Learning Hub / Unit 6
Unit 6 of 6 · Round S-06 · target day 2026-12-05
Seal it, then reveal it: a forecast others can check
A forecast is only useful to others if they can check what you said before the answer was known. In this unit you seal a prediction about the next ONI value with a sha256 fingerprint, reveal it when NOAA CPC publishes, and turn the result into a short piece of work: original forecast, observation, error, revision and what you still do not know.
一次预测要对别人有用,别人必须能核对你在答案揭晓之前说了什么。本单元你用 sha256 指纹封存一个关于下一期 ONI 的预测,在 NOAA CPC 发布时开封,再把结果做成一份短作品:原预测、实测、误差、修订,以及仍然不知道的事。
60-minute lesson + 20-minute round touchReal data: NOAA CPC ONI to JAS 2026Team example marked SYNTHETIC
Goals and exit evidence
- Tell the story of one forecast in order: original prediction, observation, error, revision.
- Explain what a sha256 fingerprint proves (the text did not change) and what it does not (when you wrote it, or that you were right).
- Make a chart that does not mislead, and state your contribution and any tool or AI help.
Exit evidence: one page of work, a three-minute explanation, a version record and a publication choice (private is a complete choice). Expression adds no forecast points.
- 按顺序讲清一次预测:原预测、实测、误差、修订。
- 说明 sha256 指纹能证明什么(内容没改过)、不能证明什么(何时写的、是否正确)。
- 制作不误导的图表,说明自己的贡献和工具/AI 帮助。
出口证据:一页作品、三分钟说明、版本记录和公开选择(保持私密也是完整的选择)。表达不增加预测分。
Story hook
Imagine two people who both say after the fact: "I knew the index would go up." One wrote it down a month ago, put the note in an envelope and gave the envelope to a friend. The other did not. Only one of them can show it. A sha256 fingerprint is a digital envelope: change one character of the text and the fingerprint changes completely. If you share the first 8 characters with your teacher before the reveal, nobody, including you, can quietly change the prediction afterwards.
想象两个人事后都说:"我早知道指数会上升。"一个人一个月前写下来,装进信封交给朋友;另一个没有。只有前者能拿出证据。sha256 指纹就是数字信封:文本改一个字符,指纹就完全不同。开封前把前 8 位告诉老师,之后谁(包括你自己)都无法悄悄改掉预测。
Concepts
1 · Seal, reveal, score
Prediction first, data later, nothing changed, with a time stamp and a hash. This is the core of every challenge round: the server records when your forecast arrived; the hash records what it said.
预测在前、数据在后、不改动、带时间戳和哈希。这是每一轮比赛的内核:服务器记录预测何时到达,哈希记录内容是什么。
2 · Versions are kept
Changing your mind before the cutoff makes a new version; the latest on-time version counts. A revision after the reveal is kept as a learning revision, clearly labelled. It never replaces the scored version.
截止前改主意会产生新版本,最后一个准时版本计分。揭晓后的修订作为学习修订保留并清楚标注,永远不替换计分版本。
3 · Four honest boundaries
Synthetic and real are always labelled apart. One round does not show skill. Cannot tell is not the same as does not exist. Data records have their own biases, so a question should use variables measured the same way throughout.
合成与真实永远分开标注。单轮不说明技能。"看不出"不等于"不存在"。数据记录本身有偏差,题目应只用口径一致的变量。
Worked example: what was known before sealing
Our copy of the CPC ONI table (copied 2026-10-20) ends at the season shown below. The next season, ASO (August–September–October) 2026, can only be computed after October ends, so it was unknown when the copy was made. Two simple reference guesses, computed live from the table, give you a baseline to beat, just like M0 for Boston highs.
我们保存的 CPC ONI 表(2026-10-20 复制)到下面所示的季为止。下一季 ASO(8–9–10 月)2026 要等 10 月结束后才能计算,复制时尚未可知。下面两个简单参考猜测由页面实时计算,就像 Boston 最高温的 M0 一样,是你要胜过的基准。
| Season | ONI anomaly °C | Change from previous |
|---|---|---|
| Loading data… | ||
CPC's October 2026 outlooks forecast RONI, a related but different index (see Unit 5), so their numbers are not directly a forecast of ONI. Say which index your prediction is about.
