Every quant recruiting cycle, interviewers ask questions that sound almost too simple. Candidates from top programs (MIT, Princeton, Stanford) hear them and think there must be a trick. There is, but it's not the kind they expect. The trick is that the answer requires genuine understanding, not pattern matching.
Here are four real questions that tripped up more candidates than you'd think.
Question 1: The 10 Strategy Backtest
You generate 10 arbitrary trading strategies. 5 of them perform well on your test set. You then run those 5 on a validation set, and 2 perform well. Should you put those 2 into production?
Why People Get It Wrong
Most candidates say yes. They reason that you tested on out-of-sample data, 2 strategies survived, so they must be real. Some even cite the train-test-validation split as textbook best practice.
The Answer
No. The problem is multiple hypothesis testing. You started with 10 arbitrary strategies, meaning they have no economic rationale behind them. By pure chance, some will look good on any finite dataset.
If each random strategy has a 50% chance of looking good on the test set, you'd expect ~5 to pass. If each of those has a 40% chance of passing validation, you'd expect ~2 to survive. The 2 survivors aren't validated. They're the expected number of false positives from running 10 random experiments.
The key word in the question is arbitrary. If the strategies were derived from some economic hypothesis, the calculus changes. But random strategies that survive multiple rounds of testing are almost certainly overfitting noise. This is exactly why funds that mine thousands of signals end up with strategies that blow up in live trading.
Question 2: Apple vs. ExxonMobil
You expect Apple to return 20% over the next year and ExxonMobil to return 19%. Should you trade Apple, ExxonMobil, or some combination?
Why People Get It Wrong
The instinctive answer is Apple. It has the higher expected return with the same risk. Candidates who've taken one finance course will say this confidently.
The Answer
You should trade a combination of both. This is Markowitz 101, but people forget it the moment concrete numbers appear.
Even though Apple has a slightly higher expected return, unless Apple and ExxonMobil are perfectly correlated (they are not; one is tech, the other is energy), a portfolio combining both will have lower variance than either alone while sacrificing very little in expected return.
The math is simple. If both have volatility σ and correlation ρ < 1, a 50/50 portfolio has volatility σ√((1 + ρ)/2), which is strictly less than σ. You give up 0.5% in expected return (going from 20% to 19.5%) but your Sharpe ratio almost certainly improves. The diversification benefit is free. Not taking it is leaving risk-adjusted return on the table.
The deeper lesson: in quantitative finance, you almost never want to make a single concentrated bet. The edge comes from combining imperfectly correlated positions.
Question 3: The RSI Strategy in a Bull Market
The S&P 500 has gone up 20% every year for the last 5 years. Your simple RSI strategy takes a trade on the S&P once every week. It makes good money on longs but not on shorts. Why?
Why People Get It Wrong
Candidates start debugging the RSI parameters. Maybe the lookback period is wrong. Maybe the overbought threshold needs tuning. They try to fix the signal.
The Answer
The strategy is not generating alpha. It is capturing beta. When the S&P goes up 20% per year for 5 consecutive years, any strategy that goes long will make money. Your RSI signal is largely irrelevant. You could flip a coin to decide your longs and still show a profit, because the underlying market drifted up ~100% over the period.
The shorts lose money for the same reason: you are fighting a massive secular trend. No mean-reversion signal will consistently profit on the short side when the market moves relentlessly in one direction.
This is the single most common mistake in retail and junior quant backtesting: confusing market direction with strategy performance. A proper evaluation would benchmark the long P&L against simply being long the S&P. If your RSI longs don't outperform buy-and-hold, your strategy adds nothing. Similarly, the short side should be evaluated against a baseline of always being short, in which case both your strategy and the baseline lose money, confirming there is no edge on either side.
The broader point: always decompose your returns into beta (market exposure) and alpha (genuine edge). A backtest in a trending market will make almost any directional strategy look brilliant.
Question 4: The 50% Drawdown
You have a single pairs trading strategy on Pepsi and Coca-Cola. It has a 50% drawdown but returns 100% overall. Does the drawdown matter if you know the final return is 100%?
Why People Get It Wrong
Candidates go one of two ways. Some say the drawdown is irrelevant because the final return is 100%. Others launch into a long discussion about margin calls, leverage, and risk limits. Both miss the actual point.
The Answer
The drawdown would be irrelevant if this were a pure arbitrage. In a true arbitrage, convergence is guaranteed. You know with certainty that the prices will realign, so any drawdown along the way is just temporary mark-to-market noise. You hold through it and collect your return.
But pairs trading on Pepsi and Coca-Cola is not a pure arbitrage. There is no contractual or structural guarantee that the spread will converge. The two stocks are correlated because they operate in the same industry, but that correlation can break down permanently. Pepsi could acquire a company, change its capital structure, enter a new market, or face a product recall. The historical relationship you are trading on is statistical, not structural.
This means when you are sitting at a 50% drawdown, you have no way of knowing whether the spread will recover. The 100% return you are counting on is a backtest result, not a guaranteed outcome. The drawdown could be the beginning of permanent divergence, not a temporary dip on the way to profit.
That is the entire answer. The drawdown matters because you cannot distinguish between a temporary drawdown in a converging trade and the start of a permanent loss in a diverging one. Only in pure arbitrage, where convergence is certain, can you safely ignore the path and focus on the endpoint.
The Common Thread
All four questions test the same thing: whether you can see past surface-level results to understand what's actually happening. Overfitting, diversification, alpha vs. beta, drawdown risk, and the bias-variance tradeoff are not advanced concepts. They are taught in every introductory quant course. But when presented as simple scenarios instead of textbook formulas, many candidates fail to apply them.
That's exactly why these questions work as interview filters. They separate people who understand quantitative finance from people who memorized it.