Practice Test 2

In one sentence, what fundamentally distinguishes quantitative trading from traditional discretionary trading?

• It always uses faster computers and co-located servers
• It replaces intuition and news with data, probability and systematic testing
• It guarantees a profit on every single trade
• It only trades equities and never other asset classes

The five stages of the quant research workflow are best described as:

• Buy, hold, sell, repeat, scale
• Hypothesis, lawyer review, compliance, deploy, retire
• Collect data, analyze patterns, build signals, backtest, deploy and monitor
• Predict, gamble, hope, cash out, restart

"Alpha" in quant trading most precisely refers to:

• A statistical edge that produces returns beyond the benchmark
• The first trade placed each morning
• The volatility of the overall market
• The fee a broker charges per order

In an OHLCV bar, the letter V stands for:

• Volatility over the bar
• Value-at-risk for the bar
• VWAP of the bar
• Volume traded during the bar

The bid-ask spread represents:

• A guaranteed arbitrage profit
• A transaction cost equal to ask minus bid
• The daily price range of the asset
• The dividend paid to shareholders

Why do quants usually analyze returns instead of raw prices?

• Returns are always positive, which simplifies math
• Prices cannot be stored in a database
• Returns normalize movement and allow fair comparison across assets
• Returns remove the need for any backtesting

Volatility, as used in this course, is measured as:

• The standard deviation of returns
• The average of the closing prices
• The total traded volume
• The highest price minus the lowest price ever seen

A simple moving average is primarily used to:

• Guarantee future prices
• Increase the volatility of a series
• Remove all real information from prices
• Smooth noisy prices to reveal the underlying trend

The classic momentum trading rule is:

• BUY when volume is below average
• BUY when the short-term MA rises above the long-term MA
• SELL whenever the price goes up
• BUY only on Mondays

Mean reversion strategies are built on the assumption that:

• Trends always continue indefinitely
• Prices are completely random and unpredictable
• Extreme moves tend to revert back toward a normal level
• Volume always leads price by exactly one day

Which bias is generally considered the most dangerous in backtesting?

• Look-ahead bias — using information not available at the time
• Rounding prices to two decimals
• Using a font that is too small in the report
• Trading too few assets

Survivorship bias occurs when a study:

• Includes too many losing trades
• Only examines assets or funds that survived, ignoring those that failed
• Uses log returns instead of simple returns
• Runs the backtest on a single day of data

Net return is correctly computed as:

• Gross return plus transaction costs plus slippage
• Gross return times slippage
• Slippage minus gross return
• Gross return minus transaction costs minus slippage

A z-score of an observation measures:

• The total traded volume in dollars
• The probability the trade wins
• How many standard deviations the value sits from its mean
• The dividend yield of the asset

In a typical mean-reversion z-score strategy, you would BUY when:

• $z < -2$ (the asset is unusually cheap)
• $z > +2$ (the asset is unusually expensive)
• $z = 0$ exactly
• $z$ is any positive number

The Sharpe ratio measures:

• Total dollars earned over the year
• Excess return earned per unit of risk
• The number of trades placed per day
• The bid-ask spread of the portfolio

Maximum drawdown describes:

• The largest single winning trade
• The average daily return
• The largest peak-to-trough decline in equity
• The number of consecutive winning days

Diversification reduces risk the most when the combined assets are:

• Perfectly positively correlated (corr = 1)
• Identical copies of one another
• All in the same sector
• Lowly or negatively correlated

A market (versus limit) order prioritizes:

• Execution speed — it crosses the spread to trade now
• Price — it rests in the book until matched
• Avoiding all transaction costs
• Trading only at the daily close

Statistical arbitrage differs from pure arbitrage because it is:

• Completely risk-free and instantaneous
• Only legal in certain countries
• Probabilistic and driven by statistics rather than guaranteed
• Impossible to backtest

In pairs trading, the trader is fundamentally betting on:

• The absolute price level of one stock
• The relationship (spread) between two related instruments
• The total market index direction only
• A single stock's dividend schedule

Overfitting in a trading model is best described as:

• Using too little historical data
• Trading too few instruments
• A model that is too simple to capture anything
• Memorizing noise — great on training data, poor out-of-sample

Which is a hallmark of a good feature for an ML trading model?

