GOOGL / MODEL RESEARCH / 2026-10-09
Google估值模型:方法、核算与逐代验证Google valuation models: methods, calculations and iteration evidenceGoogle估值模型:方法、核算與逐代驗證
这里保存模型为什么改变、怎样计算、何时采用或拒绝。2026-10-09为本次网页整理日期;下列保留预测的输入为2026-10-06,并没有在发布日重新计算。This page records why models changed, how they calculate and when they were adopted or rejected. October 9, 2026 is the publication-preparation date; the retained forecast uses October 6 inputs and was not recalculated on publication day.這裏保存模型為什麼改變、怎樣計算、何時採用或拒絕。2026-10-09為本次網頁整理日期;下列保留預測的輸入為2026-10-06,並沒有在發佈日重新計算。
G19:滞后MSFT倍数状态,有限改善G18G19: lagged MSFT multiple state, a small improvement on G18G19:滯後MSFT倍數狀態,有限改善G18
研究日期2026-10-12;Google财务、利率和价格评价仍使用2026-10-06冻结截面,最新Google财务季度2026Q2。MSFT状态来自前一完整月末2026-09-30,不能称10月12日最新行情重算。Research date: October 12, 2026. Google financial inputs, rates and price evaluation remain frozen at October 6, 2026, with latest Google financial quarter 2026Q2. The MSFT state is from the preceding complete month-end, September 30, 2026; these are not October 12 live-market calculations.研究日期2026-10-12;Google財務、利率和價格評價仍使用2026-10-06凍結截面,最新Google財務季度2026Q2。MSFT狀態來自前一完整月末2026-09-30,不能稱10月12日最新行情重算。
| 期限与用途Horizon and role期限與用途 | G19候选估值G19 candidate valueG19候選估值 | 冻结Core EPSFrozen Core EPS凍結Core EPS | 同一条件PEShared conditional P/E同一條件PE | 估值日期 / 截止日Valuation date / deadline估值日期 / 截止日 |
|---|---|---|---|---|
| 0M · 条件公平估计Conditional fair estimate條件公平估計 | $321.28 | 12.9502 | 24.8090× | 2026-10-06 |
| 6M · 期限估值Horizon valuation期限估值 | $340.47 | 13.7238 | 24.8090× | 2027-04-06 |
| 12M · 期限估值Horizon valuation期限估值 | $377.67 | 15.2232 | 24.8090× | 2027-10-06 |
五年6M触达15/18,12M触达13/16,与G18均相同。平均绝对峰值差6M由21.39%降至21.34%,12M由31.08%降至30.88%,改善有限,仍高于G05。当前同行系数为0.00000000,所以当前三期限值与G18完全相同;历史20个季度中有14个起点同行系数为正,回测结果可以微幅改变。Five-year hits are 15/18 at 6M and 13/16 at 12M, unchanged from G18. Mean absolute peak gaps fall from 21.39% to 21.34% at 6M and from 31.08% to 30.88% at 12M. The improvement is small and gaps remain above G05. The current peer slope is 0.00000000, leaving current values identical to G18. It is positive at 14 of the twenty historical quarterly origins, so retrospective scores can change slightly.五年6M觸達15/18,12M觸達13/16,與G18均相同。平均絕對峯值差6M由21.39%降至21.34%,12M由31.08%降至30.88%,改善有限,仍高於G05。當前同行係數為0.00000000,所以當前三期限值與G18完全相同;歷史20個季度中有14個起點同行係數為正,回測結果可以微幅改變。
G19的18项冻结条件中,11项通过、7项失败,未替换主预测。个股页继续分别保留G16的0M条件公平估计$318.51与G05的6M $368.45、12M $464.42;三个报价标注各自版本,不组合成新模型。11 of 18 frozen conditions pass and 7 fail. G19 does not replace the retained forecasts. The equity page retains separate G16 conditional 0M value $318.51 and G05 6M / 12M forecasts $368.45 / $464.42, each labeled with its own version rather than combined into a new model.G19的18項凍結條件中,11項通過、7項失敗,未替換主預測。個股頁繼續分別保留G16的0M條件公平估計$318.51與G05的6M $368.45、12M $464.42;三個報價標註各自版本,不組合成新模型。
G18:财报兑现差驱动独立PE,改善G16但未替换G05G18: profit realization in independent P/E; improves G16 but does not replace G05G18:財報兑現差驅動獨立PE,改善G16但未替換G05
研究日期2026-10-12;财务、利率及价格评价截面仍为2026-10-06,最新财务季度2026Q2。以下不是10月12日新行情重算。Research date: October 12, 2026. Financial inputs, rates and price evaluation remain frozen at October 6, 2026, with latest financial quarter 2026Q2. These are not updated October 12 market calculations.研究日期2026-10-12;財務、利率及價格評價截面仍為2026-10-06,最新財務季度2026Q2。以下不是10月12日新行情重算。
| 期限与用途Horizon and role期限與用途 | G18候选计算值G18 candidate valueG18候選計算值 | 原经营EPS代理Unchanged operating-EPS proxy原經營EPS代理 | 同一独立条件PEShared independent conditional P/E同一獨立條件PE | 参考日期 / 截止日Snapshot / deadline參考日期 / 截止日 |
|---|---|---|---|---|
| 0M · 条件公平估计候选Conditional fair-estimate candidate條件公平估計候選 | $321.28 | 12.9502 | 24.8090× | 2026-10-06 |
| 6M · 期限估值候选Horizon valuation candidate期限估值候選 | $340.47 | 13.7238 | 24.8090× | 2027-04-06 |
| 12M · 期限估值候选Horizon valuation candidate期限估值候選 | $377.67 | 15.2232 | 24.8090× | 2027-10-06 |
G18相对G16改善:6M触达由14/18升至15/18,平均峰值差由24.29%降至21.39%;12M仍为13/16触达,峰值差由34.25%降至31.08%。但G05的6M为16/18、峰值差15.80%,12M峰值差26.59%,仍优于G18。新特征有增量作用,尚不足以替换主预测。G18 improves on G16: 6M hits rise from 14/18 to 15/18 and mean peak gap falls from 24.29% to 21.39%. 12M remains at 13/16 hits, with peak gap falling from 34.25% to 31.08%. G05 still has 16/18 hits and a 15.80% peak gap at 6M, and a 26.59% peak gap at 12M. The new feature has an incremental effect but does not justify replacing the retained forecasts.G18相對G16改善:6M觸達由14/18升至15/18,平均峯值差由24.29%降至21.39%;12M仍為13/16觸達,峯值差由34.25%降至31.08%。但G05的6M為16/18、峯值差15.80%,12M峯值差26.59%,仍優於G18。新特徵有增量作用,尚不足以替換主預測。
G18的18项替换条件中,11项通过、7项失败。个股页继续分别显示G16的0M条件公平估计$318.51与G05的6M $368.45、12M $464.42主预测;不同模块标注各自版本,不组合成一套新模型。Of G18’s eighteen replacement conditions, eleven pass and seven fail. The equity page continues to display G16’s separate conditional 0M estimate of $318.51 and G05’s retained 6M / 12M forecasts of $368.45 / $464.42. The modules retain their own version labels and do not form a combined new model.G18的18項替換條件中,11項通過、7項失敗。個股頁繼續分別顯示G16的0M條件公平估計$318.51與G05的6M $368.45、12M $464.42主預測;不同模塊標註各自版本,不組合成一套新模型。
G17:独立PE的期限触达校准,未替换保留预测G17: attainable-window calibration of independent P/E; retained forecasts unchangedG17:獨立PE的期限觸達校準,未替換保留預測
研究日期2026-10-12;财务、利率与价格评价截止仍冻结于2026-10-06,最新财务期2026Q2。以下不是10月12日新行情或新财报后的重算。Research date: October 12, 2026. Financial inputs, rates and price evaluation remain frozen at October 6, 2026, with latest financial period 2026Q2. These are not refreshed October 12 quotes or a post-earnings revaluation.研究日期2026-10-12;財務、利率與價格評價截止仍凍結於2026-10-06,最新財務期2026Q2。以下不是10月12日新行情或新財報後的重算。
| 期限与用途Horizon and role期限與用途 | 经营EPS代理Operating-EPS proxy經營EPS代理 | 独立PE / 目标PEIndependent / target P/E獨立PE / 目標PE | G17计算值G17 calculated valueG17計算值 | 参考日期 / 截止日Snapshot / deadline參考日期 / 截止日 |
|---|---|---|---|---|
| 0M · G16条件公平估计,沿用G16 conditional estimate, unchangedG16條件公平估計,沿用 | 12.9502 | 24.5946× | $318.51 | 2026-10-06 |
| 6M · G17期间触达候选,未采用G17 attainable candidate, not adoptedG17期間觸達候選,未採用 | 13.7238 | 21.3879× | $293.52 | 2027-04-06 |
| 12M · G17期间触达候选,未采用G17 attainable candidate, not adoptedG17期間觸達候選,未採用 | 15.2232 | 21.8197× | $332.16 | 2027-10-06 |
6M触达率由G16的77.78%提升至88.89%,只追平G05;峰值差扩大至30.86%,差于G05的15.80%和G16的24.29%。12M触达率降至75.00%,低于两基准的81.25%;峰值差扩大至48.13%。当前目标变低并没有同时改善触达与最高价接近程度。6M hit rate rises from G16’s 77.78% to 88.89%, only matching G05. Peak gap widens to 30.86%, worse than G05’s 15.80% and G16’s 24.29%. 12M hits fall to 75.00%, below both references at 81.25%, while peak gap widens to 48.13%. Lower current targets did not jointly improve touch and proximity to the window high.6M觸達率由G16的77.78%提升至88.89%,只追平G05;峯值差擴大至30.86%,差於G05的15.80%和G16的24.29%。12M觸達率降至75.00%,低於兩基準的81.25%;峯值差擴大至48.13%。當前目標變低並沒有同時改善觸達與最高價接近程度。
G17未通过替换门槛:13项失败、5项通过。个股页继续单列G16的0M条件公平估计$318.51,以及G05保留的6M $368.45、12M $464.42。不同版本分别标示用途,不构成一套混合新模型。G17 failed replacement: thirteen gates fail and five pass. The equity page continues to show G16’s separate conditional 0M estimate of $318.51 and G05’s retained 6M / 12M forecasts of $368.45 / $464.42. Versions retain separate purposes; they do not constitute one combined model.G17未通過替換門檻:13項失敗、5項通過。個股頁繼續單列G16的0M條件公平估計$318.51,以及G05保留的6M $368.45、12M $464.42。不同版本分別標示用途,不構成一套混合新模型。
查看期限PE、五年季度逐次记录与验证Read horizon P/E, all five-year quarterly records and validation查看期限PE、五年季度逐次記錄與驗證
G16:独立0M条件公平估值,未来预测仍保留G05G16: independent conditional 0M estimate; G05 future forecasts retainedG16:獨立0M條件公平估值,未來預測仍保留G05
研究整理日期2026-10-09;盈利、利率和行情输入仍固定于2026-10-06,最新财务期2026Q2。这不是发布日的新行情或重新估值。Research publication date: October 9, 2026. Earnings, rate and market inputs remain frozen at October 6, 2026, with latest financial period 2026Q2. These are not publication-day quotes or refreshed values.研究整理日期2026-10-09;盈利、利率和行情輸入仍固定於2026-10-06,最新財務期2026Q2。這不是發佈日的新行情或重新估值。
| 当前估值口径Current estimate當前估值口徑 | NTM经营EPS代理NTM operating-EPS proxyNTM經營EPS代理 | 独立估计PEEstimated independent P/E獨立估計PE | 美元/股USD per share美元/股 |
|---|---|---|---|
| G16 · 0M条件公平价G16 · conditional 0M estimateG16 · 0M條件公平價 | 12.9502 | 24.5946× | $318.51 |
0M采用当前预测NTM经营EPS×独立估计PE,不把当天收盘价写成合理价值。PE依据历史市场倍数和已知成长、利率状态估计,独立于当天价格;它仍是历史市场条件公平估值,未完成DCF内在价值证明。0M uses the current forecast NTM operating-EPS proxy × an independently estimated P/E, rather than calling that day’s close fair value. The multiple comes from historical market valuations and known growth/rate states, independently of today’s price. It remains a historically calibrated conditional estimate, not validated intrinsic DCF value.0M採用當前預測NTM經營EPS×獨立估計PE,不把當天收盤價寫成合理價值。PE依據歷史市場倍數和已知成長、利率狀態估計,獨立於當天價格;它仍是歷史市場條件公平估值,未完成DCF內在價值證明。
