Real Numbers: Expected Time Between Exams and a Half-Court Shot Challenge

Sawyer solves an exam-scheduling problem in several ways, then asks how many basketball attempts it takes to make one shot or two in a row.

SaSawyer연사 · Jane Street
채널
Jane Street
게시일
2020년 7월 17일
영상
10:32
읽기
3분
슬라이드
4
원문
영어
YouTube에서 열기
스크롤하면 영상이 오른쪽 아래로 작아져 계속 재생됩니다. 자막은 읽기 언어를 따릅니다.
읽기 언어2개 언어로 읽을 수 있음
이 글은 아직 한국어 번역이 없어 영어 원문으로 표시합니다.
요약

30초 요약

Sawyer finds the expected separation between two independently scheduled exams in a five-day week to be 1.6 days, and between the earliest and latest of three to be 2.4 days. He uses a table, a tail-sum formula, symmetry, and linearity of expectation before leaving a half-court shooting problem for viewers.

  1. 1The half-court challengeDanielle makes 20% of her attempts; the episode asks for the expected number of attempts to make one and, as a bonus, two in a row.
  2. 2Twenty-five exam schedulesA 5-by-5 table gives an expected separation of 40/25 = 1.6 days for two exams.
  3. 3A tail-sum identityFor nonnegative integer-valued X, E[X] = Σ P(X ≥ i), summing from i = 1.
  4. 4Three exams span 2.4 daysThe tail-sum and symmetry method extends to the earliest and latest of three exams.
  5. 5Use pairwise distancesThe three pairwise gaps add up to twice the earliest-to-latest span, giving a second route to 2.4 days.

핵심 정리

  1. Two independent uniformly scheduled exams over five weekdays are separated by an expected 1.6 days.
  2. For a nonnegative integer-valued random variable, its expectation is the sum of its tail probabilities.
  3. Symmetry connects the expected earliest and latest exam days.
  4. With three independent exams, the expected span from earliest to latest is 2.4 days.
  5. For three ordered dates, the sum of all pairwise distances is twice the span.
전문

챕터별로 다듬은 전문 읽기

원본 자막만 보기
01

A basketball challenge

Sawyer poses an expected-attempts problem with a 20% half-court success rate. A bonus asks about two consecutive successes.

Welcome to Real Numbers. Our expected-value sequence ends with a basketball problem: Danielle finishes practice by taking half-court shots until she makes one.

The half-court problem
슬라이드 1 / 4The half-court problemThe challenge asks for expected attempts until the first successful shot at a 20% success rate.

Danielle makes 20% of her half-court attempts. What is the expected number of attempts before she makes one? As a bonus, what is the expected number if she stops only after making two in a row? This episode leaves the questions for viewers to solve.

02

Two exam dates

The 25 equally likely weekday pairs yield an expected separation of 1.6 days. A second method uses the expected latest and earliest dates.

First, let’s solve last week’s problem. Jayden has two final exams, each independently and uniformly scheduled on one of the five weekdays. What is the expected number of days between them?

There are 25 equally likely combinations. Make a 5-by-5 table, with the day of the math exam along one axis and the statistics exam along the other, and fill each entry with the distance between those days.

The diagonal has five zeros, for exams on the same day. The next diagonals contain eight ones, then six twos, four threes, and two fours. The fours correspond to one exam on Monday and the other on Friday.

Every entry has probability 1/25, so the expectation is (5×0 + 8×1 + 6×2 + 4×3 + 2×4)/25 = 40/25 = 8/5, or 1.6 days.

The two-exam distance table
슬라이드 2 / 4The two-exam distance tableThe table enumerates the 25 equally likely schedules and groups their day gaps.

For a third exam, directly enumerating a 5-by-5-by-5 table is more cumbersome. Instead, number weekdays from 1 to 5. If L is the later day and K the earlier day, the gap is L − K, and linearity gives E[L − K] = E[L] − E[K].

The later day is the maximum of the two dates. In the table of maxima, the number of entries equal to 3 is 3² − 2²: the entries inside the upper-left 3-by-3 square but outside the 2-by-2 square. The same pattern holds for each day.

Summing those maxima gives 1×1² + 2×(2²−1²) + 3×(3²−2²) + 4×(4²−3²) + 5×(5²−4²). Rearranging yields 125 − 16 − 9 − 4 − 1 = 95.