• Relevant, stable and intuitive
• Random and frequently changing
• Contains future information
• Impossible to explain to anyone

Walk-forward validation is preferred over a single train/test split because it:

• Removes the need to gather any data
• Always produces higher backtest returns
• Repeatedly retrains and tests on rolling windows, mimicking live use
• Lets you legally use future prices

Tick data differs from OHLC data in that tick data:

• Is a once-a-day summary of prices
• Records every individual trade or quote event
• Contains only the daily closing price
• Cannot be stored in a database

kdb+/q is favoured for tick storage primarily because it is:

• A spreadsheet program like Excel
• A relational database optimized for images
• A blockchain ledger
• A columnar time-series database built for speed

Hit rate (win rate) of a strategy is defined as:

• Winning trades divided by total trades
• Total profit divided by total loss
• Average win minus average loss
• Number of trades per day

"Signal decay" refers to the phenomenon where:

• A signal's code is deleted by accident
• Volatility falls to zero permanently
• An alpha weakens over time as markets adapt and competitors arbitrage it
• The data feed is delayed by one second

The Information Coefficient (IC) of a signal measures:

• The total dollars earned
• The latency of the trading system
• The number of features used
• The correlation between predicted and realized returns

Why is "a good backtest" not the same as "a good strategy"?

• Because backtests are illegal in most markets
• Because biases, overfitting and regime change can inflate historical results that fail live
• Because live trading always beats backtests
• Because backtests cannot use price data

A researcher finds a feature that scores beautifully in-sample but its predictive power vanishes the moment new data arrives. The single most likely diagnosis is:

• The market is broken and should be avoided
• The feature is too simple and needs more parameters
• Overfitting — the model learned noise specific to the training set
• The data feed has the wrong timezone

You compute today's signal using today's closing price and then trade on today's close in your backtest. What problem have you introduced?

• Look-ahead bias — you used information not available before the trade
• Survivorship bias
• Excessive diversification
• Nothing; this is the correct approach

Two strategies each return 12% annually. Strategy X has volatility 8%; Strategy Y has volatility 20%. With a risk-free rate of 0%, which statement is correct?

• Y is better because higher volatility means higher potential
• They are equally good because returns are equal
• Neither can be compared without the number of trades
• X is better — its Sharpe ratio of 1.5 beats Y's 0.6

A signal has a 75% hit rate but loses money overall. The most plausible explanation is:

• Hit rate over 50% always means profit, so the data is wrong
• The few losing trades are far larger than the many small winners
• The strategy traded too few times to matter
• Volatility was negative

To convert raw price and volume data into predictive inputs for a model, a quant would engineer features such as:

• The company's CEO name and logo color
• Tomorrow's closing price
• Returns, rolling volatility and the volume ratio $V_t/\text{avg}(V)$
• The exchange's street address

A backtest of a "buy current S&P 500 members over the last 20 years" strategy looks spectacular. The hidden flaw is:

• Survivorship bias — today's members exclude companies that were dropped or went bankrupt
• Look-ahead bias in the closing price
• The data set is too large for a database
• There is no flaw; current members are representative

A pair of stocks has been 0.95 correlated for years, yet a long-short spread trade on them keeps losing as the spread drifts ever wider. The most likely issue is:

• Correlation is too high to trade
• The stocks should be bought, not shorted
• Volume was ignored
• The pair is correlated but not cointegrated, so the spread is not mean-reverting

A strategy earns 9% gross. Round-trip transaction costs total 2% and slippage costs another 1.5% over the period. The net return is:

• 12.5%
• 5.5%
• 9%
• 7.5%

Which experimental setup best guards against fooling yourself when evaluating a new signal?

• Optimize all parameters on the full data set, then report that result
• Pick the single best in-sample year and extrapolate it
• Hold out unseen data and validate with out-of-sample / walk-forward testing
• Trust a smooth equity curve as proof of robustness

An equities desk sees the 10-day MA cross above the 50-day MA on heavy volume. Which signal family does this combination belong to?

• Momentum / trend-following with volume confirmation
• Mean reversion
• Pure arbitrage
• Dividend capture

A signal's edge has steadily shrunk from a Sharpe of 2.0 three years ago to 0.4 today, with no code changes. The best interpretation is:

• The backtest engine is broken
• The risk-free rate must have gone negative
• Diversification increased on its own
• Signal decay — the alpha is being competed away as the market adapts

A model uses a 20-day rolling average of returns as a feature, but the window is computed including the current and future days by mistake. This is an example of:

• Survivorship bias
• Data leakage — future information has entered the feature
• Diversification
• Healthy feature engineering

You combine a momentum strategy, a mean-reversion strategy and a pairs strategy into one book. The main benefit you expect is:

• Guaranteed elimination of all losses
• A higher transaction cost, which is good
• Diversification across uncorrelated alpha sources, smoothing returns
• A perfectly straight equity line every day

A volume spike is best detected by computing which feature and comparing it to a threshold?

• Volume ratio $V_t / \text{avg}(V)$ over a trailing window
• The closing price minus the opening price
• The log of the bid-ask spread
• The dividend yield

Which scenario most clearly indicates a strategy has been over-optimized rather than genuinely robust?