G16未通过替换未来预测的冻结门槛。6M、12M继续保留G05的$368.45、$464.42;G16的独立0M身份不代表它对未来价格更准确。G16 failed the frozen future-forecast replacement screen. G05 remains the 6M / 12M reference at $368.45 / $464.42. Independence of G16’s 0M estimate does not establish greater future-price accuracy.G16未通過替換未來預測的凍結門檻。6M、12M繼續保留G05的$368.45、$464.42;G16的獨立0M身份不代表它對未來價格更準確。
查看PE计算、五年与三年季度验证Read the P/E calculation and five-/three-year quarterly evaluation查看PE計算、五年與三年季度驗證
保留G05:有日期的研究价格Retained G05: dated research prices保留G05:有日期的研究價格
输入截面2026-10-06,最新财务期2026Q2。0M $347.68为该日市场收盘锚;6M $368.45、12M $464.42为条件研究预测。未采用候选和条件试算不进入这张主表。Input snapshot: October 6, 2026; latest financial quarter: 2026Q2. 0M $347.68 is that date’s market-close anchor. 6M $368.45 and 12M $464.42 are conditional research forecasts. Unadopted candidates and document trials do not enter this main table.輸入截面2026-10-06,最新財務期2026Q2。0M $347.68為該日市場收盤錨;6M $368.45、12M $464.42為條件研究預測。未採用候選和條件試算不進入這張主表。
| 期限Horizon期限 | 美元/股USD per share美元/股 | 目标日期Target date目標日期 | 目标日之后一年盈利窗口(右端不含)Twelve-month earnings window after target (end exclusive)目標日之後一年盈利窗口(右端不含) |
|---|---|---|---|
| 0M | $347.68 | 2026-10-06 | 2026-10-06 → 2027-10-06 |
| 6M | $368.45 | 2027-04-06 | 2027-04-06 → 2028-04-06 |
| 12M | $464.42 | 2027-10-06 | 2027-10-06 → 2028-10-06 |
0M不是独立公允价值。Core EPS是规范化营业利润×0.85再除起点股数的研究代理,不是官方GAAP或供应商调整后EPS。6M和12M指目标日期的价格;盈利分母均为该目标日之后12个月,不能把6M EPS理解为六个月累计盈利。0M is not an independent fair value. Core EPS is a research proxy based on normalized operating profit × 0.85 divided by origin shares; it is neither reported GAAP nor vendor-adjusted EPS. 6M and 12M denote target-date prices; the earnings windows each cover the following twelve months, not six months of cumulative earnings.0M不是獨立公允價值。Core EPS是規範化營業利潤×0.85再除起點股數的研究代理,不是官方GAAP或供應商調整後EPS。6M和12M指目標日期的價格;盈利分母均為該目標日之後12個月,不能把6M EPS理解為六個月累計盈利。
单独查看最新财报后季度档案:2026-07-23View the separate latest post-earnings quarterly vintage: July 23, 2026單獨查看最新財報後季度檔案:2026-07-23
用户要求每季财报后一次的统一回测,最后一个起点为2026-07-23。该G05档案为0M $317.69、6M $338.82、12M $423.88。它与上方10月6日研究快照有不同价格锚、可用来源和目标日期,不能拼接。The quarterly-only post-earnings comparison has its last origin on July 23, 2026. That G05 vintage is 0M $317.69, 6M $338.82 and 12M $423.88. Its market anchor, available source inputs and target dates differ from the October 6 research snapshot; do not splice the two.用户要求每季財報後一次的統一回測,最後一個起點為2026-07-23。該G05檔案為0M $317.69、6M $338.82、12M $423.88。它與上方10月6日研究快照有不同價格錨、可用來源和目標日期,不能拼接。
| 期限Horizon期限 | 美元/股USD per share美元/股 | 目标日期Target date目標日期 | 目标日之后一年盈利窗口(右端不含)Twelve-month earnings window after target (end exclusive)目標日之後一年盈利窗口(右端不含) |
|---|---|---|---|
| 0M | $317.69 | 2026-07-23 | 2026-07-23 → 2027-07-23 |
| 6M | $338.82 | 2027-01-23 | 2027-01-23 → 2028-01-23 |
| 12M | $423.88 | 2027-07-23 | 2027-07-23 → 2028-07-23 |
G05怎样从盈利得到价格How G05 converts earnings into pricesG05怎樣從盈利得到價格
先用最新已披露收入和营业利润估计盈利路径,再用同口径起点PE映射到价格。0M/6M保留原4季度增长半衰期;12M单独使用8季度半衰期及自己的EPS0锚。参数不是根据本次网页成绩重新选择。Forecast an earnings path from known revenue and operating profit, then map it to prices with an origin multiple on the same earnings basis. 0M/6M retain the four-quarter growth half-life. 12M separately uses eight quarters and its own EPS0 anchor. Parameters were not reselected for this publication.先用最新已披露收入和營業利潤估計盈利路徑,再用同口徑起點PE映射到價格。0M/6M保留原4季度增長半衰期;12M單獨使用8季度半衰期及自己的EPS0錨。參數不是根據本次網頁成績重新選擇。
| 起点假设Origin assumption起點假設 | 数值Value數值 |
|---|---|
| TTM营业利润率TTM operating marginTTM營業利潤率 | 33.11% |
| 税率代理Tax proxy税率代理 | 15% |
| 固定稀释股数Fixed diluted shares固定稀釋股數 | 12.309 B |
6M:原HL4盈利 × 自身市场PE6M: original HL4 earnings × own market P/E6M:原HL4盈利 × 自身市場PE
13.723786467919583 × 26.847421347114157 = 368.4482777820606
6M不新增价格校准,PE=P0/CoreEPS0。这个倍数保持不变是研究条件,不是已经验证的合理PE。6M adds no price calibration; P/E=P0/CoreEPS0. Holding this multiple fixed is a research condition, not a verified fair P/E.6M不新增價格校準,PE=P0/CoreEPS0。這個倍數保持不變是研究條件,不是已經驗證的合理PE。
12M:F8自身盈利锚与自身历史校准12M: F8 own earnings anchor and own historical calibration12M:F8自身盈利錨與自身歷史校準
EPS0,F8 = 13.185966850625862 PE8 = 347.68 / EPS0,F8 = 26.367425607739765 Raw12M = 15.223176609354857 × 26.367425607739765 = 401.39597676064824 Calibration factor = exp(0.145841178688327) = 1.1570124152282537 Final12M = Raw12M × Calibration factor = 464.4201285347416 Equivalent P/E = Final12M / EPS12,F8 = 30.50743878576229
因此未校准12M条件价为$401.40,主表校准后研究价为$464.42。26.3674倍是F8自身市场锚;30.5074倍只是价格校准等价PE,不是独立公允倍数。不能把G01的EPS0或26.8474倍旧锚乘进F8分支,也不能重复叠加旧G03校准。The uncalibrated 12M conditional price is therefore $401.40; the main table uses the calibrated $464.42 research price. 26.3674× is F8’s own market anchor. 30.5074× is only a calibration-equivalent P/E, not an independent fair multiple. Do not use G01 EPS0 or the old 26.8474× anchor in the F8 branch, or apply G03 calibration again.因此未校準12M條件價為$401.40,主表校準後研究價為$464.42。26.3674倍是F8自身市場錨;30.5074倍只是價格校準等價PE,不是獨立公允倍數。不能把G01的EPS0或26.8474倍舊錨乘進F8分支,也不能重複疊加舊G03校準。
| 符号Symbol符號 | 普通语言说明Plain-language meaning普通語言説明 |
|---|---|
| NOPAT | 营业利润乘假设税后比例;研究盈利代理Operating profit times the assumed after-tax fraction; a research earnings proxy營業利潤乘假設税後比例;研究盈利代理 |
| g0 / s_j | 起点季度增长锚 / 收入季节性Origin growth anchor / revenue seasonality起點季度增長錨 / 收入季節性 |
| c / exp(c) | 已成熟历史残差的对数修正 / 乘性价格因子Log correction from mature historical residuals / multiplicative price factor已成熟歷史殘差的對數修正 / 乘性價格因子 |
| Kish / h | 沿用的重叠样本折扣启发值,不是真实独立样本数A heuristic overlap discount, not the number of independent observations沿用的重疊樣本折扣啓發值,不是真實獨立樣本數 |
| EV / k | 经营企业价值 / 经营贴现率代理Operating enterprise value / operating discount-rate proxy經營企業價值 / 經營貼現率代理 |
G01–G15逐代模型Every generation: G01–G15G01–G15逐代模型
下表的G01–G14数字统一是原2026-10-06研究截面,G15为单独标注的6月30日条件试算。打开每代查看研究目的、方法、公式、冻结参数、两套日期明确的数值和原采用决定。原失败结论没有因后续评价目标改变而回写。G01–G14 values below use the original October 6 research snapshot. G15 is the separately labeled June 30 conditional trial. Open each generation for its question, method, formulas, frozen parameters, separately dated outputs and original adoption decision. Later changes to the evaluation objective do not rewrite earlier rejections.下表的G01–G14數字統一是原2026-10-06研究截面,G15為單獨標註的6月30日條件試算。打開每代查看研究目的、方法、公式、凍結參數、兩套日期明確的數值和原採用決定。原失敗結論沒有因後續評價目標改變而回寫。
| 版本Version版本 | 研究方向Research direction研究方向 | 身份Status身份 | 6M | 12M |
|---|---|---|---|---|
| G01G01G01 | G01 · H09-RDA 基本面与同口径PE基准G01 · H09-RDA earnings and constant-basis P/E benchmarkG01 · H09-RDA 基本面與同口徑PE基準 | 原基准应用Original benchmark application原基準應用 | $368.45 | $382.84 |
| G02G02G02 | G02 · 营业利润率趋势G02 · Operating-margin trendG02 · 營業利潤率趨勢 | 保留未采用版本Preserved, not adopted保留未採用版本 | $370.60 | $386.63 |
| G03G03G03 | G03 · 按已兑现期末价格误差校准G03 · Mature terminal-price residual calibrationG03 · 按已兑現期末價格誤差校準 | 保留未采用版本Preserved, not adopted保留未採用版本 | $405.62 | $446.30 |
| G04G04G04 | G04 · 12M均线状态校准G04 · 12M moving-average state calibrationG04 · 12M均線狀態校準 | 保留未采用版本Preserved, not adopted保留未採用版本 | $368.45 | $444.68 |
| G05G05G05 | G05 · 12M增长持续期与自身校准G05 · 12M growth duration and own-path calibrationG05 · 12M增長持續期與自身校準 | 保留研究模型Retained research model保留研究模型 | $368.45 | $464.42 |
| G06G06G06 | G06 · 6M统一F8路径校准G06 · F8 path with separate 6M calibrationG06 · 6M統一F8路徑校準 | 保留未采用版本Preserved, not adopted保留未採用版本 | $411.29 | $464.42 |
| G07G07G07 | G07 · 6M三专家动态组合G07 · Dynamic 6M three-expert mixtureG07 · 6M三專家動態組合 | 保留未采用版本Preserved, not adopted保留未採用版本 | $381.16 | $464.42 |
| G08G08G08 | G08 · 趋势与预期盈利增长双状态岭校准G08 · Trend and expected-earnings ridge statesG08 · 趨勢與預期盈利增長雙狀態嶺校準 | 保留未采用版本Preserved, not adopted保留未採用版本 | $411.87 | $462.23 |
| G09G09G09 | G09 · 双状态Huber稳健校准G09 · Huber calibration of two statesG09 · 雙狀態Huber穩健校準 | 保留未采用版本Preserved, not adopted保留未採用版本 | $406.88 | $456.34 |
| G10G10G10 | G10 · 分部成长与资本回报SOTP-DCFG10 · Segment growth and reinvestment SOTP-DCFG10 · 分部成長與資本回報SOTP-DCF | 保留未采用版本Preserved, not adopted保留未採用版本 | $182.60 | $190.52 |
| G11G11G11 | G11 · 固定G05盈利的国债利率状态G11 · Treasury-rate state with fixed G05 earningsG11 · 固定G05盈利的國債利率狀態 | 保留未采用版本Preserved, not adopted保留未採用版本 | $413.91 | $545.04 |
| G12G12G12 | G12 · 利率状态绝对对数损失G12 · Absolute-log loss for the yield stateG12 · 利率狀態絕對對數損失 | 保留未采用版本Preserved, not adopted保留未採用版本 | $412.60 | $603.73 |