03

Tail sums and three exams

Sawyer introduces the tail-sum identity and applies it to three exams. Symmetry gives an expected span of 2.4 days.

This rearrangement reflects a general tail-sum identity: for a nonnegative integer-valued random variable X, E[X] = Σ from i = 1 to infinity of P(X ≥ i). It follows by expressing P(X = i) as P(X ≥ i) − P(X ≥ i+1) and rearranging the expectation sum.

The tail-sum identity
슬라이드 3 / 4The tail-sum identityThe formula expresses the expectation of a nonnegative integer-valued variable as a sum of tail probabilities.

For two exams, the expected later day is 95/25 = 19/5 = 3.8. Reversing the week maps day i to 6 − i and swaps earliest and latest, so the expected earlier day is 6 − 3.8 = 2.2. Their difference is again 1.6 days.

With three independent exams, the probability that the latest day is less than i is ((i−1)/5)³: all three exams must lie in the first i−1 days. Thus P(L ≥ i) = 1 − (i−1)³/125.

Latest of three exam days
슬라이드 4 / 4Latest of three exam daysThe calculation uses the complement that all three exam dates fall before day i.

Summing the five tail probabilities gives E[L] = 5 − (1³ + 2³ + 3³ + 4³)/125 = 4.2. By symmetry E[K] = 6 − 4.2 = 1.8, so the expected span from earliest to latest is 2.4 days.

04

A pairwise-distance shortcut

The three pairwise gaps sum to twice the full span. Linearity recovers the same three-exam result.

There is another route using the two-exam answer. For three ordered exam days, the earliest-to-middle gap plus the middle-to-latest gap equals the earliest-to-latest gap. Adding the full gap once more shows that the three pairwise distances sum to twice the full span.

Each pair has expected separation 8/5 days, regardless of which class is earliest. By linearity, twice the expected span is 3 × 8/5; the span is therefore 3/2 × 8/5 = 12/5 = 2.4 days, matching the previous method.

For this episode’s challenge, return to Danielle’s 20% half-court success rate. Work out the expected attempts for one successful shot and, for the bonus, for two consecutive successful shots, and show your reasoning.

같은 채널의 다른 글

1:48:29EN
JSJane Street· 2026년 10월 16일How AI Is Changing Programming: John Krazit and Ron MinskyRon Minsky, John Krazit · Jane StreetJane Street’s AI tooling lead and the Signals and Threads host examine agent infrastructure, security, code review and the changing experience of writing software.코딩 에이전트Developer productivity평가(Evals)+2
1:36:37EN
JSJane Street· 2026년 9월 23일Aaron Bauer on Teaching Software Engineering in the Age of AIRon Minsky, Aaron Bauer · Jane StreetHow learning goals, active practice and editor design shape engineering education when models can generate the code.AI in EducationSoftware engineeringEducation+2
슬라이드 9장1:05:45EN
JSJane Street· 2026년 9월 11일Cyrus Omar on Live Programming and a Computational Commons for Humans and AICyrus OmarA demonstration of Hazel shows how immediate feedback, structured collaboration and carefully scoped permissions could support people and AI agents working together.Programming LanguagesHuman-AI Collaboration개발 워크플로와 테스트+2
1:25:00EN
JSJane Street· 2026년 9월 2일Eric Mannes on Turning Messy External Data Into Trading ResearchRon, Eric Mannes · Jane StreetFrom commodity futures and an Excel risk system to point-in-time datasets and AI, Eric Mannes explains the domain knowledge behind data engineering at Jane Street.Data quality & curation개발 워크플로와 테스트Investing+2
슬라이드 7장1:16:29EN
JSJane Street· 2026년 8월 26일Grant Sanderson and Alok Puranik: Positional Encodings Through Group TheoryGrant Sanderson, Alok Puranik · 3Blue1BrownStarting with assumptions about attention, they derive matrix exponentials and connect rotations, decay and Jordan blocks to RoPE and ALiBi.Transformers & attentionMathematics머신러닝 기초+1
27:39EN
JSJane Street· 2026년 8월 21일3Blue1Brown and Jane Street on Trading, Machine Learning and CultureGrant Sanderson, Nitya, In Young, Craig, Alok · 3Blue1BrownGrant Sanderson talks with Jane Street colleagues about market making, shared compute, research and the paths that brought them to the firm.Machine LearningFintechCompany Culture+2