How to Learn Advanced English Vocabulary (Chinese Edition) | Lesson No.043 | Master Prediction, Forecast, Projection, Estimate and Scenario Without Turning a Possible Future into a Promise | 第043课：掌握预测、预报、推算、估计与情景，避免把“未来可能”写成“必然会发生”

Series ID: EDKS-ADV-VOC-ZH-0043 · Advanced English Vocabulary (Chinese Edition) · Lesson No.043 · C1 → C2 · 简体中文辅助

A forecast says what you expect. A projection says what follows if the assumptions hold. A scenario describes a coherent possible future.
forecast 说你预计会发生什么；projection 说如果假设成立会得到什么；scenario 描绘一种内部一致、可能发生的未来。

Advanced English has many future-facing nouns—prediction, forecast, projection, estimate, outlook, scenario, baseline, target, plan, expectation, assumption, extrapolation. They are not interchangeable. Some try to estimate the actual future; some deliberately ask “what if?” under assumptions; some describe a range of possible futures without claiming which one will occur; some are goals rather than expectations.

The IPCC glossary makes the distinction unusually explicit: a climate prediction or forecast attempts to estimate the actual future evolution; a projection is conditional on assumptions and scenarios that may or may not occur; a scenario is a plausible, internally consistent description of how the future may develop and is neither a prediction nor a forecast. Current ONS population guidance likewise states that national population projections are not forecasts and inevitably differ from actual future outcomes to some extent.

Never turn a conditional future into an unconditional promise.

不要把有条件的未来，改写成无条件的承诺。

This lesson owns the Mandarin-supported C1–C2 lexical system for future statement types: prediction, forecast, projection, estimate, scenario, outlook, baseline, assumption, horizon, interval, range, variant, sensitivity, extrapolation and nowcast. Lesson 035 owns probability/likelihood; Lesson 036 owns historical trends; Lesson 043 owns what kind of future claim is being made.

Part I — Build the future-claim map | 第一部分：建立“未来陈述地图”

1. Prediction is the broad attempt to say what will happen | prediction

A prediction is a statement or model output intended to say what future outcome is expected, often with uncertainty. In science it may be formal and probabilistic; in everyday speech it can be informal.

2. Forecast is prediction with a domain/time horizon | forecast

Forecast is conventional in weather, economics, sales, demand and finance. It usually aims to estimate the actual future, often over a named horizon.

3. Forecast is a noun and verb | forecast grammar

“The forecast is 3% growth.” “We forecast 3% growth.” “Growth is forecast to reach 3%.” These frames differ from project, which can also mean estimate but often carries conditional-model meaning in official projections.

4. Projection is conditional future evolution | projection

A projection asks what future values follow from assumptions, scenarios or model inputs. It does not necessarily claim that those assumptions are the most likely future.

5. Projected does not mean promised | projected

“Population is projected to reach…” means the projection produces that value under stated assumptions. It does not mean the population will certainly reach it.

6. Scenario is a coherent possible future | scenario

A scenario combines assumptions about drivers and relationships into an internally coherent future. The IPCC explicitly warns that scenarios are neither predictions nor forecasts.

7. Scenario analysis compares possible worlds | scenario analysis

Scenario analysis asks how outcomes change under different plausible conditions. It is useful under deep uncertainty when assigning reliable probabilities is difficult.

8. Estimate is broader than future language | estimate

An estimate is a calculated/inferred approximate value. It can concern the present, past or future. “Estimated population in 2025” need not be a forecast if 2025 has already occurred but complete data are unavailable.

9. Estimation and forecasting are different jobs | estimate vs forecast

Estimation often infers unknown quantities from data; forecasting extends a model into future observations. The methods can overlap but the target differs.

10. Outlook is a qualified view of future conditions | outlook

Outlook is common in economic, weather, market and industry reports. It often combines forecasts, risks and qualitative expectations rather than naming one exact numerical method.

11. Expectation is a belief/model quantity, depending domain | expectation

“We expect demand to rise” is a judgment about the future. In probability, expected value is a technical mathematical mean and is not automatically the most likely outcome.

12. Anticipate is human/organisational expectation | anticipate

“We anticipate higher demand” expresses expectation based on available information. It is less method-specific than forecast.