| G13G13G13 | G13 · 固定盈利的现金转化状态G13 · Cash-conversion state with fixed earningsG13 · 固定盈利的現金轉化狀態 | 保留未采用版本Preserved, not adopted保留未採用版本 | $411.13 | $587.03 |
| G14G14G14 | G14 · 文档固定PE适配G14 · Document fixed-P/E adaptationG14 · 文檔固定PE適配 | 保留未采用版本Preserved, not adopted保留未採用版本 | $349.96 | $395.80 |
| G15G15G15 | G15 · 文档宏观倍数与组合条件试验G15 · Document macro-multiple and ensemble trialG15 · 文檔宏觀倍數與組合條件試驗 | 条件试验,正式不可估算Conditional trial; formal model not estimable條件試驗,正式不可估算 | 正式值空;条件$371.52Formal null; conditional $371.52正式值空;條件$371.52 | 正式值空;条件$375.14Formal null; conditional $375.14正式值空;條件$375.14 |
G01 · H09-RDA 基本面与同口径PE基准G01 · H09-RDA earnings and constant-basis P/E benchmarkG01 · H09-RDA 基本面與同口徑PE基準 · 原基准应用Original benchmark application原基準應用
为了研究什么Research question為了研究什麼
回答盈利增长而市场愿付同口径倍数不变时,6M和12M价格应如何随经营盈利滚动。Study how 6M and 12M prices roll with operating earnings when the market pays the same earnings-basis multiple.回答盈利增長而市場願付同口徑倍數不變時,6M和12M價格應如何隨經營盈利滾動。
具体方法与本代改变Method and changes具體方法與本代改變
以最新可用季度同比日收入作为增长锚,12个连续季度拟合收入季节性;增长贡献按4季半衰期减弱。固定最新TTM营业利润率、15%税率及起点股数,按真实季度日历计算目标日之后一年的经营盈利。Anchor growth to the latest known quarterly daily-revenue year-over-year change, fit seasonality from 12 consecutive quarters, and damp growth with a four-quarter half-life. Hold the latest TTM operating margin, 15% tax proxy and origin shares; allocate earnings over the twelve months after each target date.以最新可用季度同比日收入作為增長錨,12個連續季度擬合收入季節性;增長貢獻按4季半衰期減弱。固定最新TTM營業利潤率、15%税率及起點股數,按真實季度日曆計算目標日之後一年的經營盈利。
GOOGL最初基准;用户确认G01为H09-RDA/G00的别名,没有另一份独立G01算法。The original GOOGL benchmark. The user confirmed G01 aliases H09-RDA / G00; it is not a separate algorithm.GOOGL最初基準;用户確認G01為H09-RDA/G00的別名,沒有另一份獨立G01算法。
核算公式Calculation formulas核算公式
CoreEPS_h = NOPAT_[target,target+12M) / origin diluted shares PE0 = P0 / CoreEPS0; P_h = CoreEPS_h × PE0 P_h = P0 × NOPAT_h / NOPAT_0 D_j = D_(j−1) × exp(s_j−s_(j−1)+g0×2^(−(j−1)/4)), j≥2
冻结参数Frozen parameters凍結參數
| 参数Parameter參數 | 取值与规则Value and rule取值與規則 |
|---|---|
| 增长半衰期Growth half-life增長半衰期 | 4季度4 quarters4季度 |
| 季节性历史Seasonality lookback季節性歷史 | 12连续季度12 consecutive quarters12連續季度 |
| 税率代理Tax proxy税率代理 | 15%15%15% |
| 股份Shares股份 | 固定起点股数;缺失EPS/PE留空Fixed origin shares; missing EPS/P/E remain null固定起點股數;缺失EPS/PE留空 |
本代计算值与日期Version outputs and dates本代計算值與日期
| 期限Horizon期限 | 原研究2026-10-06Original research: 2026-10-06原研究2026-10-06 | 原研究目标日期Original target date原研究目標日期 | 季度重建2026-07-23Quarterly reconstruction: 2026-07-23季度重建2026-07-23 | 季度目标日期Quarterly target date季度目標日期 |
|---|---|---|---|---|
| 0M | $347.68 | 2026-10-06 | $317.69 | 2026-07-23 |
| 6M | $368.45 | 2027-04-06 | $338.82 | 2027-01-23 |
| 12M | $382.84 | 2027-10-06 | $354.83 | 2027-07-23 |
上述两个输入截面分别保存,不能混为发布日的当前行情。0M是该截面的市场锚;本版未采用时,候选6M/12M不替换G05主表。These input snapshots are separately preserved, not publication-day market quotes. 0M is the snapshot’s market anchor. An unadopted candidate’s 6M/12M outputs do not replace the G05 main table.上述兩個輸入截面分別保存,不能混為發佈日的當前行情。0M是該截面的市場錨;本版未採用時,候選6M/12M不替換G05主表。
采用记录及研究结果Adoption record and findings採用記錄及研究結果
原基准保留;G05后续成为保留研究预测。H09-RDA框架归通用模型,G01在GOOGL谱系中作应用引用。Preserved benchmark, subsequently superseded by G05 as the retained research forecast. H09-RDA belongs in the general-model category, with this GOOGL application referenced here.原基準保留;G05後續成為保留研究預測。H09-RDA框架歸通用模型,G01在GOOGL譜系中作應用引用。
盈利与价格倍数分开定义;0M是市场锚,没有独立估计合理PE。Keep earnings and pricing multiples distinct. 0M is the market anchor; the model does not independently estimate a fair P/E.盈利與價格倍數分開定義;0M是市場錨,沒有獨立估計合理PE。
H09-RDA通用框架入口H09-RDA general frameworkH09-RDA通用框架入口G02 · 营业利润率趋势G02 · Operating-margin trendG02 · 營業利潤率趨勢 · 保留未采用版本Preserved, not adopted保留未採用版本
为了研究什么Research question為了研究什麼
检验近期利润率扩张能否缓解基本面预测低估。Test whether recent margin expansion can reduce underprediction.檢驗近期利潤率擴張能否緩解基本面預測低估。
具体方法与本代改变Method and changes具體方法與本代改變
其余G01输入与规则保持。用11个连续已知季度组成8个滚动TTM营业利润率,在logit空间拟合斜率;未来趋势增量按4季半衰期减少,每版用自身EPS0重新计算市场PE。Keep all other G01 assumptions. Fit a logit trend to eight rolling TTM margins formed from eleven known consecutive quarters. Future trend increments damp with a four-quarter half-life. Re-anchor P/E to this candidate’s own EPS0.其餘G01輸入與規則保持。用11個連續已知季度組成8個滾動TTM營業利潤率,在logit空間擬合斜率;未來趨勢增量按4季半衰期減少,每版用自身EPS0重新計算市場PE。
只修改未来营业利润率路径;没有用较高EPS乘旧版PE。Change only the future margin path; never multiply higher candidate EPS by the previous model’s P/E.只修改未來營業利潤率路徑;沒有用較高EPS乘舊版PE。
核算公式Calculation formulas核算公式
logit(m_j)=logit(m0)+β×Σ_(k=1..j−1)2^(−k/4) P_h = EPS_h,new × (P0 / EPS_0,new)
冻结参数Frozen parameters凍結參數
| 参数Parameter參數 | 取值与规则Value and rule取值與規則 |
|---|---|
| 利润率趋势样本Margin-trend observations利潤率趨勢樣本 | 8个滚动TTM8 rolling TTM observations8個滾動TTM |
| 利润率趋势半衰期Margin-trend half-life利潤率趨勢半衰期 | 4季度4 quarters4季度 |
本代计算值与日期Version outputs and dates本代計算值與日期
| 期限Horizon期限 | 原研究2026-10-06Original research: 2026-10-06原研究2026-10-06 | 原研究目标日期Original target date原研究目標日期 | 季度重建2026-07-23Quarterly reconstruction: 2026-07-23季度重建2026-07-23 | 季度目标日期Quarterly target date季度目標日期 |
|---|---|---|---|---|
| 0M | $347.68 | 2026-10-06 | $317.69 | 2026-07-23 |
| 6M | $370.60 | 2027-04-06 | $341.12 | 2027-01-23 |
| 12M | $386.63 | 2027-10-06 | $358.87 | 2027-07-23 |
上述两个输入截面分别保存,不能混为发布日的当前行情。0M是该截面的市场锚;本版未采用时,候选6M/12M不替换G05主表。These input snapshots are separately preserved, not publication-day market quotes. 0M is the snapshot’s market anchor. An unadopted candidate’s 6M/12M outputs do not replace the G05 main table.上述兩個輸入截面分別保存,不能混為發佈日的當前行情。0M是該截面的市場錨;本版未採用時,候選6M/12M不替換G05主表。
采用记录及研究结果Adoption record and findings採用記錄及研究結果
未采用。6M无改善,中位价格误差和经营盈利尾部风险恶化;保留原基准。Not adopted: 6M did not improve, while median price errors and operating-earnings tail errors worsened. Retain the original benchmark.未採用。6M無改善,中位價格誤差和經營盈利尾部風險惡化;保留原基準。
利润率历史趋势延伸并不自动提高价格预测准确性。Extrapolating a historical margin trend does not automatically improve price forecasts.利潤率歷史趨勢延伸並不自動提高價格預測準確性。
G03 · 按已兑现期末价格误差校准G03 · Mature terminal-price residual calibrationG03 · 按已兑現期末價格誤差校準 · 保留未采用版本Preserved, not adopted保留未採用版本
为了研究什么Research question為了研究什麼
分别检验6M/12M系统性低估能否通过过去已兑现残差缓解。Test whether already realized errors can reduce systematic 6M and 12M underprediction.分別檢驗6M/12M系統性低估能否通過過去已兑現殘差緩解。
具体方法与本代改变Method and changes具體方法與本代改變
保持G01盈利路径,仅用实际终点严格早于新起点的历史预测误差,裁剪、近期加权、向零收缩;6M和12M各自训练。分析师目标不参加拟合。Preserve G01 earnings. Clip, recency-weight and shrink historical forecast residuals whose actual terminal dates are strictly earlier than the new origin. Train the two horizons separately; analyst targets never enter fitting.保持G01盈利路徑,僅用實際終點嚴格早於新起點的歷史預測誤差,裁剪、近期加權、向零收縮;6M和12M各自訓練。分析師目標不參加擬合。
只增加价格误差修正;等价PE提高是校准的代数结果,非独立公允PE。Add only a price-residual correction. A higher equivalent P/E is an algebraic description, not an independently fair multiple.只增加價格誤差修正;等價PE提高是校準的代數結果,非獨立公允PE。
核算公式Calculation formulas核算公式
r_s=clip(ln(P_actual,s/P_raw,s),±ln1.5); w_s=2^(−age_days/365.25) Kish=(Σw)^2/Σw²; a=Kish/h; λ=a/(a+2) c_h=clip(λ×weighted_mean(r),±ln1.5); P_h=P_G01,h×exp(c_h)
冻结参数Frozen parameters凍結參數
| 参数Parameter參數 | 取值与规则Value and rule取值與規則 |
|---|---|
| 残差裁剪Residual clipping殘差裁剪 | ±ln1.5±ln1.5±ln1.5 |
| 近期权重半衰期Recency half-life近期權重半衰期 | 365.25天365.25 days365.25天 |
| 零先验强度Zero-prior strength零先驗強度 | 222 |
| 最低训练覆盖Minimum training coverage最低訓練覆蓋 | 12成熟标签、2非重叠段12 mature labels, 2 nonoverlapping episodes12成熟標籤、2非重疊段 |
本代计算值与日期Version outputs and dates本代計算值與日期
| 期限Horizon期限 | 原研究2026-10-06Original research: 2026-10-06原研究2026-10-06 | 原研究目标日期Original target date原研究目標日期 | 季度重建2026-07-23Quarterly reconstruction: 2026-07-23季度重建2026-07-23 | 季度目标日期Quarterly target date季度目標日期 |
|---|---|---|---|---|
| 0M | $347.68 | 2026-10-06 | $317.69 | 2026-07-23 |
| 6M | $405.62 | 2027-04-06 | $377.24 | 2027-01-23 |
| 12M | $446.30 | 2027-10-06 | $408.68 | 2027-07-23 |
上述两个输入截面分别保存,不能混为发布日的当前行情。0M是该截面的市场锚;本版未采用时,候选6M/12M不替换G05主表。These input snapshots are separately preserved, not publication-day market quotes. 0M is the snapshot’s market anchor. An unadopted candidate’s 6M/12M outputs do not replace the G05 main table.上述兩個輸入截面分別保存,不能混為發佈日的當前行情。0M是該截面的市場錨;本版未採用時,候選6M/12M不替換G05主表。
采用记录及研究结果Adoption record and findings採用記錄及研究結果
