Algorithmic & Quant Trading
Level: 1 (Recognition) Time Limit: 20 minutes Total Marks: 30
Section A — Multiple Choice (1 mark each) — 10 marks
Q1. Algorithmic trading is best defined as:
- A) Trading only during market opening hours
- B) Using computer programs to execute orders based on predefined rules
- C) Predicting prices with astrology
- D) Manually placing trades faster than others
Q2. Which of the following is NOT a core component of a trading system?
- A) Signal generation
- B) Risk management
- C) Order execution
- D) A lucky charm
Q3. In pairs trading, the two selected securities should ideally be:
- A) Completely uncorrelated
- B) Cointegrated
- C) From different countries only
- D) Negatively priced
Q4. A mean-reversion strategy assumes that prices tend to:
- A) Trend upward forever
- B) Return toward an average level
- C) Move randomly with no pattern
- D) Only fall during recessions
Q5. A momentum strategy typically buys assets that have recently:
- A) Fallen sharply
- B) Risen in price
- C) Paid dividends
- D) Been delisted
Q6. Overfitting in strategy design means the model:
- A) Generalises well to new data
- B) Fits historical noise and fails on unseen data
- C) Uses too few parameters
- D) Ignores the training data
Q7. Walk-forward analysis is mainly used to:
- A) Increase leverage
- B) Test a strategy on rolling out-of-sample periods
- C) Guarantee profits
- D) Eliminate transaction costs
Q8. Statistical arbitrage relies primarily on:
- A) Insider tips
- B) Statistical relationships and mean-reverting spreads
- C) Single-stock long-term investing
- D) Fixed interest coupons
Q9. A key caution when using machine learning in trading is:
- A) It always beats simple models
- B) The risk of overfitting to limited noisy data
- C) It requires no data
- D) It removes the need for risk management
Q10. In cointegration, the spread between two cointegrated series is expected to be:
- A) Explosive
- B) Stationary (mean-reverting)
- C) Always increasing
- D) Random with a trend
Section B — Matching (1 mark each) — 8 marks
Q11–Q18. Match each term in Column X to its correct description in Column Y.
| Column X | Column Y |
|---|---|
| Q11. Signal generation | A. Splitting data into rolling in-sample/out-of-sample windows |
| Q12. Backtesting | B. Rules that decide when to enter/exit trades |
| Q13. Curve fitting | C. Testing a strategy on historical data |
| Q14. Walk-forward analysis | D. Long the underperformer, short the outperformer of a pair |
| Q15. Pairs trade | E. Excessively tuning parameters to past data |
| Q16. Momentum | F. Buying recent winners, selling recent losers |
| Q17. Risk management | G. Controlling position size and loss limits |
| Q18. Cointegration | H. A stable long-run equilibrium relationship between series |
(Write answers as Q11–…, Q12–…, etc.)
Section C — True/False WITH Justification (2 marks each: 1 T/F + 1 justification) — 12 marks
Q19. "Two highly correlated stocks are always cointegrated." True or False? Justify.
Q20. "A higher in-sample backtest return always means better live performance." True or False? Justify.
Q21. "Walk-forward analysis helps reduce the risk of overfitting." True or False? Justify.
Q22. "In a mean-reversion model, a very high z-score of the spread signals a potential entry." True or False? Justify.
Q23. "Adding more parameters to a strategy always improves robustness." True or False? Justify.
Q24. "Momentum and mean-reversion strategies rely on opposite assumptions about price behaviour." True or False? Justify.
Answer keyMark scheme & solutions
Section A (1 mark each)
Q1 — B. Algorithmic trading = automated execution via predefined rules. (1 mark)
Q2 — D. A lucky charm is not a system component; the rest are core modules. (1 mark)
Q3 — B. Pairs trading needs a stable statistical link — cointegration — not mere correlation. (1 mark)
Q4 — B. Mean reversion assumes prices revert to an average. (1 mark)
Q5 — B. Momentum buys recent winners (rising prices). (1 mark)
Q6 — B. Overfitting = fitting noise; poor out-of-sample performance. (1 mark)
Q7 — B. Walk-forward tests on rolling out-of-sample windows. (1 mark)
Q8 — B. Stat-arb exploits statistical relationships / mean-reverting spreads. (1 mark)
Q9 — B. ML caution = overfitting noisy, limited financial data. (1 mark)
Q10 — B. A cointegrated spread is stationary/mean-reverting. (1 mark)
Section B (1 mark each)
| Q | Answer |
|---|---|
| Q11 | B |
| Q12 | C |
| Q13 | E |
| Q14 | A |
| Q15 | D |
| Q16 | F |
| Q17 | G |
| Q18 | H |
Why: Each term maps to its defining function — signals=rules, backtest=historical test, curve fit=over-tuning, walk-forward=rolling windows, pairs=long/short pair, momentum=winners, risk mgmt=sizing/limits, cointegration=long-run equilibrium.
Section C (2 marks each: 1 for T/F, 1 for justification)
Q19 — FALSE. (1) Correlation measures short-term co-movement, but cointegration requires a stationary linear combination (stable long-run equilibrium). Correlated series can still drift apart, so correlation ≠ cointegration. (1)
Q20 — FALSE. (1) High in-sample return may reflect overfitting to past noise; live/out-of-sample results are what matter. (1)
Q21 — TRUE. (1) By repeatedly optimising in-sample then validating out-of-sample, walk-forward exposes strategies that only fit historical noise, reducing overfitting risk. (1)
Q22 — TRUE. (1) A large |z-score| means the spread is far from its mean; a mean-reversion model bets it reverts, so it triggers an entry (fade the deviation). (1)
Q23 — FALSE. (1) More parameters increase flexibility and the chance of curve-fitting noise, typically reducing robustness, not improving it. (1)
Q24 — TRUE. (1) Momentum assumes trends persist (winners keep winning); mean-reversion assumes deviations reverse — opposite behavioural assumptions. (1)
[
{"claim":"Section A has 10 one-mark questions summing to 10","code":"result = (10*1 == 10)"},
{"claim":"Section B has 8 one-mark matching items summing to 8","code":"result = (8*1 == 8)"},
{"claim":"Section C has 6 questions at 2 marks each summing to 12","code":"result = (6*2 == 12)"},
{"claim":"Total marks equal 30","code":"secA=10; secB=8; secC=6*2; result = (secA+secB+secC == 30)"},
{"claim":"Total question count is between 15 and 20 inclusive","code":"n=10+8+6; result = (15 <= n <= 20)"}
]