The overall experience of this Two Sigma OA is different from what I have done before. BCGXOA Very similar, the questions are not tricky, but the requirements for completeness of ideas and implementation details are relatively high. The time is relatively ample. Under normal circumstances, it is completely feasible to pass all three questions in about 30 minutes, as long as you don't get distracted by the questions at the beginning.
Two Sigma OA timeline reference
It usually takes 3-7 days after submission to receive the OA invitation. If it is faster, it will arrive in two or three days. OA usually gives a completion window of 5-7 days, but the amount of questions is not large and can be completed in about 30 minutes if the thinking is clear. After submitting an OA, the feedback time is usually within 1–2 weeks, ranging from a few days to two weeks.
After passing the OA, the process will advance directly to HR or technical interviews. The overall rhythm is crisp and smooth.
Q1: Linear Interpolator
Given a two-dimensional node, process the same x node according to the rules, and calculate the y value corresponding to the target x through interpolation/extrapolation
Idea: First pair and sort the x and y nodes, filter the corresponding y values according to rules to generate a deduplication list for the same x node, then use binary search to quickly locate adjacent nodes, substitute the linear interpolation formula to calculate the result, and implement the entire process manually without relying on ready-made libraries.
Q2：Daily Temperature By Town
Based on multi-town and NYC temperature data, answer 5 statistics/prediction questions + select 5 towns with excellent joint predictions
Idea: Part 1 relies on standard deviation, conditional screening to find the median, and single/double town linear regression to find the smallest error. Part 2 uses a greedy algorithm to iteratively select towns with the smallest error, and finally determines 5 towns with excellent joint predictability.
Q3: No-intercept OLS regression coefficient calculation
The third question is relatively speaking the question with the purest formula. The question gives two sets of asset return data and requires the calculation of OLS regression coefficients without intercept.
The core is the standard formula of intercept-free OLS. As long as you understand what the numerator and denominator correspond to, the code implementation is actually very straightforward. When there is a single set of data, I directly use numpy's vectorization operation to calculate the dot product and sum of squares; if there are multiple sets of data, I first calculate the required cumulative terms and then uniformly substitute them into the formula, which not only ensures efficiency but also avoids repeated calculations.
Compared with the first two questions, this question is more like confirming whether you really understand basic statistical modeling, rather than just "tune the library".
A little extra about OA preparation
If you are preparing for OA from Two Sigma or other quantitative/data companies, we also provide it here OA ancillary services , at the same time, it has compiled and updated a large number of real OA questions and solution ideas for a long time, covering common high-frequency test points, helping you to solve complex questions steadily in a limited time. For OA with many rules, heavy details, and low fault tolerance, if someone is there to accompany you, the overall pass rate will usually be much higher.