13. Estimate can be point or interval | point estimate

A point estimate gives one value; an interval gives a range reflecting statistical uncertainty under a method. Do not confuse an interval estimate with scenario variants.

14. Point forecast gives one future value | point forecast

A point forecast may say next-quarter sales = 10,000 units. It compresses uncertainty into one value and is often paired with a prediction/forecast interval.

15. Forecast interval describes uncertainty around future observations | forecast interval

Exact terminology varies by field; prediction interval commonly concerns a future observation/value, while confidence intervals commonly concern estimated parameters/means. Do not swap them automatically.

16. Confidence interval is not a scenario range | confidence interval

A confidence interval comes from a statistical procedure with a defined coverage interpretation. A scenario range may simply span alternative assumptions. Same visual “range,” different meaning.

17. Uncertainty range is generic and needs method | uncertainty range

Uncertainty range is useful broad language, but readers need to know whether it reflects statistical sampling, model uncertainty, scenario spread, expert judgment or sensitivity analysis.

18. Horizon is how far into the future | forecast horizon

A one-day weather forecast and a ten-year economic projection operate at different horizons. Uncertainty usually grows as the horizon extends, though exact behaviour depends on the system/model.

19. Lead time is time between forecast and event | lead time

In operations/weather, lead time can mean how far ahead the forecast is issued. Do not confuse it with causal lead–lag from Lesson 041.

20. Baseline is a comparison future/reference | baseline

A baseline scenario/projection gives a reference against which alternatives are compared. It is not automatically “what will happen” or “business as usual” unless defined that way.

21. Reference scenario serves comparison | reference scenario

Reference/baseline scenarios establish a common comparison path. They may be chosen for analytical convenience rather than probability.

22. Principal projection is not guaranteed most likely | principal projection

Official agencies may publish a principal/central projection based on assumptions considered appropriate at publication time. “Principal” does not automatically mean a quantified modal outcome.

23. Variant projection changes assumptions | variant

ONS publishes variant population projections using alternative fertility, mortality and migration assumptions. These are scenario alternatives, not necessarily probability bounds.

24. High and low variants are not confidence limits automatically | variants

A “high” and “low” scenario can illustrate assumptions rather than a statistical 95% interval. Never relabel variants as confidence bounds without documentation.

25. Assumption is an input condition | assumption

Projections are conditional on assumptions. Make the assumption visible: “Under the high-migration assumption…” is more honest than “Population will…”

26. Conditional means if assumptions hold | conditional

Conditional future claims use if, under, assuming, conditional on. Removing those phrases can turn a projection into a false forecast.

27. Sensitivity analysis changes inputs to test output response | sensitivity

Sensitivity analysis asks how conclusions change when assumptions/parameters change. It does not assign probability to alternatives unless explicitly designed to do so.

28. Stress test examines adverse hypothetical conditions | stress test

A stress test deliberately explores severe conditions to test resilience. It is not a forecast that the adverse scenario will occur.

29. Best-case and worst-case are planning constructs | best/worst case

Best/worst-case scenarios depend on chosen assumptions. They are not statistical maxima/minima or probability intervals unless formally defined.

30. Base case is a planning reference | base case

A business base case is often management’s central planning assumption set. It may be informed by forecasts but remains a scenario/plan, not guaranteed outcome.

31. Target is desired, not predicted | target

A target says what an organisation wants to achieve. “Target sales = 1m” is not “forecast sales = 1m.” Mixing target and forecast hides performance gaps.

32. Goal is normative, not predictive | goal

A goal describes intended achievement. It is not evidence about what will happen.

33. Plan is intended action, not expected outcome | plan

Plans influence futures but do not predict them. “We plan to open five sites” differs from “we forecast five sites will be operating.”

34. Budget is a planning number | budget

Budget revenue may be a target/assumption used for resource allocation. It should not be described as an unbiased forecast unless the organisation explicitly treats it that way.

35. Guidance is management expectation/communication | guidance

Corporate guidance communicates management expectations/ranges under assumptions. It is neither a contractual promise nor a generic scenario.

36. Commitment is stronger than forecast | commitment

“We commit to reduce emissions 50%” is an intended obligation/action. “We project a 50% reduction” is a conditional model result. Do not swap them.

