Execution and Slippage for Quant Interviews: Market Impact, Algorithms, Cost of Trading

Execution and Slippage for Quant Interviews: Market Impact, Algorithms, and the Cost of Trading

Execution is where strategy meets reality. A profitable signal on paper can be unprofitable in practice if execution is bad enough; a mediocre signal with excellent execution can outperform a strong signal traded poorly. For quant trading interviews at hedge funds and prop shops with serious execution programs — Two Sigma, Citadel, Millennium pods, Tower, Bridgewater, plus the algorithmic execution teams at Goldman, JPMorgan, Morgan Stanley — understanding execution mechanics is essential. This guide covers the slippage decomposition, common execution algorithms, market impact models, and the practical trade-offs that come up in interviews.

The Slippage Decomposition

Slippage is the difference between the “paper” price (what your strategy assumed) and the price actually achieved. It decomposes into several components:

Spread cost

The bid-ask spread. If you cross the spread (buy at the offer, sell at the bid), you pay half the spread per side as a baseline cost. Tighter spreads in liquid stocks mean lower spread cost; wider spreads in illiquid stocks mean higher.

Market impact

Your trade moves the price against you. Buying pushes the price up; selling pushes it down. Impact has two flavors:

Temporary: price snaps back after your trade because the move was driven by liquidity demand, not new information.
Permanent: price stays moved because the market updated its estimate of fair value based on your trade.

Strong candidates discuss this distinction. Temporary impact is partially recoverable; permanent impact is paid permanently.

Timing risk / drift

The price moves while you’re trying to execute, possibly against you. If you take 30 minutes to fill a large order, the price might drift 50bps in that time independent of your execution. Splitting orders across time reduces impact but increases timing risk.

Opportunity cost

If your execution is slow enough that the strategy decays before you finish, you’ve missed alpha. Time-sensitive signals decay quickly; execution must respect this.

Fees and rebates

Exchange fees, regulatory fees, broker commissions. Maker rebates can offset some costs (paying you to provide liquidity). The net cost of trading varies by venue, order type, and fee structure.

The Almgren-Chriss Framework

The classical framework for optimal execution. Trade off impact (which falls with slower execution) against timing risk (which rises with slower execution).

Setup: you need to liquidate Q shares over time T. You choose a trading schedule x(t) with x(0) = Q and x(T) = 0. The cost has two components:

Impact cost: proportional to ∫(dx/dt)² dt (faster trading = quadratically more impact)
Variance cost: proportional to ∫x(t)² dt (more inventory held = more risk)

Optimal solution: an exponentially-decaying schedule that trades fast initially when impact is dominated by inventory risk and slows as inventory falls. The “Almgren-Chriss” trajectory is the optimal balance.

For interviews, you don’t need to derive the solution; you need to understand the trade-off and discuss when faster vs slower execution makes sense.

Common Execution Algorithms

VWAP

Volume-Weighted Average Price. Schedule trading in proportion to historical intraday volume profile. Goal: achieve average execution price close to the day’s VWAP. Common benchmark; widely understood. Weak when volume profile differs from historical average (e.g., on event days).

TWAP

Time-Weighted Average Price. Schedule evenly over time. Simple; suitable when volume profile is uncertain. Lacks the volume-following intelligence of VWAP.

Implementation Shortfall (IS)

Minimize the gap between decision price (when the strategy fired) and actual execution. Captures the timing-risk vs impact trade-off explicitly. Almgren-Chriss is an IS-style algorithm.

Participation rate (POV)

Trade a fixed percentage of market volume (e.g., “10% POV” trades 10% of every executed share). Adapts to market conditions; doesn’t commit to a schedule.

Aggressive (liquidity-taking)

Cross the spread to fill immediately. High impact but low timing risk. Suitable when alpha is decaying fast.

Passive (liquidity-providing)

Post limit orders, wait for fills. Low spread cost but high adverse-selection risk: the orders that get filled are disproportionately adverse to you. Suitable when alpha is stable and you can wait.

Market Impact Models

Linear (Kyle’s lambda)

Impact proportional to volume traded: Δprice ∝ Q × λ. The simplest model; useful as a first-order approximation. Lambda is calibrated empirically per stock.

Square-root law

Empirically observed impact is closer to Δprice ∝ √(Q / V) where V is daily volume. Larger trades have more impact, but the relationship is concave: doubling the trade size doesn’t double impact. The square-root law has been re-confirmed in many empirical studies.

Almgren-Chriss propagator models

Distinguish temporary and permanent impact explicitly. Temporary impact decays after the trade; permanent impact persists. More complex to calibrate but more realistic for short-horizon trading.

