chriss trade-off, TWAP/VWAP benchmarks, and honest impact modeling.
Leads to: Execution cost is the bridge from paper signals to realized P&L; it underpins Capstones 7–8.
Learning Objectives
Click a status chip to cycle: Not started → In progress → Studied → Practiced → Needs review → Mastered.
- Explain the execution trade-off between market impact and timing (volatility) risk.
- Derive the mean–variance execution objective and interpret the risk-aversion parameter.
- Implement a scheduled execution (TWAP and a front/back-loaded schedule) and measure impact.
- Compute implementation shortfall against an arrival-price benchmark.
- Explain why aggressive execution raises impact while slow execution raises risk.
Key Vocabulary
- Market impact
- The adverse price move caused by your own trading; larger and faster orders push price against you.
- Timing risk
- The risk that price drifts while you execute slowly; grows with volatility and horizon.
- Implementation shortfall
- Realized cost vs. the arrival (decision) price: the honest all-in measure of execution quality.
- TWAP / VWAP
- Time- (or volume-) weighted average price: standard execution benchmarks and schedules.
- Almgren–Chriss
- A mean–variance framework trading expected impact cost against the variance of execution cost.
- Participation rate
- The fraction of market volume your order represents; higher rates mean more impact.
Intuition & Motivation
The classic Almgren–Chriss formulation makes it a mean–variance problem: minimize expected impact cost plus a risk-aversion times the variance of cost. Low risk-aversion → trade slowly, accept timing risk to save impact; high risk-aversion → trade fast, pay impact to cut risk. The honest yardstick for any schedule is implementation shortfall: what did the whole execution cost relative to the price when you decided to trade?
The brief
Simulate executing a large parent order under a simple market-impact model. Compare schedules (immediate, TWAP, and a risk-adjusted Almgren–Chriss schedule) on the mean and variance of implementation shortfall. Deliverable: code, an efficient-frontier plot of expected cost vs. cost variance, and a note on which schedule suits which risk-aversion.
Required theory recap
- Impact (Phase 16): temporary impact \(\propto\) trading rate; permanent impact shifts the mid-price.
- Timing risk (Phase 8/12): price variance over the execution horizon \(\propto\sigma^2 T\).
- Mean–variance (Phase 10): minimize \(\E[\text{cost}]+\lambda\,\Var[\text{cost}]\); convex in the schedule.
- Benchmarks: TWAP (uniform in time), VWAP (proportional to volume), arrival price (for shortfall).
The execution trade-off, made precise
Split a parent order of \(X\) shares into a schedule \(x_1,\dots,x_n\) summing to \(X\). A simple linear temporary-impact model says each child order pays impact proportional to its size, and while shares remain un-executed they carry volatility risk. The Almgren–Chriss objective is:
Here \(\eta\) is the impact coefficient, \(\sigma\) the volatility, and \(\lambda\ge0\) the risk-aversion. The first term punishes trading fast (large \(x_k\)); the second punishes holding inventory (large remaining position) in a volatile market.
Implementation shortfall: the honest scorecard
Every execution is graded against the arrival price - the price at the moment of the decision. Beating VWAP can still lose money if the whole market moved; shortfall captures the true cost, including the impact you caused and the drift you suffered. It ties directly back to the transaction-cost realism of Capstone 5: this is where those ‘costs’ come from.
Impact is modeled explicitly and conservatively; every schedule is scored by shortfall against arrival price; the frontier is produced with fixed seeds (17.3) so it is reproducible.
- Executing a large order in one clip and ignoring the market impact it causes.
- Benchmarking only against VWAP and declaring victory while losing money vs. arrival price.
- Assuming zero impact - the same optimism as ignoring transaction costs (Capstone 5).
- Treating the schedule as risk-free; slow execution carries real timing (volatility) risk.
- Choosing \(\lambda\) to make a backtest look good rather than from genuine risk appetite.
- Confusing temporary impact (relaxes after you stop) with permanent impact (moves the mid for everyone).
- Always report implementation shortfall against arrival price, not just vs. VWAP.
- Trade slower when impact dominates (large order, thin market); faster when volatility/timing risk dominates.
- Model impact conservatively; underestimating it is how paper strategies die in production.
- Present an execution efficient frontier (expected cost vs. variance), and pick a point by stated risk aversion.
- Feed realized execution costs back into the strategy’s net returns (18.12) - execution is part of the P&L, not an afterthought.
Interactive: TWAP vs. front-loaded execution and shortfall
Build the schedules and an impact-based shortfall, and confirm the trade-off: faster execution lowers timing-risk variance but raises expected impact cost.
Knowledge Check
Practical Exercise
You must sell 500,000 shares of a stock that trades ~5,000,000 shares/day, with daily volatility 2%. (a) Qualitatively, should you lean toward fast or slow execution, and what two quantities drive the decision? (b) Sketch how the expected-cost-vs-cost-variance frontier shifts if the stock’s volatility doubles. (c) State the single benchmark you would grade the execution against and why.
(a) The order is 10% of daily volume - substantial, so impact is a real concern, pushing toward slower execution; but 2% daily vol means meaningful timing risk over a long horizon, pushing toward faster. The two driving quantities are the participation rate (order size vs. market volume, driving impact) and the volatility over the execution horizon (driving timing risk). A moderate, possibly front-loaded schedule balances them; the exact point comes from your risk-aversion \(\lambda\).
(b) Doubling volatility inflates the timing-risk (variance) term for any given schedule, so the whole efficient frontier shifts up/right (more cost variance at each expected-cost level). The optimal response is to trade faster (front-load more) to cut exposure, accepting higher expected impact - the frontier’s minimum-cost-variance region moves toward quicker schedules.
(c) Grade against implementation shortfall vs. arrival price: it is the honest all-in cost, capturing both the impact you caused and any adverse drift while you traded. VWAP alone can flatter a bad execution when the whole market moved against you.
Lesson Summary
Retrieval Practice
Close the lesson and answer from memory before checking. This is deliberate, effortful recall - the single highest-yield study action.
A: Market impact (rises when you trade fast) versus timing/volatility risk (rises when you trade slow). The Almgren–Chriss objective minimizes expected impact plus \(\lambda\) times cost variance; higher \(\lambda\) → faster/front-loaded execution, lower \(\lambda\) → slower TWAP-like execution.
A: Against the arrival (decision) price. It captures the all-in cost - both the impact your own trading caused and any adverse market drift during execution - unlike VWAP, which can flatter an execution when the whole market moved.
Completion Checklist
- I can explain the core ideas in my own words
- I worked the derivations/examples by hand
- I completed the interactive workbench(es)
- I passed the knowledge check
Source References
This lesson synthesizes and paraphrases concepts from the sources below. No copyrighted text, problem sets, or solutions are reproduced. Return to the originals for full depth.
- Algorithmic and High-Frequency Trading (Cartea, Jaimungal & Penalva, 2015) current - Ch. 6-8 - Optimal execution, market impact, the Almgren–Chriss framework, and implementation shortfall.
- Additional Exercises for Convex Optimization (Boyd & Vandenberghe, 2016) current - Ch. 4 - Mean–variance / convex optimization structure of the execution scheduling problem.
- Monte Carlo Methods in Financial Engineering (Paul Glasserman, 2004) foundational - Ch. 3 - Simulating price paths with impact to evaluate execution schedules by Monte Carlo.