37. Extrapolation extends an observed relationship beyond observed range/time | extrapolate

Extrapolation uses a fitted pattern beyond the data range. It becomes increasingly assumption-dependent and can fail when relationships change.

38. Interpolation estimates inside observed range | interpolate

Interpolation estimates between observed points. It is generally less assumption-heavy than far extrapolation but still model-dependent.

39. Trend extrapolation is not full forecast | extrapolated trend

Continuing a historical trend forward assumes the underlying process persists. A richer forecast may include explanatory variables, seasonality, structural breaks or expert judgment.

40. Nowcast estimates present/very recent state | nowcast

A nowcast estimates current or very recent conditions before complete observations are available. It is common in economics and meteorology. It is not a long-range forecast.

41. Hindcast/backcast reconstructs past under a model | hindcast/backcast

Hindcasting applies models to past periods to evaluate/reconstruct behaviour. Backcasting can also mean planning backwards from a desired future in sustainability/strategy, so domain matters.

42. Simulation generates model-consistent outcomes | simulation

A simulation explores outcomes under model rules/inputs. A simulated future is not automatically a prediction; it may belong to a scenario experiment.

43. Model output is not automatically forecast | model output

Models can generate historical fits, counterfactuals, scenarios, projections or forecasts. Label the output by its purpose and assumptions.

44. Expected value is not most likely value | expected value

The expected value is a probability-weighted mean. It can be a value that rarely or never occurs exactly. Do not translate “expected” into “most likely.”

45. Part I checkpoint: identify the future-claim type before reading the number | 第一部分检查点

Ask: Is this intended to estimate the actual future, show a conditional future, explore a plausible scenario, approximate an unknown quantity, set a target, or test an adverse hypothetical? The noun tells you what the number is allowed to claim.

Part II — Future language across real domains | 第二部分：真实领域中的未来表达

A future-facing number must be read with its method, assumptions and horizon. Forecasts, projections, scenarios and planning figures answer different questions.

46. Weather forecast aims at the actual near future | 天气预报

A weather forecast aims to estimate actual future conditions at a stated location and lead time. Probability and ranges express uncertainty around that forecast, not alternative planning scenarios.

47. Seasonal outlook is broader than tomorrow’s forecast | seasonal outlook

Longer-range outlooks may express probabilities of broad categories rather than precise daily conditions. Outlook signals a broader, lower-resolution expectation.

48. Climate projection is conditional on scenarios | climate projection

The IPCC defines climate projections as modelled responses conditional on scenarios and assumptions that may or may not be realised. “Projected warming under scenario X” must keep the scenario.

49. Climate scenario is not a prediction of human choices | climate scenario

An emissions scenario can describe a plausible path of technology, policy, population and emissions. It does not assert that society will choose that path.

50. Population projection is conditional demographic arithmetic | population projection

ONS states that national population projections show potential future population levels based on assumptions about fertility, mortality and migration; they are not forecasts and will differ from actual future outcomes to some extent.

51. Population variants are assumption variants | variants

High and low demographic variants show consequences of different assumptions. They are not automatically probability intervals around the principal projection.

52. Principal projection is a reference path, not certainty | principal projection

A principal projection can be analytically central without being a quantified “most likely” outcome. The label describes the official assumption set, not certainty.

53. Longer horizons usually carry more uncertainty | horizon

ONS cautions that population projections become more uncertain further into the future. Long-run figures should therefore not be paraphrased with false precision.

54. Economic forecast aims at future indicators | economic forecast

Forecasts of growth, prices or employment usually aim to estimate actual future values, often conditional on policy or market assumptions. Horizon and assumptions belong to the claim.

55. Consensus forecast aggregates forecaster expectations | consensus forecast

A consensus forecast combines several forecasts, often by mean or median. It summarises expectations but does not eliminate shared model errors or common blind spots.

56. Nowcast estimates the present or very recent past | nowcast

A nowcast fills the delay before complete observations are available. It is common in economics and weather. It is not a long-horizon forecast.

57. Sales forecast is not a sales target | sales forecast

A sales forecast estimates likely future sales. A target states desired performance. Mixing them hides the gap between expectation and ambition.

58. Demand forecast informs capacity and inventory | demand forecast

Demand forecasts estimate future demand under current information. Capacity decisions often need an uncertainty interval rather than only one point value.