原三年期末价格门槛未通过;6M中位误差变大,12M中位改善4.64%低于当时5%门槛。五年季度新标准下12M排名第一,未自动晋升。Rejected under the original three-year terminal-price gates: 6M median error worsened; 12M median improvement of 4.64% was below the then required 5%. It ranks first for 12M in the later five-year quarterly touch study, without automatic promotion.原三年期末價格門檻未通過;6M中位誤差變大,12M中位改善4.64%低於當時5%門檻。五年季度新標準下12M排名第一,未自動晉升。
平均低估可缓解,但典型误差与不同市场阶段仍需控制。Average underprediction can shrink while typical errors and regime differences remain.平均低估可緩解,但典型誤差與不同市場階段仍需控制。
G04 · 12M均线状态校准G04 · 12M moving-average state calibrationG04 · 12M均線狀態校準 · 保留未采用版本Preserved, not adopted保留未採用版本
为了研究什么Research question為了研究什麼
检验市场趋势状态能否解释12M整体校准之外的残差。Test whether trend state explains residuals beyond pooled 12M calibration.檢驗市場趨勢狀態能否解釋12M整體校準之外的殘差。
具体方法与本代改变Method and changes具體方法與本代改變
0M/6M精确保持G01。12M用起点截至当日200个交易日收盘均线,分UP/DOWN,在全体与同状态残差之间做两层收缩。Keep G01 exactly at 0M and 6M. At 12M, classify the origin using the last 200 trading-day closes and shrink same-state residual information toward the full pool.0M/6M精確保持G01。12M用起點截至當日200個交易日收盤均線,分UP/DOWN,在全體與同狀態殘差之間做兩層收縮。
增加12M状态,沿用G03裁剪、近期加权与收缩;未知训练状态仍留全体池。Add a 12M state while preserving G03 clipping, recency weights and shrinkage; unknown training states remain in the overall pool.增加12M狀態,沿用G03裁剪、近期加權與收縮;未知訓練狀態仍留全體池。
核算公式Calculation formulas核算公式
UP if P0≥SMA200; DOWN otherwise c4=clip[λ_G(μ_G+λ_S(μ_S−μ_G)),±ln1.5] P_12=P_G01,12×exp(c4); P_6=P_G01,6
冻结参数Frozen parameters凍結參數
| 参数Parameter參數 | 取值与规则Value and rule取值與規則 |
|---|---|
| 均线Moving average均線 | 200交易日200 trading days200交易日 |
| 同状态最低覆盖Minimum same-state coverage同狀態最低覆蓋 | 12成熟标签、2非重叠段12 mature labels, 2 nonoverlapping episodes12成熟標籤、2非重疊段 |
| 稀疏状态回退Sparse-state fallback稀疏狀態回退 | G03全池校准G03 pooled calibrationG03全池校準 |
本代计算值与日期Version outputs and dates本代計算值與日期
| 期限Horizon期限 | 原研究2026-10-06Original research: 2026-10-06原研究2026-10-06 | 原研究目标日期Original target date原研究目標日期 | 季度重建2026-07-23Quarterly reconstruction: 2026-07-23季度重建2026-07-23 | 季度目标日期Quarterly target date季度目標日期 |
|---|---|---|---|---|
| 0M | $347.68 | 2026-10-06 | $317.69 | 2026-07-23 |
| 6M | $368.45 | 2027-04-06 | $338.82 | 2027-01-23 |
| 12M | $444.68 | 2027-10-06 | $412.53 | 2027-07-23 |
上述两个输入截面分别保存,不能混为发布日的当前行情。0M是该截面的市场锚;本版未采用时,候选6M/12M不替换G05主表。These input snapshots are separately preserved, not publication-day market quotes. 0M is the snapshot’s market anchor. An unadopted candidate’s 6M/12M outputs do not replace the G05 main table.上述兩個輸入截面分別保存,不能混為發佈日的當前行情。0M是該截面的市場錨;本版未採用時,候選6M/12M不替換G05主表。
采用记录及研究结果Adoption record and findings採用記錄及研究結果
原三年期末误差采用测试拒绝;五年财报后季度触达优先综合排名第一,G05第二,两者触达数相同、综合峰值差仅差约0.067个百分点。未自动替换G05。Rejected under the original terminal-error screen. First in the later five-year post-earnings quarterly touch ranking, ahead of G05 by only about 0.067 percentage points of combined peak gap; hit counts are identical. G05 is not automatically replaced.原三年期末誤差採用測試拒絕;五年財報後季度觸達優先綜合排名第一,G05第二,兩者觸達數相同、綜合峯值差僅差約0.067個百分點。未自動替換G05。
研究排序取决于评价目标;微小历史领先不等于显著或前瞻优势。Historical rankings depend on the objective. A tiny lead is not evidence of significant or prospective superiority.研究排序取決於評價目標;微小歷史領先不等於顯著或前瞻優勢。
G05 · 12M增长持续期与自身校准G05 · 12M growth duration and own-path calibrationG05 · 12M增長持續期與自身校準 · 保留研究模型Retained research model保留研究模型
为了研究什么Research question為了研究什麼
检验更长的增长持续期能否改善12M盈利与价格低估。Test whether longer growth persistence improves 12M earnings and price underprediction.檢驗更長的增長持續期能否改善12M盈利與價格低估。
具体方法与本代改变Method and changes具體方法與本代改變
0M/6M保持G01的4季增长路径;仅12M使用8季增长半衰期F8,并以F8自身EPS0重锚、F8自己的已兑现残差重训。Keep the original four-quarter path at 0M and 6M. Only 12M uses the F8 eight-quarter growth half-life, its own EPS0 / P/E anchor, and calibration refitted to its own already realized residuals.0M/6M保持G01的4季增長路徑;僅12M使用8季增長半衰期F8,並以F8自身EPS0重錨、F8自己的已兑現殘差重訓。
12M盈利持续期与自身锚改变;不是直接把G03校准乘更高EPS。Change 12M earnings duration and its own anchor; do not apply G03 calibration to higher earnings.12M盈利持續期與自身錨改變;不是直接把G03校準乘更高EPS。
核算公式Calculation formulas核算公式
D_j=D_(j−1)×exp(s_j−s_(j−1)+g0×2^(−(j−1)/8)), j≥2 PE8=P0/EPS0,F8; P_12=EPS12,F8×PE8×exp(c8) P_0=P0; P_6=P_G01,6
冻结参数Frozen parameters凍結參數
| 参数Parameter參數 | 取值与规则Value and rule取值與規則 |
|---|---|
| 6M增长半衰期6M growth half-life6M增長半衰期 | 4季度4 quarters4季度 |
| 12M增长半衰期12M growth half-life12M增長半衰期 | 8季度8 quarters8季度 |
| 12M校准12M calibration12M校準 | F8自身残差、同G03裁剪/权重/先验F8 own residuals, G03 clipping / weights / priorF8自身殘差、同G03裁剪/權重/先驗 |
本代计算值与日期Version outputs and dates本代計算值與日期
| 期限Horizon期限 | 原研究2026-10-06Original research: 2026-10-06原研究2026-10-06 | 原研究目标日期Original target date原研究目標日期 | 季度重建2026-07-23Quarterly reconstruction: 2026-07-23季度重建2026-07-23 | 季度目标日期Quarterly target date季度目標日期 |
|---|---|---|---|---|
| 0M | $347.68 | 2026-10-06 | $317.69 | 2026-07-23 |
| 6M | $368.45 | 2027-04-06 | $338.82 | 2027-01-23 |
| 12M | $464.42 | 2027-10-06 | $423.88 | 2027-07-23 |
上述两个输入截面分别保存,不能混为发布日的当前行情。0M是该截面的市场锚;本版未采用时,候选6M/12M不替换G05主表。These input snapshots are separately preserved, not publication-day market quotes. 0M is the snapshot’s market anchor. An unadopted candidate’s 6M/12M outputs do not replace the G05 main table.上述兩個輸入截面分別保存,不能混為發佈日的當前行情。0M是該截面的市場錨;本版未採用時,候選6M/12M不替換G05主表。
采用记录及研究结果Adoption record and findings採用記錄及研究結果
保留研究预测。按首算前声明的新12M优先筛选通过;未获独立前瞻验证或正式Champion认证。G06–G15没有正式替换整套G05。Retained research forecast under a new 12M-first screen declared before calculation. No independent prospective validation or formal Champion certification. G06–G15 have not formally replaced the complete G05 policy.保留研究預測。按首算前聲明的新12M優先篩選通過;未獲獨立前瞻驗證或正式Champion認證。G06–G15沒有正式替換整套G05。
原三年12M MAPE降至25.8435%,但相对G03仅改善约0.49个百分点;定价误差仍大。Original three-year 12M MAPE fell to 25.8435%, only about 0.49 percentage points better than G03. Pricing errors remain large.原三年12M MAPE降至25.8435%,但相對G03僅改善約0.49個百分點;定價誤差仍大。
G06 · 6M统一F8路径校准G06 · F8 path with separate 6M calibrationG06 · 6M統一F8路徑校準 · 保留未采用版本Preserved, not adopted保留未採用版本
为了研究什么Research question為了研究什麼
用统一8季增长路径优化6M,同时精确保留G05的12M。Improve 6M on one eight-quarter earnings path while preserving G05 at 12M.用統一8季增長路徑優化6M,同時精確保留G05的12M。
具体方法与本代改变Method and changes具體方法與本代改變
0M/6M/12M均以F8盈利为参考;新增F8自身6M残差校准。6M的重叠折扣Kish/6,12M仍Kish/12,各自训练,不折半12M修正。Use F8 earnings references at all three horizons. Add F8’s own 6M residual calibration with Kish/6; 12M retains Kish/12. Train separately rather than halving a 12M correction.0M/6M/12M均以F8盈利為參考;新增F8自身6M殘差校準。6M的重疊折扣Kish/6,12M仍Kish/12,各自訓練,不折半12M修正。
6M盈利路径与校准一起改变;没有拼接未通过的6M到主模型。Change both the 6M earnings path and its correction; never splice a rejected 6M into the main model.6M盈利路徑與校準一起改變;沒有拼接未通過的6M到主模型。
核算公式Calculation formulas核算公式
P6=EPS6,F8×PE8×exp(c6); P12=G05_12
冻结参数Frozen parameters凍結參數
| 参数Parameter參數 | 取值与规则Value and rule取值與規則 |
|---|---|
| 增长半衰期Growth half-life增長半衰期 | 8季度8 quarters8季度 |
| 6M校准6M calibration6M校準 | 独立F8残差;Kish/6Separate F8 residuals; Kish/6獨立F8殘差;Kish/6 |
本代计算值与日期Version outputs and dates本代計算值與日期
| 期限Horizon期限 | 原研究2026-10-06Original research: 2026-10-06原研究2026-10-06 | 原研究目标日期Original target date原研究目標日期 | 季度重建2026-07-23Quarterly reconstruction: 2026-07-23季度重建2026-07-23 | 季度目标日期Quarterly target date季度目標日期 |
|---|---|---|---|---|
| 0M | $347.68 | 2026-10-06 | $317.69 | 2026-07-23 |
| 6M | $411.29 | 2027-04-06 | $381.86 | 2027-01-23 |
| 12M | $464.42 | 2027-10-06 | $423.88 | 2027-07-23 |
上述两个输入截面分别保存,不能混为发布日的当前行情。0M是该截面的市场锚;本版未采用时,候选6M/12M不替换G05主表。These input snapshots are separately preserved, not publication-day market quotes. 0M is the snapshot’s market anchor. An unadopted candidate’s 6M/12M outputs do not replace the G05 main table.上述兩個輸入截面分別保存,不能混為發佈日的當前行情。0M是該截面的市場錨;本版未採用時,候選6M/12M不替換G05主表。
采用记录及研究结果Adoption record and findings採用記錄及研究結果
未采用;6M MAPE改善至15.9947%,但中位误差由12.4568%升至14.5681%,2024起点表现恶化。保留G05。Not adopted: 6M MAPE improved to 15.9947%, but median error rose from 12.4568% to 14.5681%, and 2024-origin results worsened. Retain G05.未採用;6M MAPE改善至15.9947%,但中位誤差由12.4568%升至14.5681%,2024起點表現惡化。保留G05。
平均误差改善不足以抵消典型误差和阶段风险。A lower mean error does not offset worse typical errors and regime risk.平均誤差改善不足以抵消典型誤差和階段風險。
G07 · 6M三专家动态组合G07 · Dynamic 6M three-expert mixtureG07 · 6M三專家動態組合 · 保留未采用版本Preserved, not adopted保留未採用版本
为了研究什么Research question為了研究什麼
检验组合多条相关预测能否减少6M校准过度。Test whether mixing correlated forecasts can reduce excessive 6M calibration.檢驗組合多條相關預測能否減少6M校準過度。
具体方法与本代改变Method and changes具體方法與本代改變