CPC 2026 年 10 月的展望预报的是 RONI——相关但不同的指数(见第 5 单元),其数字不能直接当作 ONI 预报。写预测时说清是哪个指数。
Activity · Seal your prediction
Write two predictions
How the hash is made: the object without its sha256 field is written as compact JSON with sorted keys (all values are strings), then sha256 is taken of the UTF-8 bytes. This is the same canonical form the challenge server uses. The seal lives only in this browser; tell your teacher the first 8 characters now. A browser seal is a classroom exercise, not a server receipt: it proves the text, not the time.
哈希怎样计算:去掉 sha256 字段后,按键排序写成紧凑 JSON(所有值都是字符串),对 UTF-8 字节求 sha256。这与比赛服务器的规范形式相同。封存件只保存在本浏览器;请现在把前 8 位告诉老师。浏览器封存是课堂练习,不是服务器回执:它证明内容,不证明时间。
Activity · Reveal and compare
When CPC publishes the ASO 2026 ONI value and the November discussion, paste your sealed JSON, verify the hash, then enter the official values exactly as published with the source URL and the date you read them. Do not change anything in the sealed text.
CPC 发布 2026 年 ASO 的 ONI 值和 11 月讨论后,粘入封存 JSON、验哈希,再按官方原文填入数值,并写上来源 URL 和读取日期。不要改动封存文本中的任何字。
| Question | Sealed | Confidence | Official | Difference |
|---|---|---|---|---|
| Verify the hash first. | ||||
How to read it: a miss does not mean you learned less. Look at three things: what your reason was, which part of it the facts overturned, and whether your confidence matched the result. A high-confidence miss is the most valuable one to write down.
读法:猜错不等于学得差。看三件事:当时依据是什么、哪条依据被事实推翻、把握度和结果是否匹配。高把握却错,最值得写下来。
Reflection
Check for understanding
SYNTHETIC team Cyan Cloud sealed three days at one cutoff. V1 = 8, 10, 10 (before cutoff); V2 = 9, 12, 11 (last on-time version); V3 = 12, 14, 13 (after the reveal). Synthetic observations: 12, 14, 13 °C.
合成团队"青色云朵"在同一截止封存三天。V1 = 8、10、10(截止前);V2 = 9、12、11(最后准时版);V3 = 12、14、13(揭晓后)。合成实测:12、14、13 °C。
1. Which version is scored? Find the errors and the MAE.哪个版本计分?求误差与 MAE。
Answer
V2. Errors 3, 2, 2 °C; MAE = 7/3 ≈ 2.33 °C. All three forecasts were too low: one sign of a same-direction bias this time, not proof it will happen again.
V2。误差 3、2、2 °C;MAE = 7/3 ≈ 2.33 °C。三次都偏低:这一次有同方向偏差,不证明下次也会。
2. Someone charts only day 2 because day 1 had the biggest error. What is wrong?有人只画第 2 天,因为第 1 天误差最大。问题在哪?
Answer
Selective presentation. A result must show every day that was agreed in advance.
选择性呈现。结果必须展示事先约定的全部有效日。
3. May the team present V3 as its final forecast without a label?团队能否把 V3 当作最终预测展示而不标注?
Answer
No. V3 can be shown as a learning revision, clearly marked as not the scored version.
不能。V3 可以作为学习修订展示,但必须明确标注它不是计分版本。
4. Your hash matches. Does that prove you wrote the prediction before the reveal?哈希一致。这能证明你在揭晓前写下了预测吗?
Answer
Not by itself. It proves the text is unchanged since the hash was made. The time comes from an outside record: the first 8 characters you gave your teacher before the reveal, or the server's received_at in a real round.
单靠哈希不能。它证明自生成哈希后内容没变;时间要靠外部记录:揭晓前交给老师的前 8 位,或正式轮次中服务器的 received_at。
5. "My MAE was 2.33, so I am a reliable forecaster." Rewrite it."我这次 MAE 2.33,所以我是可靠的预报员。"请改写。
Answer
"This is the error of one three-day task. I need more independent tasks and a comparison before I can judge whether my method is stable."
"这是一次三日任务的误差;我还需要更多独立任务和对照,才能判断方法是否稳定。"
6. A guardian does not agree to publication. Can the learner still finish the unit?监护人不同意公开,学生还能完成本单元吗?
Answer
Yes. The work and feedback stay private. Not publishing never lowers a forecast score or removes a learning opportunity.
能。作品和反馈保持私密。不公开不会降低预测成绩,也不减少学习机会。
A one-page piece of work
- Title: what my forecast missed.
