2.3 Execution Algorithms: VWAP, TWAP, POV, Implementation Shortfall

Basic Prerequisites: Market Microstructure & Order Types and Calculus & Statistics. (these are the folder-level prerequisites for pages 02–06; page 01 states its own, smaller, entry requirements)

1. Intuition & Practical Objective

An institution that decides to buy $50,000,000 of a mid-cap stock holds a parent order: a block far larger than anything the market would accept at one price. Sweeping it into the book at once detonates the price; trickling it out too slowly leaves it exposed to drift. Execution algorithms chop the parent into child orders distributed over time, each one tiny relative to prevailing liquidity, so that the realized average price is close to a chosen benchmark - the volume-weighted average price (VWAP), the time-weighted average price (TWAP), a constant participation rate (POV), or the decision-time arrival price (implementation shortfall).

This folder is the execution-algorithms topic-folder for Pillar 2. It is a hub: it gives you the (a) fast algorithm-comparison table below (job #1), the (b) core benchmark formulas, and (c) routes you to six sub-pages that climb from zero-knowledge intuition through the three workhorse schedules, implementation shortfall, scheduling and volume profiles, failure modes / gaming, and the benchmark-aware extensions.

The one-sentence essence. “Execution is the physics of turning a decision (price + size) into a realization; TWAP splits by clock, VWAP by forecast volume, POV by live volume, and implementation shortfall measures the gap between the paper portfolio and the real one as execution cost plus opportunity cost.”

Scope note (vs the sibling folder). This folder is the benchmark-and-schedule view - what algorithm, measured against what. For the optimal schedule under impact and volatility (the calculus), see Optimal Execution & Almgren–Chriss. The two meet where VWAP/TWAP become special cases of an Almgren–Chriss timetable.

2. Mathematical Ground Truth & Derivations

Notation. $B$ = number of time buckets (e.g. $B=13$ half-hours in the US session); $K$ = number of market transactions; $p_k,v_k$ = transaction price and volume; $t=1,\dots ,B$ bucket index; $V$ = parent size; $q_t$ = child size in bucket $t$; ${\varphi }_t$ = forecast fraction of day volume in bucket $t$ (${\sum }_t{\varphi }_t=1$); $\rho$ = participation rate; $m_0,m_t$ = arrival and terminal benchmark prices; $\bar{p}$ = average execution price; $\kappa$ = fraction filled.

Algorithm comparison (job #1 lookup):

VWAP benchmark (Foucault eq 2.7). The market’s day VWAP is the volume-weighted mean of transaction prices,

equivalently over buckets $\text{VWAP}={\sum }_{t=1}^B{\varphi }_t{p}_t$ when ${\varphi }_t$ is realized volume share. Key flaw (Hasbrouck Ch 14; Foucault): VWAP depends on your own realized volume - a broker that handles a large share of the day sets the benchmark and “always wins.” It is gameable (Harris 2003).

VWAP slippage / tracking error. The engine’s target. Its average execution price $\bar{p}={\sum }_t{w}_t^{\text{exec}}{p}_t$ (weights = child shares) beats the benchmark when $\bar{p}<\text{VWAP}$ for a buy. The schedule’s tracking error is the RMS deviation of its child-allocation weights from realized volume shares,

TWAP’s weights are flat, so ${\text{TE}}_{\text{TWAP}}>0$ whenever the profile is U-shaped; a perfect VWAP engine has ${\text{TE}}_{\text{VWAP}}=0$.

Implementation shortfall (Perold 1988; Foucault eq 2.29). With $\kappa$ the fraction of the desired order filled at average price $\bar{p}$, between decision price $m_0$ and terminal price $m_t$:

3. Computational Implementation - the parent-order cost of slicing

Node: sweeping $X=10^6$ shares buys at the entire temporary-impact cost in one shot; TWAP equal-slicing over 50 buckets pays it on slices of $X/50$. Stdlib only; re-executed and reproduced (full scripts on each sub-page).

4. Failure Modes & First-Principles Breakdowns

Hub signposts - the full analysis is in 05 · Failure Modes & Practice. In one line each:

1. Gaming a predictable schedule - a deterministic TWAP heartbeat (or a VWAP engine that leaks its profile) can be front-run by snipers who buy just ahead of every child.
2. Volume-profile misestimation - a U-shaped historical profile fails on a CPI/event day; the engine under-trades the open and is forced to dump at the close.
3. Adverse selection by speed - slow, schedule-driven participation pays the laddered price on an informed trend, so “low impact” coexists with “wrong side.”

5. References

• Perold, André F. - “The implementation shortfall: Paper versus reality,” Journal of Portfolio Management 14(3), 4–9 (1988). *The origin of the IS benchmark
• Hasbrouck, Joel - Empirical Market Microstructure (2007)
• Foucault, Pagano, Roëll - Market Liquidity (2013)
• Kissell, Glantz & Malamut - Optimal Trading Strategies (2003) and Kissell - The Science of Algorithmic Trading (2014). The desk standard on building/slicing around VWAP/TWAP/POV/IS.
• Johnson, Barry - Algorithmic Trading & DMA (2010). INT
• Almgren & Chriss - “Optimal execution of portfolio transactions,” Journal of Risk 3(2), 5–40 (2000). The scheduling calculus underlying arrival-price (IS) execution.

6. Connected Graph Bridges

• Foundational base: Market Microstructure & Order Types · Market Impact & Depth
• Sibling topic: Optimal Execution & Almgren–Chriss (the optimal schedule; this folder’s IS view is its measurement side)
• Sibling topic: Queue Position & Fill Probability (how a child order actually gets filled)
• Transaction costs & turnover: Constraints & Transaction Costs
• Sub-pages (in-folder): 01 From Zero · 02 TWAP·VWAP·POV · 03 Implementation Shortfall · 04 Scheduling & Volume Profiles · 05 Failure Modes & Practice · 06 Advanced Extensions

Beginner: start at 01 · Practitioner: start at 05