59. Multi-year business projection is conditional | business projection

A business projection can model revenue, cost and cash outcomes under assumptions. The label projection is useful when the numbers are scenario-dependent rather than intended as unbiased forecasts.

60. Budget is a planning number | budget

A budget may encode targets, constraints or negotiated plans. A forecast should update expectations as information changes; a budget may remain fixed for accountability.

61. Rolling forecast updates as evidence arrives | rolling forecast

A rolling forecast extends the horizon and updates assumptions repeatedly. It differs from a static annual plan.

62. Scenario planning explores strategic uncertainty | scenario planning

Strategic scenarios can represent technology disruption, regulation changes or supply shocks without assigning probabilities. Their purpose is preparedness, not point prediction.

63. Stress scenario is deliberately adverse | stress scenario

A stress test asks whether a system remains viable under severe hypothetical conditions. An adverse scenario can be intentionally unlikely.

64. Scenario likelihood may be unspecified | deep uncertainty

Some scenarios deliberately have no probability attached. Saying one is “60% likely” adds a claim that the scenario framework may not support.

65. Simulation remains conditional on model and inputs | simulation

A simulation generates outcomes under model rules and assumptions. Simulated futures can support projections, scenarios or forecasts depending how the inputs and outputs are interpreted.

66. Ensemble forecast explores forecast uncertainty | ensemble

An ensemble combines multiple runs or models. Spread among members can inform uncertainty, but member frequency is not automatically a calibrated probability distribution.

67. Model ensemble and scenario ensemble answer different questions | ensemble types

Multiple model runs under one scenario examine model/initial-condition uncertainty. Multiple scenarios examine assumption uncertainty. A chart can contain both; label them separately.

68. Prediction interval differs from scenario envelope | interval vs envelope

A statistical prediction interval has a coverage interpretation under a model. A scenario envelope may simply span outcomes from chosen assumptions. Same visual band, different semantics.

69. Confidence band differs from future-outcome band | confidence band

A confidence band often describes uncertainty in an estimated mean or model curve. A future-observation interval also includes outcome variability and is usually wider.

70. Fan chart needs its legend | fan chart

Fan charts can show probability or uncertainty bands around a central forecast. Colour widths may encode percentiles, probability mass or another convention. Read the legend before paraphrasing.

71. Forecast calibration asks whether probabilities match outcomes | calibration

If events given 70% probability occur about 70% of the time across comparable cases, the forecasts are calibrated in a broad sense. Calibration is different from how narrow the probabilities are.

72. Sharpness is confidence concentration, not correctness | sharpness

A narrow forecast distribution is sharp. Sharpness is useful only when calibration is acceptable; confidently wrong forecasts are not better.

73. Forecast error is known after reality arrives | forecast error

Forecast error compares a forecast with the realised value. Before the outcome happens, uncertainty is not the same thing as known error.

74. Residual and forecast error are related but not identical | residual

A residual often compares fitted and observed values within a modelled sample; a forecast error compares out-of-sample/future forecast with what later occurs. Context matters.

75. Backtesting evaluates out-of-sample performance | backtesting

A forecasting method should be tested on periods not used to fit it where possible. Excellent in-sample fit does not guarantee future accuracy.

76. Hindcast tests a model against known past outcomes | hindcast

Running a model from an earlier state and comparing with historical observations can support credibility. It does not eliminate structural uncertainty about future conditions.

77. Extrapolation can fail at structural breaks | structural break

A relation fitted before a policy, technology or behavioural change may not continue afterward. Forecasts need mechanisms for regime change rather than blind trend extension.

78. Forecast revision is normal when evidence changes | revision

A rational forecast updates when new data or assumptions arrive. Revision is not automatically failure; refusing to update can be worse.

79. Forecast vintage identifies the information available at the time | vintage

Forecasts made in January and June use different information sets. Fair evaluation should compare the forecast with the data available when it was issued.

80. Data revisions complicate evaluation | real-time data

Official indicators can be revised after initial release. A forecaster working with preliminary data should not be judged as if later revisions were already known.

81. Conditional forecast says “if X, then Y” | conditional forecast

A forecast can be conditioned on policy, exchange rate, weather or other future drivers. The conditioning assumption must remain visible.

