中级Intermediate在线实验室Online lab可打印Printable
金香蕉探险：只有成功者的地图可靠吗？
Golden Banana: Can We Trust a Map Made Only by Returners?
先看返回报告，再打开完整任务板。用三次探险机会补充证据，区分“没报告”和“没有人去”。
Read returning reports, then open the full mission board. Spend three expeditions gathering evidence and separate no report from no attempt.
A、B 入口各有三份返回报告，C 一份也没有。先别选“最安全”的路：我们知道每个入口出发过几次吗？
Gates A and B each have three returning reports; C has none. Before calling a route safest, do we know how many missions left each gate?
开始之前Before you start
给返回报告找回出发次数这个分母，用三次探险预算补证据，并准确说出数据尚不能支持的结论。
Recover departure denominators, spend three expedition tokens gathering evidence and state what the data still cannot establish.
认识分数和简单百分比；先学训练数据或模型评估更好，不需要编程。
Read fractions and simple percentages; training data or model evaluation helps. No coding required.
一步一步试试看Try it step by step
- 先猜（5分钟）：只看返回者报告：A3份、B3份、C0份。每组先写下一次选哪个入口，以及还缺哪项信息；不能把 C 没报告说成没人去过。Predict (5 min): show only returned reports A=3,B=3,C=0. Choose a gate and name missing information; no C reports does not mean nobody tried C.
- 动手（10分钟）：打开完整任务板：A出发5回3，B出发3回3，C出发2回0。补分母、算观察比例，再看原决定是否需要调整。Try (10 min): reveal departures/returns A=5/3,B=3/3,C=2/0. Supply denominators, calculate observed fractions and reconsider the decision.
- 观察（10分钟）：每组只有三枚新探险筹码。先说清本次目标是提高返回数量还是探索少见入口，再派一次、看一次结果，完整记录两种结果。Observe (10 min): give each group three new expedition tokens. State whether the goal is returns or information about less-known gates. Send one mission, reveal one outcome and record both kinds.
- 挑战（10分钟）：另做纸上题，回到初始 C 出发2回0，再增加一次返回，比例变为1/3；不沿用自由探险后的记录。能保证安全或定位过去的陷阱吗？Challenge (10 min): as a separate paper puzzle, restart from initial C=0 returns / 2 departures and add one return, giving 1/3. Do not reuse the freely explored board. Does this guarantee safety or locate earlier traps?
一起算一个例子Work through one example
某入口4次出发、2次返回，观察返回比例2/4=50%。三份返回报告本身没有告诉我们分母；“返回”也不等于已经找到金香蕉。
Two returns from four departures give an observed 2/4=50%. Three return reports alone do not reveal the denominator, and returning need not mean finding the golden banana.
先想一想，再看答案Think first, then check
B 的历史3/3能保证下一次成功吗？C 的0/2能证明整条路没有出口吗？如果只保留返回者，漏掉了哪些数据？
Does B’s past 3/3 guarantee the next return? Does C’s 0/2 prove no exit? Which data disappear when we keep only returners?
查看参考答案Show the expected answer
都不能。A观察3/5=60%，B3/3=100%，C0/2=0%；这些是小样本记录，不是真实安全概率。只看返回份数时A和B同为3，无法比较比例。独立纸上题从初始C的0/2加一次返回，得到1/3≈33.3%，不沿用自由探险后的数据，也不额外消耗实验筹码；仍不能定位过去失败处或保证未来。未返回任务也属于分母；应保留入口、出发次数和结果。三次实验目标若是补信息，可探索C；若优先返回，可基于现有记录考虑B，但都无保证。
Neither. Observed fractions are A=3/5=60%, B=3/3=100%, C=0/2=0%; tiny samples are not true safety probabilities. Equal return counts of three do not compare rates. The separate paper puzzle adds one return to initial C=0/2, giving 1/3≈33.3%; it neither reuses the freely explored board nor spends another lab token. It does not locate earlier failures or guarantee the future. Unreturned missions belong in the denominator; retain gate, departure and outcome. Exploring C may gather information; prioritizing returns may favor B from current evidence, without guarantees.
这个实验说明了什么What this experiment shows
课堂版使用固定任务卡练习信息不完整，不模拟原项目的随机迷宫、走失位置或实时多人房间。原游戏的未返回记录不会直接给出陷阱坐标；数据缺失可以由收集过程造成。
This classroom version uses fixed mission cards, not the original random maze or live multiplayer rooms. Unreturned reports do not reveal trap coordinates. Missingness can arise from how data are collected.
学习单Worksheets
学习单用英文，可打印给孩子。教师指南随页面语言下载：当前为中文版。
Print the English worksheets for the children. The guide follows the page language: English is selected.
下载学习单（PDF）Download worksheets (PDF)教师指南（中文 PDF）Teacher guide (English PDF)
继续玩原项目Explore the original game
课堂实验可以离线使用。原项目有更多模式，可由老师带领继续探索。
The classroom lab works offline. The original project has more modes for teacher-led exploration.
视频与文章Videos and reading
- Lucas Academy: original game 原项目与玩法original project and game