Test the Hypothesis
Reported by candidates from Point72's online assessment. Pattern, common pitfall, and the honest play if you blank under the timer.
Point72 reported this one in July 2026, and the detail that jumps out is the output: a two-element string array with "Yes" or "No" plus a magnitude to exactly two decimals. It's a two-sample t-test dressed up as a coding question. If you've got an OA invite, expect to implement the stats by hand, not call a library. The math is short but easy to fumble under a timer. If you blank on the formula or the critical value, StealthCoder runs invisibly on your desktop during the live OA and gives you a working solution as a safety net.
The problem
Given two arrays and a confidence level, determine whether the means of these series are significantly different using a two-tailed t-test. Additionally, output the magnitude by which the hypothesis passes or fails. The magnitude is defined as the minimum absolute difference between the computed t-statistic and the t-statistic at the confidence level at the two tails. Return an array with two values. The first element is "Yes" if the means are significantly different, or "No" if they are not. The second element is the magnitude rounded to 2 decimals. For FastPrep's multi-language runner, return both values as strings, with the magnitude formatted to exactly two decimal places. Source note: The screenshot has a tiny rounding/number typo in the samples, so FastPrep uses the mathematically consistent values while keeping the same t-test rules and source image visible below. Function testHypothesis(n: int, x: float[], y: float[], confidence_level: float) → String[] Complete the function testHypothesis with the following parameters: ParameterTypeDescription nintthe number of data points x[n]floatan array of data points y[n]floatanother array of data points confidence_levelfloatthe confidence level to test the hypothesis Examples Example 1 n = 4 x = [4.461, 7.757, 17.317, 4.151] y = [8.911, 12.68, -10.593, 17.048] confidence_level = 0.95 return = ["No", "2.47"] Performing a t-test at the 95% confidence level shows that the means are not significantly different, and the magnitude by which the null hypothesis passes is 2.47. Example 2 n = 10 x = [0.878, 2.817, 0.121, -0.945, -0.055, -1.456, 0.35, -1.478, -1.687, -1.889] y = [1.142, -0.492, -0.988, -0.729, -0.846, -0.157, 0.5, 1.188, -1.075, -0.189] confidence_level = 0.95 return = ["No", "1.82"] Performing a t-test at the 95% confidence level shows that the means are not significantly different, and the magnitude by which the null hypothesis passes is 1.82.
Reported by candidates. Source: FastPrep
Pattern and pitfall
The trick is that it's a math and simulation problem, not a data-structure one. Compute the mean and sample variance of each array, build the t-statistic, then compare its absolute value to the two-tailed critical value for your confidence level. The magnitude is the absolute gap between your t and that critical value. The pitfalls are all in the details. Use sample variance (divide by n-1), not population variance. Decide which t-test variant matches the examples, and check it against both samples before trusting it. Pick the right degrees of freedom. Handle the critical value without a stats library, which probably means a lookup table or an inverse t approximation. Format the magnitude to exactly two decimals as a string. If the formula or critical value logic slips away mid-assessment, StealthCoder is the hedge that reads the problem and hands you a solution while the proctor sees nothing.
If this hits your live OA and you blank, StealthCoder solves it in seconds, invisible to the proctor.
You can drill Test the Hypothesis cold, or you can hedge it. StealthCoder runs invisibly during screen share and surfaces a working solution in under 2 seconds. The proctor sees the IDE. They don't see what's behind it. Built by an Amazon engineer who would have shipped this the night before his JPMorgan OA if he'd had it.
Get StealthCoderRelated leaked OAs
You've seen the question.
Make sure you actually pass Point72's OA.
Point72 reuses patterns across OAs. Built by an Amazon engineer who would have shipped this the night before his JPMorgan OA if he'd had it. Works on HackerRank, CodeSignal, CoderPad, and Karat.
Test the Hypothesis FAQ
What's the trick in the Point72 Test the Hypothesis problem?+
It's a plain two-sample t-test. Compute each mean and sample variance, get the t-statistic, then compare its absolute value to the two-tailed critical value at your confidence level. The magnitude is the absolute difference between the two. No clever algorithm, just careful math.
How do I verify my t-test variant is the right one?+
Run both examples by hand or in code. Example 1 should give No with 2.47, and Example 2 should give No with 1.82. If your variant or degrees of freedom don't reproduce both, switch variants until they do. Two matching samples is strong evidence.
How do I get the critical t-value without a stats library?+
Assume you may not have one. Either hardcode a t-table for common confidence levels and degrees of freedom, or implement an inverse t approximation. Check which the assessment allows. A small table keyed by df and confidence is the fastest route if inputs are limited.
What output format does it want?+
Return an array of two strings. The first is "Yes" if the means are significantly different, otherwise "No". The second is the magnitude formatted to exactly two decimal places, like "2.47". Don't return a float or trim trailing zeros, since the string must match exactly.
How do I prepare in 48 hours for a stats-flavored OA like this?+
Refresh the t-statistic formula, sample variance with n-1, and degrees of freedom. Code it once from scratch and test it on both examples. Practice rounding and string formatting in your language. That covers nearly everything this Point72 question from July 2026 asks.