82. Scenario-conditioned projection is explicitly conditional | scenario-conditioned

“Under scenario A, emissions fall to…” is honest. Removing “under scenario A” converts a conditional result into an unconditional future claim.

83. Expected case and base case can differ | expected vs base

A base case can be chosen as the planning reference. An expected value can be probability-weighted. Organisations sometimes use these words loosely, so define them.

84. Most likely outcome is not expected value | mode vs mean

In a skewed future distribution, the mean/expected value can differ from the modal or most likely outcome. Lesson 038’s centre distinctions remain relevant.

85. Forecast accuracy depends on horizon | horizon performance

A model can be excellent one day ahead and poor one year ahead. Evaluate forecasts at the horizon used for decisions.

86. Forecast skill compares against a benchmark | skill

Forecast skill asks whether a method improves on a reference such as persistence, seasonal naive or climatology. Low raw error is less impressive if a trivial benchmark does equally well.

87. Naive forecast is a serious benchmark | naive forecast

A naive forecast may use the last observed value or the value from the same season last year. Complex models should demonstrate improvement over such baselines.

88. Forecast combination can improve robustness | combination

Combining forecasts from diverse methods can reduce reliance on one misspecified model. It does not guarantee accuracy; weights and model diversity matter.

89. Expert judgment can modify a model forecast | judgment

Domain experts may adjust forecasts around structural breaks or information not represented in the model. Document the adjustment so readers can distinguish model output from final forecast.

90. Part II checkpoint: future claims need type, horizon, assumptions and update rule | 第二部分检查点

Before comparing future numbers, verify that they are the same kind of claim, use the same horizon, condition on compatible assumptions and were produced with comparable information sets.

Part III — Mandarin-to-English future-claim control | 第三部分：中文母语学习者的未来表达转换

91. 预测 = prediction / forecast, depending domain | 预测

General “未来预测” may be prediction. Weather, demand, sales and economic contexts often prefer forecast. The English choice signals method/domain, not just future reference.

92. 预报 = forecast | 预报

“天气预报” → weather forecast. “短期预报” → short-term forecast. Do not translate weather forecast as projection unless the source is explicitly scenario-conditioned.

93. 预计 = expect / forecast / project, depending evidence | 预计

Chinese 预计 is broad. “预计明天到货” → expected tomorrow; “预计销售增长 5%” may be forecast to grow 5%; a model under assumptions may be projected to grow. Translate the claim type.

94. 估计 = estimate | 估计

“估计有 10,000 人” → an estimated 10,000 people. The estimate may refer to present or past unknown quantities, not only future values.

95. 估算 = estimate / approximate calculation | 估算

Use estimate or approximate calculation depending register. Avoid forecast if the task is estimating a current unknown quantity.

96. 推算 = projection / extrapolation / estimate | 推算

“根据当前趋势推算” can be extrapolate from the current trend. “人口推算” may be population projection. Formula and domain determine the English noun.

97. 推演 = simulate / scenario analysis / work through | 推演

Strategic or military/policy 推演 may be scenario exercise/simulation. It is not automatically a prediction of what will occur.

98. 情景 = scenario | 情景

“高排放情景” → high-emissions scenario. “基准情景” → baseline/reference scenario. Scenario describes a coherent conditional future, not necessarily the expected future.

99. 情境 can be context or scenario | 情境

“教学情境” → teaching context/situation; “未来情境” may be future scenario. Do not translate all 情境 as scenario.

100. 假设 = assumption / hypothesis | 假设

A projection input is an assumption. A testable scientific explanation may be a hypothesis. Chinese 假设 covers both; English separates them.

101. 前提 = premise / assumption / condition | 前提

“在该前提下” → under this assumption/condition. Premise is stronger in logical argument. For projections, assumption is usually natural.

102. 条件下 = under / conditional on | 条件

“在高迁移情景下” → under the high-migration scenario. This phrase protects conditionality and should not be deleted in summaries.

103. 基准 = baseline / benchmark / reference | 基准

A baseline is the reference path/starting condition; a benchmark is a comparison standard; a reference scenario is an analytical comparator. Choose the job, not the surface word.