将G01、F8原价、G06校准价按历史已兑现裁剪绝对对数损失动态加权,在价格对数空间几何平均;12M精确保留G05。各历史G06专家在自己的起点内再满足成熟边界。Geometrically combine G01, raw F8 and calibrated G06 prices using weights from past mature clipped absolute-log losses. Preserve G05 at 12M. Each historical G06 expert also respects its own earlier maturity boundary.將G01、F8原價、G06校準價按歷史已兑現裁剪絕對對數損失動態加權,在價格對數空間幾何平均;12M精確保留G05。各歷史G06專家在自己的起點內再滿足成熟邊界。
组合的是价格,不是EPS;三个专家相关且冷启动可能重合。Combine prices, not EPS. The experts are correlated and may coincide during cold starts.組合的是價格,不是EPS;三個專家相關且冷啓動可能重合。
核算公式Calculation formulas核算公式
L_k=weighted_mean[min(abs(ln(P_k/P_actual)),ln1.5)] weight_k ∝ prior_k×exp(−λL_k/0.05) P7_6=exp(Σweight_k×ln(P_k)); P7_12=G05_12
冻结参数Frozen parameters凍結參數
| 参数Parameter參數 | 取值与规则Value and rule取值與規則 |
|---|---|
| 专家Experts專家 | G01 / F8原价 / G06G01 / raw F8 / G06G01 / F8原價 / G06 |
| 先验权重Prior weights先驗權重 | 50% / 25% / 25%50% / 25% / 25%50% / 25% / 25% |
| 温度Temperature温度 | 0.050.050.05 |
| 冷启动Cold start冷啓動 | 固定先验几何组合Fixed-prior geometric mixture固定先驗幾何組合 |
本代计算值与日期Version outputs and dates本代計算值與日期
| 期限Horizon期限 | 原研究2026-10-06Original research: 2026-10-06原研究2026-10-06 | 原研究目标日期Original target date原研究目標日期 | 季度重建2026-07-23Quarterly reconstruction: 2026-07-23季度重建2026-07-23 | 季度目标日期Quarterly target date季度目標日期 |
|---|---|---|---|---|
| 0M | $347.68 | 2026-10-06 | $317.69 | 2026-07-23 |
| 6M | $381.16 | 2027-04-06 | $351.14 | 2027-01-23 |
| 12M | $464.42 | 2027-10-06 | $423.88 | 2027-07-23 |
上述两个输入截面分别保存,不能混为发布日的当前行情。0M是该截面的市场锚;本版未采用时,候选6M/12M不替换G05主表。These input snapshots are separately preserved, not publication-day market quotes. 0M is the snapshot’s market anchor. An unadopted candidate’s 6M/12M outputs do not replace the G05 main table.上述兩個輸入截面分別保存,不能混為發佈日的當前行情。0M是該截面的市場錨;本版未採用時,候選6M/12M不替換G05主表。
采用记录及研究结果Adoption record and findings採用記錄及研究結果
未采用;6M MAPE约16.12%,中位误差由12.46%升至12.82%,未通过筛选。保留G05。Not adopted: 6M MAPE was about 16.12%, while median error rose from 12.46% to 12.82%. Retain G05.未採用;6M MAPE約16.12%,中位誤差由12.46%升至12.82%,未通過篩選。保留G05。
动态权重对当前价格贡献很小,增加组合复杂度未形成优势。Dynamic weighting made a small current contribution and did not establish an advantage.動態權重對當前價格貢獻很小,增加組合複雜度未形成優勢。
G08 · 趋势与预期盈利增长双状态岭校准G08 · Trend and expected-earnings ridge statesG08 · 趨勢與預期盈利增長雙狀態嶺校準 · 保留未采用版本Preserved, not adopted保留未採用版本
为了研究什么Research question為了研究什麼
同时优化6M/12M,检验起点走势及模型盈利成长的残差解释力。Optimize both horizons and test origin price trend plus forecast earnings growth as residual predictors.同時優化6M/12M,檢驗起點走勢及模型盈利成長的殘差解釋力。
具体方法与本代改变Method and changes具體方法與本代改變
固定F8盈利,状态为过去6个月价格对数回报和同起点预测NOPAT12/NOPAT0对数增长;各除0.25并裁剪±2。两期限独立拟合中心化平方损失与强岭先验。Hold F8 earnings. Use past six-month log price return and same-origin forecast log NOPAT12/NOPAT0, scaled by 0.25 and clipped at ±2. Fit centered squared losses with strong ridge priors independently for each horizon.固定F8盈利,狀態為過去6個月價格對數回報和同起點預測NOPAT12/NOPAT0對數增長;各除0.25並裁剪±2。兩期限獨立擬閤中心化平方損失與強嶺先驗。
新增两项起点可知状态,无国债或联邦基金利率输入。Add two states known at the origin; neither Treasury nor federal-funds rates enter this version.新增兩項起點可知狀態,無國債或聯邦基金利率輸入。
核算公式Calculation formulas核算公式
z=clip([past_6M_log_return,ln(NOPAT12/NOPAT0)]/0.25,±2) min aΣv[y−α−β·(z−μz)]²+2α²+8||β||² P_h=rawF8_h×exp(clip(α+β·(z_current−μz),±ln1.5))
冻结参数Frozen parameters凍結參數
| 参数Parameter參數 | 取值与规则Value and rule取值與規則 |
|---|---|
| 状态States狀態 | 过去6M回报、预测盈利增长Past 6M return, forecast earnings growth過去6M回報、預測盈利增長 |
| 特征尺度及裁剪Feature scale and clipping特徵尺度及裁剪 | 0.25对数单位、±20.25 log units, ±20.25對數單位、±2 |
| 截距/斜率先验Intercept / slope priors截距/斜率先驗 | 2 / 82 / 82 / 8 |
| 损失Loss損失 | 平方残差Squared residuals平方殘差 |
本代计算值与日期Version outputs and dates本代計算值與日期
| 期限Horizon期限 | 原研究2026-10-06Original research: 2026-10-06原研究2026-10-06 | 原研究目标日期Original target date原研究目標日期 | 季度重建2026-07-23Quarterly reconstruction: 2026-07-23季度重建2026-07-23 | 季度目标日期Quarterly target date季度目標日期 |
|---|---|---|---|---|
| 0M | $347.68 | 2026-10-06 | $317.69 | 2026-07-23 |
| 6M | $411.87 | 2027-04-06 | $384.39 | 2027-01-23 |
| 12M | $462.23 | 2027-10-06 | $425.74 | 2027-07-23 |
上述两个输入截面分别保存,不能混为发布日的当前行情。0M是该截面的市场锚;本版未采用时,候选6M/12M不替换G05主表。These input snapshots are separately preserved, not publication-day market quotes. 0M is the snapshot’s market anchor. An unadopted candidate’s 6M/12M outputs do not replace the G05 main table.上述兩個輸入截面分別保存,不能混為發佈日的當前行情。0M是該截面的市場錨;本版未採用時,候選6M/12M不替換G05主表。
采用记录及研究结果Adoption record and findings採用記錄及研究結果
联合未采用;12M门槛通过,6M中位误差相对G01恶化3.73%,保留整套G05。Joint rejection: 12M passed, but the 6M median error worsened by 3.73% relative to G01. Retain the whole G05 model.聯合未採用;12M門檻通過,6M中位誤差相對G01惡化3.73%,保留整套G05。
新增状态当前贡献小,多数修正来自整体残差截距。Current state effects were small; most correction came from the pooled residual intercept.新增狀態當前貢獻小,多數修正來自整體殘差截距。
G09 · 双状态Huber稳健校准G09 · Huber calibration of two statesG09 · 雙狀態Huber穩健校準 · 保留未采用版本Preserved, not adopted保留未採用版本
为了研究什么Research question為了研究什麼
检验减少大残差影响能否改善典型误差且保留12M表现。Test whether reducing the influence of large residuals improves typical errors without losing 12M performance.檢驗減少大殘差影響能否改善典型誤差且保留12M表現。
具体方法与本代改变Method and changes具體方法與本代改變
沿G08的F8盈利、两项状态、裁剪和近期权重,只将拟合平方损失换成固定Huber;保留全部评估行。Preserve G08 F8 earnings, states, clipping and recency weights; change only the fitting loss to a fixed Huber loss and keep every evaluation row.沿G08的F8盈利、兩項狀態、裁剪和近期權重,只將擬合平方損失換成固定Huber;保留全部評估行。
唯一模型改变为拟合损失,未搜索δ或选择更好控制。Only the fitting loss changes; do not search delta or adopt a better-scoring control.唯一模型改變為擬合損失,未搜索δ或選擇更好控制。
核算公式Calculation formulas核算公式
δ=ln1.20; Lδ(r)=r² if |r|≤δ, otherwise 2δ|r|−δ² min aΣvLδ(y−α−β·(z−μz))+2α²+8||β||² P_h=rawF8_h×exp(clipped_correction)
冻结参数Frozen parameters凍結參數
| 参数Parameter參數 | 取值与规则Value and rule取值與規則 |
|---|---|
| Huber阈值Huber deltaHuber閾值 | ln1.20ln1.20ln1.20 |
| 状态与先验States and priors狀態與先驗 | 同G08Same as G08同G08 |
本代计算值与日期Version outputs and dates本代計算值與日期
| 期限Horizon期限 | 原研究2026-10-06Original research: 2026-10-06原研究2026-10-06 | 原研究目标日期Original target date原研究目標日期 | 季度重建2026-07-23Quarterly reconstruction: 2026-07-23季度重建2026-07-23 | 季度目标日期Quarterly target date季度目標日期 |
|---|---|---|---|---|
| 0M | $347.68 | 2026-10-06 | $317.69 | 2026-07-23 |
| 6M | $406.88 | 2027-04-06 | $378.69 | 2027-01-23 |
| 12M | $456.34 | 2027-10-06 | $417.70 | 2027-07-23 |
上述两个输入截面分别保存,不能混为发布日的当前行情。0M是该截面的市场锚;本版未采用时,候选6M/12M不替换G05主表。These input snapshots are separately preserved, not publication-day market quotes. 0M is the snapshot’s market anchor. An unadopted candidate’s 6M/12M outputs do not replace the G05 main table.上述兩個輸入截面分別保存,不能混為發佈日的當前行情。0M是該截面的市場錨;本版未採用時,候選6M/12M不替換G05主表。
采用记录及研究结果Adoption record and findings採用記錄及研究結果
未采用;12M通过,但6M MAPE16.1041%没有严格优于G03的15.9604%,联合失败。保留G05。Not adopted: 12M passed, but 6M MAPE of 16.1041% did not strictly beat G03 at 15.9604%. Joint selection failed; retain G05.未採用;12M通過,但6M MAPE16.1041%沒有嚴格優於G03的15.9604%,聯合失敗。保留G05。
稳健拟合可以改善典型误差,却可能削弱纠正持续低估的信息。Robust fitting can improve typical errors while weakening information needed to correct persistent underprediction.穩健擬合可以改善典型誤差,卻可能削弱糾正持續低估的信息。
G10 · 分部成长与资本回报SOTP-DCFG10 · Segment growth and reinvestment SOTP-DCFG10 · 分部成長與資本回報SOTP-DCF · 保留未采用版本Preserved, not adopted保留未採用版本
为了研究什么Research question為了研究什麼
从业务成长、资本投入与风险推导价值与经营倍数,检验分部合并估值。Derive conditional value and operating multiples from business growth, capital investment and risk.從業務成長、資本投入與風險推導價值與經營倍數,檢驗分部合併估值。
具体方法与本代改变Method and changes具體方法與本代改變
Google Services、Cloud分别预测利润与再投资;单列公司成本及Other Bets损失。一个绝对40季现金流路径随期限推进,不重新启动增长。经营EV加账面金融权益桥与已留存预测现金后除普通股。Forecast Services and Cloud profit and reinvestment separately; deduct corporate costs and Other Bets losses. Advance along one absolute forty-quarter cash-flow path without restarting growth. Add the book-value financing bridge and retained modeled cash, then divide by common shares.Google Services、Cloud分別預測利潤與再投資;單列公司成本及Other Bets損失。一個絕對40季現金流路徑隨期限推進,不重新啓動增長。經營EV加賬面金融權益橋與已留存預測現金後除普通股。
不依市场PE锚或价格残差;0M为条件DCF价值,性质与其他版本0M不同。业务分部经营倍数是EV/NOPAT,不是官方分部PE。No market P/E anchor or residual correction. 0M is a conditional DCF value, unlike other versions. Segment EV/NOPAT multiples are not reported segment P/E.不依市場PE錨或價格殘差;0M為條件DCF價值,性質與其他版本0M不同。業務分部經營倍數是EV/NOPAT,不是官方分部PE。
核算公式Calculation formulas核算公式
k=last_pre_origin_10Y+0.05; FCFF=OperatingProfit×0.85−NetReinvestment EV=PV(FCFF)+PV(TerminalValue) EV/forward NOPAT=(1−g/ROIC)/(k−g); terminal ROIC=k Value/share=(OperatingEV+FinancialAssets−FinancialDebt−PreferredClaims−NCI+RetainedForecastCash)/CommonShares
冻结参数Frozen parameters凍結參數