- Object and rules: station or index, unit, version scored.
- Result: all days, errors, MAE, and what it does not prove.
- What I did, what my teacher and any AI did.
- Next time: one change, written before the next cutoff.
- Publication: private, apply to publish, or undecided.
- 题目:我的预测哪里偏了。
- 对象与规则:站点或指数、单位、计分版本。
- 结果:全部日期、误差、MAE,以及它不能证明什么。
- 我做了什么,老师和 AI 做了什么。
- 下一次:只改一件事,在下次截止前写下。
- 公开状态:私密、申请公开或待决定。
Charts: label axes and units; error bars start at zero; a truncated temperature axis must show its range. Optional sound: map 1 °C to one MIDI semitone, keep the mapping, the table and the chart; never autoplay.
图表:标坐标轴和单位;误差柱从 0 开始;截断的温度轴要标出范围。声音选修:每 1 °C 对应一个 MIDI 半音,保留映射、原表和图;不自动播放。
Classroom: 20-minute round touch
| 0–3 | Read the S-06 card: target 2026-12-05, cutoff 2026-12-04 18:00 EST = 23:00 UTC.读 S-06 题卡:目标日 2026-12-05,截止 2026-12-04 18:00 EST = 23:00 UTC。 |
| 3–8 | Seal the Boston forecast; in the same session seal the ONI prediction above and give the teacher its first 8 characters.封存 Boston 预测;同时封存上面的 ONI 预测,把前 8 位交给老师。 |
| 8–13 | Pairs ask the five peer questions: original forecast? reveal source? error computable? revision marked? what is still unknown?两人互问五个问题:有原预测吗?揭晓来源?误差可算吗?修订标明了吗?还有什么不知道? |
| 13–18 | After CPC publishes: reveal, compare, write the reflection.CPC 发布后:开封、对照、写复盘。 |
| 18–20 | Choose a publication status; fill the five numbers.选择公开状态;填写五个数。 |
Full 60-minute lesson plan
| 0–8 | Show two pieces of work; which one can be checked? Find the missing source and missing original forecast.展示两份作品,哪份能复核?找缺来源与缺原预测的问题。 |
| 8–20 | Model an error chart; check axes and units; learners chart the Cyan Cloud data.示范误差图,检查坐标与单位;学生画青色云朵数据。 |
| 20–32 | Seal the ONI prediction; order four statements into an evidence story.封存 ONI 预测;把四段说明排成证据故事。 |
| 32–42 | Hometown prep card (see bonus unit B3).家乡准备卡(见附加单元 B3)。 |
| 42–55 | Three-minute talks in pairs with the five peer questions.两人一组三分钟讲解,用五个同伴问题核查。 |
| 55–60 | Private submission of work and version check; choose public, private or undecided.私密提交作品、确认版本;选公开、私密或待定。 |
Age bands
10–12
Four-box sheet: my guess, the observation, how far off, what I will try next. Oral answers may be written down by the teacher and marked as transcribed.
四格作品纸:我猜、实测、差多少、下次想试。可口述,由教师逐字记录并标"代录"。
13–15
One page of work with tool disclosure; check whether the chart axes mislead; seal and reveal the ONI prediction.
一页作品与工具披露;检查坐标是否误导;封存并开封 ONI 预测。
16–17
Add data versions, comparison conditions and reproducible steps; explain why the canonical JSON must sort keys before hashing.
加上资料版本、对照条件和可复现步骤;解释为何哈希前规范 JSON 要排序键。
No-device alternative. Write the prediction on a card, fold it, and have the teacher sign and date the outside; open it in class after CPC publishes. Squared paper and coloured pencils for the chart. No video, software purchase or photo upload is needed.
无设备替代。把预测写在卡片上折好,请教师在外侧签名并写日期;CPC 发布后课堂上开封。用方格纸和彩笔画图。不需要视频、购买软件或上传照片。
Five numbers (after each round)
Math map for this unit
| NBT-L2-DECIMAL | Decimal operations |
| CC-L2-DIFF | How many more or fewer |
| MD-L2-EST | Estimation and reasonableness |
| F-L2-FUNCDEF | Function concept |
| XC-L2-ASKED | Report the quantity asked |
Sources: SOURCES.json · CPC ONI · CPC ENSO Diagnostic Discussion · lesson text merged from m681 lesson 6, spec 41 unit 6 and the L6 seal-and-reveal page.