104. 基准情景 = baseline/reference scenario | 基准情景

Use when other scenarios are compared against one reference. It does not automatically mean business-as-usual or most likely.

105. 主情景 / 主要情景 = principal/central scenario, but define status | 主情景

If an institution defines a principal projection, keep the official term. Do not translate it “most likely scenario” unless probability is explicitly assigned.

106. 高低情景 = high/low variant or scenario | 高低情景

High/low variants often reflect different assumptions. They are not automatically upper/lower confidence limits.

107. 展望 = outlook | 展望

“经济展望” → economic outlook. Outlook can combine forecasts, risks and qualitative assessment; it does not necessarily correspond to one model output.

108. 目标 = target / goal, not forecast | 目标

“销售目标” → sales target. “减排目标” → emissions-reduction target. Targets express desired outcomes; do not translate them as forecasts.

109. 计划 = plan | 计划

A plan describes intended action. “计划明年开设三家门店” → plans to open three stores next year, not “forecasts three stores.”

110. 预算 = budget | 预算

A budget is a resource-allocation/planning framework. Budgeted revenue can differ from forecast revenue because one may embody ambition or constraints.

111. 指引 / 指导值 can be guidance | guidance

Corporate or institutional guidance can communicate an expected range. Translate according to domain; it is not a generic synonym for forecast.

112. 变体 = variant | 变体预测

Official statistics may publish variant projections under alternative assumptions. A variant is one conditional path, not necessarily one quantile of probability.

113. 区间 = interval / range | 区间

“预测区间” → prediction/forecast interval; “情景范围” → scenario range; “置信区间” → confidence interval. The mathematical meaning differs.

114. 范围 = range, but ask what generated it | 范围

A range can come from uncertainty, scenarios, sensitivity tests, min–max cases or empirical variation. Never assume the endpoints have probability meaning.

115. 不确定性 = uncertainty | 不确定性

Future uncertainty can arise from random variation, unknown parameters, model structure, scenario choice or unpredictable human behaviour. One “uncertainty band” can mix several sources.

116. 敏感性分析 = sensitivity analysis | 敏感性

Use when inputs/assumptions are varied to see how outputs respond. This is not a probability analysis unless probabilities are also modelled.

117. 压力测试 = stress test | 压力测试

A stress test examines resilience under deliberately adverse conditions. It is not evidence that the stress case is expected.

118. 外推 = extrapolation | 外推

Extending a relationship beyond the observed data range/time is extrapolation. The farther the extension, the more exposed the result is to structural change.

119. 插值 = interpolation | 插值

Interpolation estimates within the observed range between known points. Do not translate as projection merely because the result is model-based.

120. 模拟 = simulation | 模拟

A simulation generates outcomes from model rules. It can support forecasts, projections or scenarios, but simulation itself does not tell you the claim type.

121. 回测 = backtest / backtesting | 回测

Backtesting evaluates a forecasting or decision method against historical periods as if the future were not yet known. Avoid leaking future information into the test.

122. 修订 = revision / update | 修订

Forecast revision is normal when data or assumptions change. “Revised forecast” does not necessarily mean the first forecast was careless.

123. 预测值 = forecast/predicted value | 预测值

Use forecast value for time-series future forecasting, predicted value for model prediction more broadly. A fitted value inside the training sample may not be a forecast.

124. 实际值 = actual / realised value | 实际值

Once the outcome is observed, compare the forecast with the actual/realised value. In experimental statistics, “observed value” may be more neutral.

125. Part III checkpoint: 预测 is not one English word | 第三部分检查点

Mandarin 预测、预计、推算、估计、情景、目标 and 假设 can all point toward the future, but English asks a sharper question: Is this an expectation of the actual future, a conditional model output, a plausible what-if future, an approximate unknown, or an intended target?

Part IV — Future-claim failure laboratory | 第四部分：未来陈述失误实验室

126. Projection becomes a false promise when “under” disappears | 条件被删掉

Original: “Under the high-migration assumption, population is projected to reach X.” Unsafe: “Population will reach X.” The summary deletes the assumption and changes projection into unconditional prediction.

127. Scenario becomes forecast when probability is invented | 情景被写成预测

Three scenarios are presented without probabilities. Writing “Scenario B is the most likely” adds an unsupported ranking. Scenario analysis can be intentionally probability-free.