| 参数Parameter參數 | 取值与规则Value and rule取值與規則 |
|---|---|
| 显式预测Explicit forecast顯式預測 | 40季度40 quarters40季度 |
| 增长/利润率/资本强度半衰期Growth / margin / capital-intensity half-life增長/利潤率/資本強度半衰期 | 各8季度8 quarters each各8季度 |
| 终值名义增长Terminal nominal growth終值名義增長 | 3%3%3% |
| 终值ROICTerminal ROIC終值ROIC | 等于kEqual to k等於k |
| 经营风险溢价Operating risk premium經營風險溢價 | 5个百分点5 percentage points5個百分點 |
| 税率代理Tax proxy税率代理 | 15%15%15% |
| Other Bets资金消耗Other Bets fundingOther Bets資金消耗 | 20季度后停止;正期权价值0Stop after 20 quarters; positive option value 020季度後停止;正期權價值0 |
本代计算值与日期Version outputs and dates本代計算值與日期
| 期限Horizon期限 | 原研究2026-10-06Original research: 2026-10-06原研究2026-10-06 | 原研究目标日期Original target date原研究目標日期 | 季度重建2026-07-23Quarterly reconstruction: 2026-07-23季度重建2026-07-23 | 季度目标日期Quarterly target date季度目標日期 |
|---|---|---|---|---|
| 0M | $175.01 | 2026-10-06 | $180.35 | 2026-07-23 |
| 6M | $182.60 | 2027-04-06 | $188.23 | 2027-01-23 |
| 12M | $190.52 | 2027-10-06 | $196.43 | 2027-07-23 |
G10的0M为条件DCF价值,6M/12M沿同一现金流路径推进;不以市价锚生成。其他版本0M均为对应起点市场锚,两列分别是独立截面。G10 0M is a conditional DCF value; 6M/12M advance along the same cash-flow path and are not generated from a market anchor. Other versions use their origin market price at 0M. The two columns are separate snapshots.G10的0M為條件DCF價值,6M/12M沿同一現金流路徑推進;不以市價錨生成。其他版本0M均為對應起點市場錨,兩列分別是獨立截面。
采用记录及研究结果Adoption record and findings採用記錄及研究結果
未采用;三年6M/12M期末MAPE46.18%/50.82%,明显差于G05。保守假设提供解释,不足以替代价格预测。Not adopted: original three-year terminal MAPE of 46.18% / 50.82% was materially worse than G05. A conservative explanatory scenario does not replace a price forecast.未採用;三年6M/12M期末MAPE46.18%/50.82%,明顯差於G05。保守假設提供解釋,不足以替代價格預測。
分部再投资/ROIC未知、金融权益桥是账面代理;不得用被拒绝的低估情景证明市场高估。Segment reinvestment and ROIC are unobserved; the financing bridge uses carrying-value proxies. Rejection of a low valuation scenario does not prove market overvaluation.分部再投資/ROIC未知、金融權益橋是賬面代理;不得用被拒絕的低估情景證明市場高估。
G11 · 固定G05盈利的国债利率状态G11 · Treasury-rate state with fixed G05 earningsG11 · 固定G05盈利的國債利率狀態 · 保留未采用版本Preserved, not adopted保留未採用版本
为了研究什么Research question為了研究什麼
隔离倍数部分,检验已知10Y收益率能否解释G05剩余价格残差。Isolate multiples and test whether known 10Y yields explain G05’s remaining price residuals.隔離倍數部分,檢驗已知10Y收益率能否解釋G05剩餘價格殘差。
具体方法与本代改变Method and changes具體方法與本代改變
G05全部盈利、股份、0M/6M HL4和12M HL8及自身校准保持;新增单截距、单利率斜率的平方岭修正。收益率取起点前最后一个真实官方观察,百分点为单位。Preserve all G05 earnings, shares, HL4 at 0M/6M and HL8 plus own calibration at 12M. Add a squared-ridge intercept and one yield slope. Take the last official observation strictly before origin, measured in percentage points.G05全部盈利、股份、0M/6M HL4和12M HL8及自身校準保持;新增單截距、單利率斜率的平方嶺修正。收益率取起點前最後一個真實官方觀察,百分點為單位。
只修同一盈利对应的等价倍数;不假定未来降息,不重复机械扣当前高利率。Change only equivalent pricing multiples for the same earnings. Do not assume future rate cuts or mechanically deduct current high rates twice.只修同一盈利對應的等價倍數;不假定未來降息,不重複機械扣當前高利率。
核算公式Calculation formulas核算公式
y=clip(ln(P_actual/G05_own_origin_target),±ln1.5) min aΣv[y−α−β(yield−μyield)]²+2α²+8β² P11_h=G05_h×exp(clip(α+β(yield_current−μyield),±ln1.5))
冻结参数Frozen parameters凍結參數
| 参数Parameter參數 | 取值与规则Value and rule取值與規則 |
|---|---|
| 利率状态Yield state利率狀態 | 起点前官方10Y、百分点Pre-origin official 10Y, percentage points起點前官方10Y、百分點 |
| 截距/斜率先验Intercept / slope priors截距/斜率先驗 | 2 / 82 / 82 / 8 |
| 近期权重半衰期Recency half-life近期權重半衰期 | 365.25天365.25 days365.25天 |
本代计算值与日期Version outputs and dates本代計算值與日期
| 期限Horizon期限 | 原研究2026-10-06Original research: 2026-10-06原研究2026-10-06 | 原研究目标日期Original target date原研究目標日期 | 季度重建2026-07-23Quarterly reconstruction: 2026-07-23季度重建2026-07-23 | 季度目标日期Quarterly target date季度目標日期 |
|---|---|---|---|---|
| 0M | $347.68 | 2026-10-06 | $317.69 | 2026-07-23 |
| 6M | $413.91 | 2027-04-06 | $382.16 | 2027-01-23 |
| 12M | $545.04 | 2027-10-06 | $487.48 | 2027-07-23 |
上述两个输入截面分别保存,不能混为发布日的当前行情。0M是该截面的市场锚;本版未采用时,候选6M/12M不替换G05主表。These input snapshots are separately preserved, not publication-day market quotes. 0M is the snapshot’s market anchor. An unadopted candidate’s 6M/12M outputs do not replace the G05 main table.上述兩個輸入截面分別保存,不能混為發佈日的當前行情。0M是該截面的市場錨;本版未採用時,候選6M/12M不替換G05主表。
采用记录及研究结果Adoption record and findings採用記錄及研究結果
12M通过、6M中位误差恶化导致联合失败;保留G05。12M passed; worse 6M median error caused joint rejection. Retain G05.12M通過、6M中位誤差惡化導致聯合失敗;保留G05。
新增利率相对零斜率控制带来的MAPE改善仅6M0.06659、12M0.04455个百分点;大部分改善不能归因利率。Against the zero-slope control, the yield state improved MAPE by only 0.06659 / 0.04455 percentage points. Most mean improvement cannot be attributed to rates.新增利率相對零斜率控制帶來的MAPE改善僅6M0.06659、12M0.04455個百分點;大部分改善不能歸因利率。
G12 · 利率状态绝对对数损失G12 · Absolute-log loss for the yield stateG12 · 利率狀態絕對對數損失 · 保留未采用版本Preserved, not adopted保留未採用版本
为了研究什么Research question為了研究什麼
检验L1拟合能否修复G11典型误差,同时保持12M。Test whether L1 fitting can repair G11’s typical errors while preserving 12M.檢驗L1擬合能否修復G11典型誤差,同時保持12M。
具体方法与本代改变Method and changes具體方法與本代改變
盈利、利率状态和先验数值保持G11,唯一改变为绝对对数残差;用凸优化原始/对偶证书核验。Keep G11 earnings, yield state and numeric priors; change only the residual loss to absolute log error. Verify the convex solution with primal / dual certificates.盈利、利率狀態和先驗數值保持G11,唯一改變為絕對對數殘差;用凸優化原始/對偶證書核驗。
先验数值相同但损失尺度改变,不等于有效正则强度完全相同。The same numeric priors do not imply identical effective regularization after changing the loss scale.先驗數值相同但損失尺度改變,不等於有效正則強度完全相同。
核算公式Calculation formulas核算公式
min aΣv|y−α−β(z−μz)|+2α²+8β² P12_h=G05_h×exp(clip(α+β(z_current−μz),±ln1.5))
冻结参数Frozen parameters凍結參數
| 参数Parameter參數 | 取值与规则Value and rule取值與規則 |
|---|---|
| 损失Loss損失 | L1绝对对数残差L1 absolute-log residualsL1絕對對數殘差 |
| 其余输入及先验Other inputs and priors其餘輸入及先驗 | 同G11;先验2 / 8Same as G11; priors 2 / 8同G11;先驗2 / 8 |
本代计算值与日期Version outputs and dates本代計算值與日期
| 期限Horizon期限 | 原研究2026-10-06Original research: 2026-10-06原研究2026-10-06 | 原研究目标日期Original target date原研究目標日期 | 季度重建2026-07-23Quarterly reconstruction: 2026-07-23季度重建2026-07-23 | 季度目标日期Quarterly target date季度目標日期 |
|---|---|---|---|---|
| 0M | $347.68 | 2026-10-06 | $317.69 | 2026-07-23 |
| 6M | $412.60 | 2027-04-06 | $388.33 | 2027-01-23 |
| 12M | $603.73 | 2027-10-06 | $541.26 | 2027-07-23 |
上述两个输入截面分别保存,不能混为发布日的当前行情。0M是该截面的市场锚;本版未采用时,候选6M/12M不替换G05主表。These input snapshots are separately preserved, not publication-day market quotes. 0M is the snapshot’s market anchor. An unadopted candidate’s 6M/12M outputs do not replace the G05 main table.上述兩個輸入截面分別保存,不能混為發佈日的當前行情。0M是該截面的市場錨;本版未採用時,候選6M/12M不替換G05主表。
采用记录及研究结果Adoption record and findings採用記錄及研究結果
12M通过;6M MAPE未严格优于G03且中位误差高于G01,联合未采用。保留G05。12M passed. 6M failed to beat G03 MAPE and retained higher median error than G01; joint rejection. Retain G05.12M通過;6M MAPE未嚴格優於G03且中位誤差高於G01,聯合未採用。保留G05。
12M平均/中位改善,P90却高于G11;不存在全面胜出。12M mean and median errors improved, but P90 worsened relative to G11; improvement was not universal.12M平均/中位改善,P90卻高於G11;不存在全面勝出。
G13 · 固定盈利的现金转化状态G13 · Cash-conversion state with fixed earningsG13 · 固定盈利的現金轉化狀態 · 保留未采用版本Preserved, not adopted保留未採用版本
为了研究什么Research question為了研究什麼
检验现金转化是否提供倍数残差信息,并追加期限内触达评价。Test cash conversion as a multiple-residual predictor and append within-horizon touch evaluation.檢驗現金轉化是否提供倍數殘差信息,並追加期限內觸達評價。
具体方法与本代改变Method and changes具體方法與本代改變
保留G05盈利,把G11单利率状态换成同最新财务期连续4季现金转化比,使用平方岭修正。训练仍为到期收盘残差,触达附录只重评旧目标。Preserve G05 earnings. Replace G11’s yield state with a four-quarter cash-conversion ratio aligned to the latest profit quarter; fit squared-ridge corrections. Training still uses terminal closes; the touch appendix only rescores saved targets.保留G05盈利,把G11單利率狀態換成同最新財務期連續4季現金轉化比,使用平方嶺修正。訓練仍為到期收盤殘差,觸達附錄只重評舊目標。
相对G11只换单状态;相对G12损失也改变。未按窗口最高价重新训练。Relative to G11, replace one state. Relative to G12, the loss also changes. No fitting to window peaks occurred.相對G11只換單狀態;相對G12損失也改變。未按窗口最高價重新訓練。
核算公式Calculation formulas核算公式
q=(TTM_CFO−TTM_CashPPE_Capex)/(0.85×TTM_GAAP_OperatingProfit); z=q/0.10 min aΣv[y−α−β(z−μz)]²+2α²+8β² P13_h=G05_h×exp(clip(α+β(z_current−μz),±ln1.5))
冻结参数Frozen parameters凍結參數
| 参数Parameter參數 | 取值与规则Value and rule取值與規則 |
|---|---|
| 现金状态单位Cash-state unit現金狀態單位 | 10个百分点10 percentage points10個百分點 |
| 损失及先验Loss and priors損失及先驗 | 平方损失;2 / 8Squared loss; 2 / 8平方損失;2 / 8 |
本代计算值与日期Version outputs and dates本代計算值與日期
| 期限Horizon期限 | 原研究2026-10-06Original research: 2026-10-06原研究2026-10-06 | 原研究目标日期Original target date原研究目標日期 | 季度重建2026-07-23Quarterly reconstruction: 2026-07-23季度重建2026-07-23 | 季度目标日期Quarterly target date季度目標日期 |
