# Market Microstructure and Execution Analysis Explain how venue mechanics affect observed prices and fills. Use this workbook alongside the course. Write your answers before opening the solutions. Practical work is self-reviewed; scored knowledge checks are in the Academy. ## 1. Market mechanics ### Limit order books A market consists of trading arrangements, participants and rules, not only a price chart. Identify whether the observed book belongs to an exchange, dealer or provider. Its displayed liquidity is not automatically the liquidity of the entire gold market. ### OTC quoting An order book organizes eligible buying and selling interest under venue rules. Price priority is common, but time, allocation and special-order behavior can vary. Inspect the applicable mechanism before assuming how queue position translates into fills. ### Aggressors and passive orders Displayed quantity can change before an order reaches it. Some interest is not displayed, and cancellations can remove visible interest. Treat a book snapshot as an observation at a time, not a firm promise of future execution. ### Venue fragmentation A transaction has counterparties and an execution price, but the observer may not know their motives. Avoid assigning intention such as institutional accumulation solely from a large print; size and direction do not identify the underlying purpose. ### Worked example A snapshot shows ten units at a price, but those units are cancelled before a new order arrives. A model that fills against the old snapshot assumes liquidity that was no longer available. ### Independent exercise Specify the timestamps and event ordering needed to evaluate that order without using stale book state. My inputs and assumptions: My calculation or decision: Evidence that would change my conclusion: ## 2. Order behavior ### Queue priority A marketable order consumes eligible opposing liquidity under its terms. The average fill can worsen as size reaches additional price levels. Measure the quantity-weighted execution price rather than using only the best quote. ### Partial fills A resting limit order can wait behind other eligible interest. A traded price at its limit does not prove its queue was reached. Without order-level evidence, a fill model should state how queue uncertainty is handled. ### Adverse selection Cancellation competes with market events. A fill can occur before a cancel is effective, leaving real exposure despite the intention to withdraw. Reconcile acknowledgement and execution records rather than assuming local request order equals venue event order. ### Stop execution Adverse selection occurs when fills tend to arrive before unfavorable subsequent movement. A passive price advantage can be offset by which orders actually fill. Compare filled and unfilled opportunities without conditioning only on favorable outcomes. ### Worked example A buy for three units fills one at 100 and two at 101. Its weighted average is 100.67, approximately. Reporting the best quote of 100 as the fill understates the cost of size. ### Independent exercise Calculate the average for two units at 99 and three at 100. Explain what additional information is required to compare this with a benchmark. My inputs and assumptions: My calculation or decision: Evidence that would change my conclusion: ## 3. Measurement ### Arrival price Choose an execution benchmark before judging the result. Arrival price, decision price and a period average answer different questions. Selecting whichever benchmark makes a fill look best creates an outcome-dependent measurement. ### Slippage decomposition For a buy, a fill above the benchmark is an adverse price difference; for a sell, reverse the sign. Multiply by quantity and instrument multiplier to express cash cost. Keep fees separate if they are not embedded in the price comparison. ### Latency Separate observed facts from attribution. Total implementation shortfall can include delay, market movement and execution effects. A simple benchmark difference does not uniquely identify which component caused it. ### Execution benchmark selection Report distributions and conditions alongside averages. Size, volatility, spread and time of day can change execution quality. A mean from heterogeneous orders may conceal the subgroup in which a strategy becomes infeasible. ### Worked example A buy fills at 101 against a preselected benchmark of 100 for two contracts with multiplier ten. The adverse price difference is 20 before fees. Calling the difference one dollar ignores the exposure multiplier and size. ### Independent exercise Repeat for a sell at 99 against benchmark 100 with the same size and multiplier. State whether the difference is favorable or adverse. My inputs and assumptions: My calculation or decision: Evidence that would change my conclusion: ## 4. Execution experiment design ### Choose a benchmark before routing Design the experiment around a defined execution question and benchmark. Record all eligible orders, including rejections and non-fills. Restricting the sample to completed favorable fills can misrepresent the implementation problem. ### Separate market movement from execution cost Market movement during delay should not be silently attributed entirely to spread or commissions. Preserve decision, submission, acknowledgement and fill timestamps where available so different components can be investigated. ### Control for order size and volatility Control or stratify by size and volatility when comparing methods. Two routing policies used in very different conditions cannot be compared fairly by raw average price difference alone. ### Report uncertain or unobservable fills Some counterfactual fills are unobservable: you do not know exactly what would have happened to an order never sent. State the modeling assumptions and uncertainty rather than presenting a simulated alternative as an actual execution record. ### Worked example Policy A is used only for small quiet-market orders and B for large volatile-market orders. A lower average shortfall under A does not isolate a policy advantage. ### Independent exercise Design a more informative comparison and identify a limitation that remains even after matching observed conditions. My inputs and assumptions: My calculation or decision: Evidence that would change my conclusion: ## Course project Compare implementation shortfall under multiple order policies. ### Self-review rubric - Concepts and reasoning: 25% - Calculations, data and evidence: 30% - Process and risk controls: 25% - Limitations and communication: 20% Record one correction and one next practice task. This rubric is not automatically graded. ## Worked solutions ### Exercise 1 Track snapshot time, subsequent book changes, order submission and arrival assumptions. Apply events in the documented order and use the state available when the order could execute, not the most convenient earlier snapshot. ### Exercise 2 The average is (198+300)/5 = 99.6. A cost comparison also needs direction, benchmark price and timestamp, multiplier, fees and the order's intended quantity. ### Exercise 3 The sell receives one less per unit than the benchmark, so the adverse difference is again 20. Direction changes the sign convention; the benchmark must still have been chosen consistently. ### Exercise 4 Use predefined comparable groups or a controlled study where appropriate, retain all outcomes and report size/volatility differences. Unobserved conditions and counterfactual queue states may still limit causal attribution. ## Further reading - https://www.cmegroup.com/education/courses/introduction-to-futures