128. High/low variants become fake confidence limits | 变体 ≠ 置信区间

Alternative assumptions generate a high and low projection. Calling the range “95% confidence interval” invents a statistical interpretation that the scenario set does not possess.

129. Budget becomes forecast | 预算 ≠ 预测

A team budgets 20% growth but independently forecasts 8%. Reporting “management expects 20%” hides the distinction between resource plan and expected outcome.

130. Target becomes prediction | 目标 ≠ 会实现

A target of zero emissions by a future date is a desired/committed state. A projection of current policies may show a different path. Targets should not be presented as forecasts.

131. Forecast becomes commitment | 预测 ≠ 承诺

A forecast of opening three locations is not a contractual promise to open three locations. Expectation and commitment carry different obligations.

132. Estimate becomes observation | 估计值 ≠ 实测值

An estimated current population is modelled/inferred from incomplete information. Calling it “the observed population” can overstate data completeness.

133. Nowcast becomes forecast | 当前估计 ≠ 未来预测

An estimate of this quarter before complete data arrive is a nowcast. It targets the present/recent period, not a future quarter.

134. Point forecast hides wide uncertainty | 单点隐藏不确定性

A forecast says 100 but plausible interval is 40–180. Repeating only 100 can create false precision when decisions are sensitive to tails.

135. Wide interval does not mean bad model automatically | 宽区间

An inherently unpredictable system should produce honest wide uncertainty. A narrow interval can be worse if miscalibrated.

136. Narrow interval does not mean accurate | 过度自信

A forecast can be sharply concentrated around the wrong value. Sharpness without calibration is overconfidence.

137. Expected value is impossible as one realised outcome | 期望值陷阱

A discrete outcome has expected value 3.5 even though no observation can equal 3.5. “Expected” is mathematical averaging, not “the outcome we will see.”

138. Most likely outcome differs from expected value | 众数 vs 均值

A skewed forecast distribution can have a most likely value below its mean. “Most likely” and “expected” should not be swapped.

139. Scenario mean can be meaningless | 场景平均

Three qualitatively different scenarios are not necessarily probabilistic samples. Averaging their outcomes may create a value corresponding to no coherent scenario.

140. Extrapolation crosses a structural break | 外推失效

A historical growth trend is extended beyond a regulatory/technology shift. The continuation assumption fails. “Trend projection” should state its structural assumption.

141. Short-horizon skill is used to sell long-horizon confidence | horizon mismatch

A model performs well one month ahead and is presented as validated for ten-year projections. Forecast performance does not transfer automatically across horizons.

142. In-sample fit masquerades as forecast accuracy | overfit

A model reproduces the data used to build it almost perfectly. That is not evidence of future performance. Holdout/backtest evidence matters.

143. Backtest leaks future information | look-ahead bias

A model tuned using later information is evaluated as if that information were unknown. The reported forecasting skill is inflated.

144. Data revision creates unfair hindsight | vintage problem

A forecast made with preliminary data is judged using later revised history. Evaluation should reproduce the information available at issuance where feasible.

145. Stress case is reported as central expectation | stress-test confusion

A severe scenario is deliberately chosen to test resilience. Calling it “what analysts expect” misstates the exercise.

146. Base case is reported as most likely without probability | base-case overclaim

A planning model labels one path “base case” for operational convenience. That does not automatically mean it has the highest probability.

147. Forecast revision is treated as contradiction rather than updating | revision

A new forecast incorporates new evidence. Comparing two vintages without noting information changes can make rational updating look like inconsistency.

148. Part IV checkpoint: future-language errors are claim-type errors | 第四部分检查点

Most future-communication failures come from changing the type of claim: projection→forecast, scenario→probability, target→expectation, estimate→observation, uncertainty band→scenario range, or short-horizon validation→long-horizon guarantee.

149. Build the future-claim FENCE | 未来陈述 FENCE

150. Build a future-claim control card | 未来陈述控制卡

151. Practice A — name the future claim | 练习 A

1. Tomorrow’s rainfall probability from a weather model.
2. Population in 2050 under fixed demographic assumptions.
3. A coherent high-technology future used for planning.
4. An approximate current unemployment rate before complete data arrive.
5. A desired sales level next year.