Practical Considerations

Liquidity and market regime

Impact depends on liquidity: trading 10% of daily volume has much higher impact in a thinly-traded small-cap than in a liquid mega-cap. Real-time liquidity (current depth, recent trade flow) matters too. Algorithms that adapt to instantaneous liquidity outperform those using static parameters.

Time of day

Volume profiles are U-shaped: high at open and close, lower in the middle. VWAP-style algorithms trade more at open/close to follow this profile.

News and events

Earnings announcements, Fed decisions, and other events spike volatility and reduce liquidity. Algorithms typically pause around scheduled events; unscheduled events require real-time response.

Cross-impact

Trading one stock can affect related stocks (sector peers, ETF constituents, paired stocks). Sophisticated strategies account for this; naive ones don’t and pay for it in tracking error.

Common Interview Questions

Decompose slippage

“You executed an order and slipped 30bps from decision price. How do you decompose the slippage?” Spread cost (~5bps for liquid stocks); market impact; timing drift; fees. Strong candidates list components and discuss diagnostics: what would help isolate impact from drift?

Optimal execution trade-off

“You need to buy 1M shares of stock X today. You can finish in 10 minutes or 6 hours. Which is better?” Depends on: alpha decay (fast-decaying signal favors fast execution); volatility (high vol favors fast execution to reduce timing risk); average daily volume (low ADV favors slow execution to reduce impact). Strong candidates probe these and discuss the Almgren-Chriss trade-off.

Aggressive vs passive

“When would you use a market order vs a limit order?” Market orders for urgency, certainty, low-impact contexts. Limit orders for cost reduction when you can wait. Discuss adverse selection: passive limit orders fill disproportionately when the price moves against you.

VWAP execution

“How would you implement a VWAP algorithm?” Estimate intraday volume profile; spread your order proportionally; place child orders aggressively or passively depending on tracking error. Strong candidates discuss real challenges: profile estimation, regime shifts, tracking VWAP slippage.

Discuss square-root law

“Why does the square-root law hold empirically?” Several theoretical explanations: the assumption that participants reveal their order size gradually; the structure of order books; the equilibrium between liquidity providers and takers. Empirical robustness has more support than theoretical understanding; strong candidates can discuss the empirical evidence and the limits of theoretical explanations.

Frequently Asked Questions

How important is execution research vs alpha research?

Both matter. Execution research often has lower variance and higher capacity than alpha research: improvements in execution apply to all of a fund’s trading and don’t decay the way novel alpha does. Alpha research has higher upside per discovery but more variance. At large hedge funds, execution research teams are substantial and well-paid; some firms (Two Sigma, Citadel) treat them as central rather than support functions. Don’t dismiss execution roles as second-tier; they’re often more impactful than they look.

What books should I use for execution prep?

Robert Almgren and Neil Chriss’s papers on optimal execution are foundational. Larry Harris’s Trading and Exchanges covers execution in the broader microstructure context. Algorithmic Trading: Winning Strategies and Their Rationale by Ernie Chan covers execution from a practitioner’s view. Market Microstructure in Practice by Lehalle and Laruelle is a deeper academic treatment. For interviews, Almgren-Chriss plus Harris on microstructure is sufficient.

How do execution algorithms handle very illiquid stocks?

Carefully. In illiquid stocks, even small orders have material impact; standard algorithms (VWAP, TWAP) over-trade and pay too much. Specialized approaches: trade only when sufficient liquidity exists (wait-and-see); use larger lot sizes timed to natural liquidity events (opening cross, large-trade auction sessions); limit participation rate strictly. Some algorithms switch to non-anonymous channels (block trading desks, dark pools) where size can be transferred without on-exchange impact.

Is dark-pool execution always cheaper?

Not always. Dark pools (off-exchange venues without displayed quotes) can reduce impact for large orders by hiding intent, but they have their own costs: adverse-selection risk (informed traders may use dark pools strategically), potentially worse fill rates, fees that may be higher than displayed venues. Sophisticated execution strategies route across both lit and dark venues based on real-time conditions. Strong candidates discuss when dark pools help and when they don’t.

What’s the relationship between execution and HFT?

Some HFT activity is execution-related (executing internal orders fast and at low impact); some HFT activity competes with execution algorithms (taking liquidity that execution algos try to capture). The relationship is complex: execution algorithms try to outsmart HFT (avoid being detected and front-run); HFT firms try to detect execution patterns and trade ahead of them. The “arms race” between them is part of why execution costs have evolved over the past 20 years.