|---|---|---|---|---|
| 0M | $347.68 | 2026-10-06 | $317.69 | 2026-07-23 |
| 6M | $411.13 | 2027-04-06 | $390.32 | 2027-01-23 |
| 12M | $587.03 | 2027-10-06 | $538.09 | 2027-07-23 |
上述两个输入截面分别保存,不能混为发布日的当前行情。0M是该截面的市场锚;本版未采用时,候选6M/12M不替换G05主表。These input snapshots are separately preserved, not publication-day market quotes. 0M is the snapshot’s market anchor. An unadopted candidate’s 6M/12M outputs do not replace the G05 main table.上述兩個輸入截面分別保存,不能混為發佈日的當前行情。0M是該截面的市場錨;本版未採用時,候選6M/12M不替換G05主表。
采用记录及研究结果Adoption record and findings採用記錄及研究結果
原终点两期筛选失败。新三年月末触达6M30/30、12M20/24,G05为29/30、23/24,未同时改善,继续保留G05。Both original terminal screens failed. In the later three-year monthly touch review, G13 was 30/30 at 6M and 20/24 at 12M versus G05 29/30 and 23/24; it did not improve both. Retain G05.原終點兩期篩選失敗。新三年月末觸達6M30/30、12M20/24,G05為29/30、23/24,未同時改善,繼續保留G05。
当前现金比42.45%低于成熟训练范围62.48%–122.27%,负斜率反而推高目标,是外推关联,不能说现金质量下降创造更高合理PE。Current cash conversion of 42.45% is below the mature training range of 62.48%–122.27%. A negative slope raises prices under this extrapolation; this does not mean lower cash quality creates a higher fair P/E.當前現金比42.45%低於成熟訓練範圍62.48%–122.27%,負斜率反而推高目標,是外推關聯,不能説現金質量下降創造更高合理PE。
G14 · 文档固定PE适配G14 · Document fixed-P/E adaptationG14 · 文檔固定PE適配 · 保留未采用版本Preserved, not adopted保留未採用版本
为了研究什么Research question為了研究什麼
检验脱离市场PE锚的固定25.5/26倍能否更接近期间高点且保持触达。Test whether fixed 25.5 / 26 multiples can stay near window peaks while preserving touch rates.檢驗脱離市場PE錨的固定25.5/26倍能否更接近期間高點且保持觸達。
具体方法与本代改变Method and changes具體方法與本代改變
保留G05按日历分配的经营盈利窗口,统一使用当时原财报直接披露、匹配最新利润季的季度稀释股数;6M固定25.5倍、12M固定26倍,不叠原残差校准。Keep G05 daily-overlap operating-earnings windows. Use directly reported quarterly diluted shares matched to the profit quarter; apply 25.5 at 6M and 26 at 12M without residual correction.保留G05按日曆分配的經營盈利窗口,統一使用當時原財報直接披露、匹配最新利潤季的季度稀釋股數;6M固定25.5倍、12M固定26倍,不疊原殘差校準。
只迁移文档固定PE;文档未定义的完整Non-GAAP盈利算法和DCF不能称已复现。后续G14-Q按财报后季度测算,替代月末评价频率而不改算法。Transfer only the document’s fixed multiples; its unspecified full Non-GAAP earnings algorithm and DCF are not reproduced. G14-Q later scores post-earnings quarterly origins without changing the algorithm.只遷移文檔固定PE;文檔未定義的完整Non-GAAP盈利算法和DCF不能稱已復現。後續G14-Q按財報後季度測算,替代月末評價頻率而不改算法。
核算公式Calculation formulas核算公式
EPS_h=G05_own_origin_NOPAT_window / matched_quarter_direct_diluted_shares P6=EPS6×25.5; P12=EPS12×26; P0=market_close
冻结参数Frozen parameters凍結參數
| 参数Parameter參數 | 取值与规则Value and rule取值與規則 |
|---|---|
| 6M/12M固定PE6M / 12M fixed P/E6M/12M固定PE | 25.5 / 2625.5 / 2625.5 / 26 |
| 未来股份Future shares未來股份 | 固定;不预测回购/发行Fixed; no buyback / issuance forecast固定;不預測回購/發行 |
| 调参Tuning調參 | 无None無 |
本代计算值与日期Version outputs and dates本代計算值與日期
| 期限Horizon期限 | 原研究2026-10-06Original research: 2026-10-06原研究2026-10-06 | 原研究目标日期Original target date原研究目標日期 | 季度重建2026-07-23Quarterly reconstruction: 2026-07-23季度重建2026-07-23 | 季度目标日期Quarterly target date季度目標日期 |
|---|---|---|---|---|
| 0M | $347.68 | 2026-10-06 | $317.69 | 2026-07-23 |
| 6M | $349.96 | 2027-04-06 | $342.31 | 2027-01-23 |
| 12M | $395.80 | 2027-10-06 | $386.09 | 2027-07-23 |
上述两个输入截面分别保存,不能混为发布日的当前行情。0M是该截面的市场锚;本版未采用时,候选6M/12M不替换G05主表。These input snapshots are separately preserved, not publication-day market quotes. 0M is the snapshot’s market anchor. An unadopted candidate’s 6M/12M outputs do not replace the G05 main table.上述兩個輸入截面分別保存,不能混為發佈日的當前行情。0M是該截面的市場錨;本版未採用時,候選6M/12M不替換G05主表。
采用记录及研究结果Adoption record and findings採用記錄及研究結果
未采用。原三年触达两期均83.33%,低于G05;五年季度触达6M11/18、12M10/16,峰值差虽最小但优先触达不胜出。Not adopted: original three-year touch rates were both 83.33%, below G05. Five-year quarterly touch was 11/18 and 10/16; the smallest peak gap does not override more misses.未採用。原三年觸達兩期均83.33%,低於G05;五年季度觸達6M11/18、12M10/16,峯值差雖最小但優先觸達不勝出。
低于当前市场隐含PE的固定倍数,可能高于某些历史期隐含PE;当前保守不等于历史更易触达。A multiple below the current implied P/E may exceed the implied P/E in other historical periods; current conservatism does not imply easier historical touches.低於當前市場隱含PE的固定倍數,可能高於某些歷史期隱含PE;當前保守不等於歷史更易觸達。
G15 · 文档宏观倍数与组合条件试验G15 · Document macro-multiple and ensemble trialG15 · 文檔宏觀倍數與組合條件試驗 · 条件试验,正式不可估算Conditional trial; formal model not estimable條件試驗,正式不可估算
为了研究什么Research question為了研究什麼
核验新文档十年40季算术、实际价格窗口,以及利率模块的条件表现。Audit forty quarterly document calculations, real price windows and conditional rate-module performance.核驗新文檔十年40季算術、實際價格窗口,以及利率模塊的條件表現。
具体方法与本代改变Method and changes具體方法與本代改變
2016Q3–2026Q2季末40行,文档NTM EPS保留;用核验日线和测算日前国债收益率。分别计算固定PE、利率PE、24到28倍时间代理、50/50组合。不是正式可估算模型。Keep document NTM EPS for forty quarter-end rows from 2016Q3 to 2026Q2. Use verified daily prices and pre-origin yields. Calculate fixed P/E, rate-adjusted P/E, a 24-to-28 time proxy and a 50/50 mixture. The formal model is not estimable.2016Q3–2026Q2季末40行,文檔NTM EPS保留;用核驗日線和測算日前國債收益率。分別計算固定PE、利率PE、24到28倍時間代理、50/50組合。不是正式可估算模型。
与G14不同:加入非对称利率规则和组合,但真实SOTP业务公式缺失、NTM共识vintage缺失;季末起点也不同于财报后协议。Unlike G14, add asymmetric rate rules and an ensemble. Actual business SOTP equations and consensus vintages are missing; quarter-end origins differ from the post-earnings protocol.與G14不同:加入非對稱利率規則和組合,但真實SOTP業務公式缺失、NTM共識vintage缺失;季末起點也不同於財報後協議。
核算公式Calculation formulas核算公式
PE_macro,h=clip[base_h−1.2×max(y10−2.5,0)+0.8×max(2.5−y10,0),bounds_h] PE_time(i)=24+4×i/39, i=0…39 T_h=document_NTM_EPS×(0.5PE_macro,h+0.5PE_time(i))
冻结参数Frozen parameters凍結參數
| 参数Parameter參數 | 取值与规则Value and rule取值與規則 |
|---|---|
| 中性10Y收益率Neutral 10Y yield中性10Y收益率 | 2.5%2.5%2.5% |
| 高/低利率系数High / low rate coefficients高/低利率係數 | 每个百分点−1.2 / +0.8倍−1.2 / +0.8 P/E per percentage point每個百分點−1.2 / +0.8倍 |
| 6M倍数边界6M multiple bounds6M倍數邊界 | 18–28.518–28.518–28.5 |
| 12M倍数边界12M multiple bounds12M倍數邊界 | 19–29.519–29.519–29.5 |
| 组合Mixture組合 | 50%宏观 + 50%时间代理50% macro + 50% time proxy50%宏觀 + 50%時間代理 |
| EPS来源EPS sourceEPS來源 | 文档假设,历史快照未验证Document assumption; historical snapshots unverified文檔假設,歷史快照未驗證 |
本代计算值与日期Version outputs and dates本代計算值與日期
G15正式0M、6M、12M均为空,因历史NTM共识快照和真实分部SOTP公式缺失。2026-06-30文档条件6M $371.52、12M $375.14,仅为假设输入的核算;$357.37是该日市价参考。G15 formal 0M, 6M and 12M values are null because historical consensus NTM snapshots and actual segment SOTP formulas are missing. The June 30, 2026 conditional $371.52 / $375.14 targets calculate assumed inputs only; $357.37 is that date’s market reference.G15正式0M、6M、12M均為空,因歷史NTM共識快照和真實分部SOTP公式缺失。2026-06-30文檔條件6M $371.52、12M $375.14,僅為假設輸入的核算;$357.37是該日市價參考。
查看完整条件试验与十年窗口View the full conditional trial and ten-year windows查看完整條件試驗與十年窗口采用记录及研究结果Adoption record and findings採用記錄及研究結果
不替换。正式0M/6M/12M值为NULL,不可估算;只发布清楚标注的条件试算。补齐历史NTM快照及真实SOTP公式,并统一起点后才可比较采用。Do not replace G05. Formal 0M/6M/12M values remain null. Publish only clearly labeled conditional calculations. Verified historical NTM snapshots, actual business SOTP equations and common origins are required before replacement testing.不替換。正式0M/6M/12M值為NULL,不可估算;只發布清楚標註的條件試算。補齊歷史NTM快照及真實SOTP公式,並統一起點後才可比較採用。
核验后组合6M30/39、12M34/37,不优于固定PE;文档把早于预测起点的2026-05-18高点误放入6月底以后窗口,已纠正。The verified ensemble touched 30/39 at 6M and 34/37 at 12M, without beating fixed P/E. The document incorrectly placed a May 18, 2026 high in the window after the June 30 origin; this was corrected.核驗後組合6M30/39、12M34/37,不優於固定PE;文檔把早於預測起點的2026-05-18高點誤放入6月底以後窗口,已糾正。
五年季度回测:先触达,再看峰值差Five-year quarterly review: touches first, then peak gaps五年季度回測:先觸達,再看峯值差
2021Q3–2026Q2共20次财报;每季只在发布后的首个完整常规交易日收盘测算。观察截止2026-10-06,6M成熟18、待到期2;12M成熟16、待到期4。未来窗口没有用最后可见价格补齐。Twenty earnings events from 2021Q3 through 2026Q2, evaluated only at the close of the first complete regular session after each release. Observation cutoff: October 6, 2026. 6M has 18 mature and 2 pending windows; 12M has 16 mature and 4 pending. Future windows are never filled with the last visible price.2021Q3–2026Q2共20次財報;每季只在發佈後的首個完整常規交易日收盤測算。觀察截止2026-10-06,6M成熟18、待到期2;12M成熟16、待到期4。未來窗口沒有用最後可見價格補齊。
触达:从起点下一交易日开始,到加6或12个日历月的截止日(含),任何一天常规时段high严格高于未四舍五入目标即命中,之后回落不取消。平均触达比例是所有完整窗口“最高价÷目标”的均值,包括未命中;它不同于触达率,不应单纯最大化。Touch: from the session after origin through the inclusive six- or twelve-calendar-month deadline, any regular-session daily high strictly above the unrounded target counts; a later retreat does not cancel the hit. Mean attainment is the average of peak/target over all complete windows, including misses. It differs from the hit rate and should not simply be maximized.觸達:從起點下一交易日開始,到加6或12個日曆月的截止日(含),任何一天常規時段high嚴格高於未四捨五入目標即命中,之後回落不取消。平均觸達比例是所有完整窗口“最高價÷目標”的均值,包括未命中;它不同於觸達率,不應單純最大化。