Choose among forecast, projection, scenario, nowcast/estimate, target.

152. Practice B — restore conditionality | 练习 B

Unsafe: “Population will reach X.” Source: “Under the principal migration/fertility/mortality assumptions, the population is projected to reach X.” Restore the assumptions in three concise paraphrases.

153. Practice C — range classification | 练习 C

1. 95% prediction interval.
2. High and low assumption variants.
3. Best/worst planning cases.
4. Sensitivity range from parameter changes.

Explain why the four ranges must not be relabelled as one another.

154. Practice D — forecast vs target | 练习 D

Write a report with forecast sales 80, target 100 and budget 95. Explain what each number means and how management should interpret the gaps without merging them.

155. Practice E — Mandarin repair | 练习 E

1. 在高迁移假设下，人口预计达到……
2. 这是一种情景，不是预测。
3. 预算目标高于当前销售预测。
4. 短期预报区间较窄，但长期不确定性扩大。
5. 模型只是外推当前趋势，因此结构变化可能使结果失效。

156. A 15-minute future-language session | 15 分钟训练

157. Seven-day mastery route | 七天路线

158. First weak-link diagnostic | 弱点诊断

159. Reading harvest | 阅读提取

Collect complete frames: is forecast to, is projected under, assuming that, under the baseline scenario, variant projection, prediction interval, revised outlook, stress scenario. The qualifier is part of the claim.

160. Writing activation | 写作激活

Use: future quantity → claim type → horizon → assumptions → central value → uncertainty/variants → update date. This lets readers reconstruct exactly what kind of future number they are seeing.

161. Speaking activation | 口语激活

Explain one projection in 45 seconds without using “will” until you have explicitly said what assumptions make it conditional.

162. Paraphrase test: keep conditionality | 改述测试

Original: “Under scenario A, temperatures are projected to rise by X.” Unsafe: “Temperatures will rise by X.” Safe: “The model projects X warming if scenario A’s assumptions are realised.”

163. Summary test: preserve range meaning | 摘要测试

If a report shows scenario variants, do not compress them into “uncertainty ±X” unless the source interprets them that way. Summary must preserve what generated the range.

164. Canonical ownership boundary | 本课所有权边界

Lesson 043 owns future-claim types: prediction, forecast, projection, estimate, scenario, outlook, baseline, assumption, horizon, variants, intervals, extrapolation, nowcasting and forecast evaluation language. Lesson 035 owns likelihood/risk; Lesson 036 historical trends; Lesson 044 measurement/forecast accuracy terminology.

165. Recommended reference floor | 推荐参考资源

• IPCC AR6 WGII Glossary — prediction/forecast, projection and scenario definitions.
• IPCC AR6 WGII Chapter 1 — scenarios as “what-if” futures rather than forecasts.
• ONS — National Population Projections Quality and Methods — projections are not forecasts and uncertainty grows with horizon.
• ONS — 2024-based Variant Projections — alternative demographic assumption scenarios.

166. SEO language map | 本课关键词范围

prediction vs forecast, forecast vs projection, projection vs scenario, scenario not prediction, estimate vs forecast, forecast interval, prediction interval, future uncertainty, population projection, climate projection, baseline scenario, stress scenario, target vs forecast, 预测英语, 预报英语, 推算英语, 情景英语, 预计英语, 估计英语.

167. Your assignment | 本课作业

Find four public future-facing numbers: one forecast, one projection, one scenario output and one target. For each, build the FENCE and write a 150-word explanation showing why the nouns cannot be exchanged.

168. Why this owner earns longform depth | 为什么需要长文深度？

The learning job is not memorising four synonyms. It is preserving the epistemic status of a future statement across weather, official statistics, climate, business planning and model outputs. Removing one word—under, scenario, target, projected—can silently turn a conditional analysis into a promise.

169. Final rule one | 最后一条规则一

A forecast says what you expect. A projection says what follows if the assumptions hold. A scenario says what could coherently happen.

forecast 说你预计会发生什么；projection 说如果假设成立会得到什么；scenario 描绘一种内部一致、可能发生的未来。

170. Final rule two | 最后一条规则二

Never turn a conditional future into an unconditional promise.

不要把有条件的未来，改写成无条件的承诺。