• It uses a single moving-average length chosen on economic reasoning
• It performs similarly across many parameter settings
• It holds up on out-of-sample data and in multiple regimes
• It has 30 hand-tuned rules and an almost perfectly smooth equity curve in-sample

A stock moves from \$80 to \$92 in one day. What is the simple return?

• 12.0%
• 15.0%
• 13.0%
• 8.7%

A price goes from \$50 to \$60. What is the log return $\ln(P_t/P_{t-1})$? (use $\ln 1.2 \approx 0.1823$)

• 20.0%
• 16.7%
• 18.23%
• 10.0%

Daily returns over two days are $+10\%$ then $-10\%$. What is the cumulative (compounded) return $\prod(1+r)-1$?

• −1.0%
• 0.0%
• +1.0%
• −2.0%

Closing prices for 5 days are 20, 22, 24, 26, 28. What is the 5-day simple moving average?

• 26
• 25
• 22
• 24

Prices are 10, 12, 11, 13, 15, 14. Using a 3-day SMA, what is the most recent SMA value (last three prices)?

• 13.0
• 14.0
• 12.0
• 15.0

A strategy made 40 trades; 26 were winners. What is the hit rate?

• 74%
• 26%
• 65%
• 60%

Five trades returned $+2\%, -1\%, +3\%, -2\%, +3\%$. What is the average return per trade?

• +1.0%
• +1.8%
• +5.0%
• +0.5%

Today the 10-day MA is 105 and the 30-day MA is 110. Yesterday the 10-day MA was 108 and the 30-day MA was 109. Under a momentum MA-crossover rule (BUY when short > long, SELL when short < long), what signal fires today?

• BUY, because the short MA rose above the long MA
• HOLD, no change occurred
• BUY, because both MAs are above 100
• SELL, because the short MA is now below the long MA

A spread has mean $\mu = 50$ and standard deviation $\sigma = 4$. The current value is 60. What is the z-score, and under the rule $z>2 \Rightarrow$ SELL, $z<-2 \Rightarrow$ BUY, what action?

• $z = 2.0$, BUY
• $z = 2.5$, SELL
• $z = 2.5$, BUY
• $z = 10$, HOLD

Today's volume is 4,500,000 shares. The trailing 20-day average volume is 1,800,000 shares. What is the volume ratio, and does it exceed a spike threshold of 2.0?

• 2.0, exactly at threshold (no spike)
• 1.8, below threshold (no spike)
• 2.5, spike detected
• 0.4, no spike

A portfolio returns 14% with volatility 16%. The risk-free rate is 2%. What is the Sharpe ratio?

• 0.75
• 0.875
• 0.125
• 1.00

An equity curve peaks at \$150,000 and later falls to a trough of \$120,000 before recovering. What is the maximum drawdown?

• −25.0%
• −30.0%
• −15.0%
• −20.0%

Three trades execute at prices/sizes: 100 @ 200 shares, 102 @ 300 shares, 105 @ 500 shares. What is the VWAP?

• 102.33
• 103.10
• 105.00
• 100.00

Daily returns are $+1\%, +3\%, -1\%, +1\%$ (mean $= +1\%$). Using the sample standard deviation $\sigma=\sqrt{\frac{1}{n-1}\sum (r_i-\bar r)^2}$, what is the rolling volatility?

• 1.41%
• 2.00%
• 1.63%
• 0.80%

A strategy's gross return is 18%. Fees are 3%, slippage is 2%, and borrowing costs are 1%. What is the net return?

• 12%
• 15%
• 13%
• 24%

Over three days an asset returns $+10\%$, then $+20\%$, then $-25\%$. What is the cumulative return $\prod(1+r)-1$?

• +5.0%
• 0.0%
• +1.0%
• −1.0%

A signal generated 8 winning trades averaging $+\$150$ each and 12 losing trades averaging $-\$80$ each. What is the average return (P&L) per trade across all 20 trades?

• +\$35.00
• +\$12.00
• +\$70.00
• −\$12.00

An asset's expected per-trade profit is $\E = p\cdot W - (1-p)\cdot L$ with win probability $p=0.40$, average win $W=\$500$, average loss $L=\$150$. What is the expected value per trade?

• +\$200
• +\$350
• +\$110
• −\$90

Prices over four days are 100, 110, 99, 108. Using a z-score on the daily simple returns with mean $\bar r = +3\%$ and $\sigma = 9\%$, the most recent day's return is $(108-99)/99 \approx +9.09\%$. What is the z-score of that latest return?

• ≈ +0.68
• ≈ +1.35
• ≈ +3.00
• ≈ −0.68

A strategy with starting capital \$100,000 ends the year at \$118,000 after total costs. It took on annual volatility of 12% and the risk-free rate is 4%. What is its Sharpe ratio (using the net annual return)?

• 1.50
• 1.83
• 0.50
• 1.17