| 排名Rank排名 | 版本Version版本 | 6M触达6M hits6M觸達 | 6M最高价/目标均值6M mean peak/target6M最高價/目標均值 | 12M触达12M hits12M觸達 | 12M最高价/目标均值12M mean peak/target12M最高價/目標均值 | 综合触达率Combined hit rate綜合觸達率 | 综合绝对峰值差Combined absolute peak gap綜合絕對峯值差 |
|---|---|---|---|---|---|---|---|
| 1 | G04G04G04 | 16/18 · 88.89% | 119.64% | 13/16 · 81.25% | 135.27% | 85.07% | 21.13% |
| 2 | G05G05G05 | 16/18 · 88.89% | 119.64% | 13/16 · 81.25% | 132.52% | 85.07% | 21.20% |
| 3 | G01G01G01 | 16/18 · 88.89% | 119.64% | 13/16 · 81.25% | 138.97% | 85.07% | 21.42% |
| 4 | G07G07G07 | 14/18 · 77.78% | 118.39% | 13/16 · 81.25% | 132.52% | 79.51% | 21.28% |
| 5 | G03G03G03 | 14/18 · 77.78% | 116.07% | 13/16 · 81.25% | 134.65% | 79.51% | 21.37% |
| 6 | G06G06G06 | 14/18 · 77.78% | 116.00% | 13/16 · 81.25% | 132.52% | 79.51% | 21.81% |
| 7 | G09G09G09 | 13/18 · 72.22% | 115.23% | 13/16 · 81.25% | 131.94% | 76.74% | 21.42% |
| 8 | G02G02G02 | 13/18 · 72.22% | 117.49% | 13/16 · 81.25% | 135.16% | 76.74% | 22.50% |
| 9 | G08G08G08 | 14/18 · 77.78% | 115.86% | 12/16 · 75.00% | 131.61% | 76.39% | 21.82% |
| 10 | G10G10G10 | 14/18 · 77.78% | 169.90% | 12/16 · 75.00% | 183.87% | 76.39% | 45.53% |
| 11 | G11G11G11 | 14/18 · 77.78% | 116.58% | 10/16 · 62.50% | 130.71% | 70.14% | 22.25% |
| 12 | G13G13G13 | 14/18 · 77.78% | 118.87% | 10/16 · 62.50% | 128.84% | 70.14% | 22.80% |
| 13 | G14G14G14 | 11/18 · 61.11% | 103.46% | 10/16 · 62.50% | 108.86% | 61.81% | 19.04% |
| 14 | G12G12G12 | 11/18 · 61.11% | 116.26% | 10/16 · 62.50% | 128.47% | 61.81% | 23.54% |
先按两期限成熟触达率等权平均排序;并列再按两期限平均绝对峰值差|目标÷最高价−1|等权平均排序。G04综合第一,G05第二;两者触达次数相同,峰值差领先仅约0.067个百分点。单看6M,G01/G04/G05完全并列;单看12M,G03第一。此次排序未自动采用G04或拼接新期限。Rank first by the equal-weighted mature hit rates of both horizons, then by their equal-weighted mean absolute peak gap |target/peak−1|. G04 is first overall, G05 second, with identical hit counts and only about a 0.067-percentage-point gap advantage. G01/G04/G05 tie exactly at 6M; G03 leads at 12M. This ranking does not automatically adopt G04 or splice horizons.先按兩期限成熟觸達率等權平均排序;並列再按兩期限平均絕對峯值差|目標÷最高價−1|等權平均排序。G04綜合第一,G05第二;兩者觸達次數相同,峯值差領先僅約0.067個百分點。單看6M,G01/G04/G05完全並列;單看12M,G03第一。此次排序未自動採用G04或拼接新期限。
| 版本Version版本 | 6M未到期:已触达 / 未触达6M pending: observed hit / not yet hit6M未到期:已觸達 / 未觸達 | 12M未到期:已触达 / 未触达12M pending: observed hit / not yet hit12M未到期:已觸達 / 未觸達 |
|---|---|---|
| G04 | 2 / 0 | 2 / 2 |
| G05 | 2 / 0 | 1 / 3 |
| G01 | 2 / 0 | 3 / 1 |
| G07 | 1 / 1 | 1 / 3 |
| G03 | 1 / 1 | 2 / 2 |
| G06 | 1 / 1 | 1 / 3 |
| G09 | 1 / 1 | 2 / 2 |
| G02 | 1 / 1 | 3 / 1 |
| G08 | 1 / 1 | 2 / 2 |
| G10 | 2 / 0 | 4 / 0 |
| G11 | 1 / 1 | 1 / 3 |
| G13 | 0 / 2 | 1 / 3 |
| G14 | 2 / 0 | 3 / 1 |
| G12 | 0 / 2 | 1 / 3 |
未到期但已超过目标标为观察触达,未超过标为待观察。二者都不进入成熟触达率和完整窗口最高价比例,因为窗口最高价还未定。Pending windows that have exceeded the target are observed hits; those that have not remain unresolved. Neither enters mature hit rates or complete-window peak ratios because the final peak is not yet known.未到期但已超過目標標為觀察觸達,未超過標為待觀察。二者都不進入成熟觸達率和完整窗口最高價比例,因為窗口最高價還未定。
原三年终点价格评分:保留历史筛选记录Original three-year terminal-price scores: preserved selection history原三年終點價格評分:保留歷史篩選記錄
原研究使用2023-10至2026-09的36个月末,6M成熟30/待到期6,12M成熟24/待到期12。实际标签取目标日当天或之后首个交易日收盘。此处评价期末预测误差,不是期限内触达;原采用和拒绝记录基于这些当时冻结的门槛。The original studies used 36 month-end origins from October 2023 through September 2026: 30 mature / 6 pending at 6M and 24 mature / 12 pending at 12M. Actual labels use the close on or first session after the target date. This measures terminal-price error, not within-horizon touch; original decisions used the screens frozen at the time.原研究使用2023-10至2026-09的36個月末,6M成熟30/待到期6,12M成熟24/待到期12。實際標籤取目標日當天或之後首個交易日收盤。此處評價期末預測誤差,不是期限內觸達;原採用和拒絕記錄基於這些當時凍結的門檻。
| 版本Version版本 | 6M MAPE6M MAPE6M MAPE | 6M中位误差6M median error6M中位誤差 | 6M P90误差6M P90 error6M P90誤差 | 12M MAPE12M MAPE12M MAPE | 12M中位误差12M median error12M中位誤差 | 12M P90误差12M P90 error12M P90誤差 |
|---|---|---|---|---|---|---|
| G01 | 16.45% | 12.46% | 41.57% | 28.45% | 28.18% | 46.61% |
| G02 | 16.46% | 12.90% | 39.35% | 27.16% | 28.88% | 43.51% |
| G03 | 15.96% | 14.18% | 38.99% | 26.34% | 26.88% | 42.37% |
| G04 | 16.45% | 12.46% | 41.57% | 26.40% | 27.99% | 42.59% |
| G05 | 16.45% | 12.46% | 41.57% | 25.84% | 25.96% | 41.41% |
| G06 | 15.99% | 14.57% | 38.83% | 25.84% | 25.96% | 41.41% |
| G07 | 16.12% | 12.82% | 40.86% | 25.84% | 25.96% | 41.41% |
| G08 | 15.95% | 12.92% | 38.91% | 25.51% | 25.92% | 40.96% |
| G09 | 16.10% | 12.27% | 38.36% | 25.49% | 25.27% | 41.31% |
| G10 | 46.18% | 49.00% | 61.26% | 50.82% | 53.85% | 62.13% |
| G11 | 15.89% | 14.41% | 38.41% | 24.26% | 25.21% | 37.09% |
| G12 | 16.42% | 13.63% | 38.03% | 23.73% | 23.05% | 40.63% |
| G13 | 15.63% | 11.90% | 41.42% | 25.21% | 26.46% | 40.60% |
| G14 | 16.50% | 16.00% | 31.23% | 18.58% | 18.80% | 29.95% |
MAPE以实际价格为分母,越小越好;不能用100%−MAPE写成准确率。中位误差表示典型记录,P90是历史误差的第90百分位,不是未来置信区间。月末窗口重叠,且过去成绩被多轮研究看过,因此不是独立新样本外验证。MAPE uses actual price as the denominator and is lower-is-better; 100%−MAPE is not an accuracy rate. The median represents typical records, and P90 is a historical error percentile, not a future confidence interval. Month-end windows overlap and their results were reused across iterations; this is not independent new out-of-sample validation.MAPE以實際價格為分母,越小越好;不能用100%−MAPE寫成準確率。中位誤差表示典型記錄,P90是歷史誤差的第90百分位,不是未來置信區間。月末窗口重疊,且過去成績被多輪研究看過,因此不是獨立新樣本外驗證。
分析师比较独立保存:36个真实观察日期的模型与共识目标差只衡量意见距离;24个成熟12M共同终点才衡量兑现价格误差。G05共同终点MAPE25.18%,分析师22.78%;同日目标平均绝对差7.86%。分析师没有参与校准,也不能用12M共识评价6M。商业原始逐日记录保留内部。Analyst comparisons remain separate: model/consensus gaps at 36 actual observation dates measure disagreement; only the 24 mature matched 12M terminal prices measure realized forecast error. G05 matched-terminal MAPE was 25.18%, versus 22.78% for analysts; mean absolute same-date target difference was 7.86%. Analysts were not used for calibration, and 12M consensus cannot evaluate 6M. Commercial daily source records remain internal.分析師比較獨立保存:36個真實觀察日期的模型與共識目標差只衡量意見距離;24個成熟12M共同終點才衡量兑現價格誤差。G05共同終點MAPE25.18%,分析師22.78%;同日目標平均絕對差7.86%。分析師沒有參與校準,也不能用12M共識評價6M。商業原始逐日記錄保留內部。
G15:文档十年条件试算,未替换主模型G15: ten-year document-input trial, not a replacementG15:文檔十年條件試算,未替換主模型
正式模型不可估算:历史NTM共识快照、预测日期和调整口径尚未验证;真实分部SOTP公式缺失。以下数字是2026-06-30文档行的条件计算,不是2026-10-09新预测。The formal model is not estimable: historical NTM consensus snapshots, forecast dates and adjustment basis remain unverified, and actual segment SOTP equations are missing. These values calculate the June 30, 2026 document row; they are not new October 9 forecasts.正式模型不可估算:歷史NTM共識快照、預測日期和調整口徑尚未驗證;真實分部SOTP公式缺失。以下數字是2026-06-30文檔行的條件計算,不是2026-10-09新預測。
| 身份Status身份 | 0M | 6M | 12M |
|---|---|---|---|
| 正式模型值Formal model values正式模型值 | 未定义Not defined未定義 | 不可估算Not estimable不可估算 | 不可估算Not estimable不可估算 |
| 文档条件试算Document conditional trial文檔條件試算 | $357.37市价参考$357.37 market reference$357.37市價參考 | $371.52 | $375.14 |
| 条件目标截止日Conditional target deadline條件目標截止日 | 2026-06-30 | 2026-12-30 | 2027-06-30 |
保留文档EPS假设$14.50,使用6月29日已观察10Y国债4.38%,中性2.50%,高利率每个百分点减1.2倍;临时时间倍数为28。时间代理24+4×季度序号/39只复现文档汇总,不能冒称业务分部SOTP。Keep the document EPS assumption of $14.50 and the June 29 observed 10Y Treasury yield of 4.38%. The neutral yield is 2.50%; each percentage point above it subtracts 1.2 P/E turns. The provisional time multiple is 28. The 24+4×quarter-index/39 proxy reproduces document summaries but is not a business-segment SOTP.保留文檔EPS假設$14.50,使用6月29日已觀察10Y國債4.38%,中性2.50%,高利率每個百分點減1.2倍;臨時時間倍數為28。時間代理24+4×季度序號/39只復現文檔彙總,不能冒稱業務分部SOTP。
6M = 14.50 × [0.5 × (25.5 − 1.2 × (4.38 − 2.50)) + 0.5 × 28] = 371.519 12M = 14.50 × [0.5 × (26.0 − 1.2 × (4.38 − 2.50)) + 0.5 × 28] = 375.144
十年试算包括2016Q3–2026Q2的40个季末,非财报后测算;6M完整39、未到期1,12M完整37、未到期3。使用核验日线及起点前已观察利率,EPS仍为文档未验证值。不能与五年财报后模型排名直接比较。The ten-year trial covers forty quarter-ends from 2016Q3 through 2026Q2, rather than post-earnings origins. It has 39 complete / 1 pending windows at 6M and 37 complete / 3 pending at 12M. Daily prices and pre-origin observed yields are verified; EPS remains an unverified document input. It cannot be ranked directly against the five-year post-earnings models.十年試算包括2016Q3–2026Q2的40個季末,非財報後測算;6M完整39、未到期1,12M完整37、未到期3。使用核驗日線及起點前已觀察利率,EPS仍為文檔未驗證值。不能與五年財報後模型排名直接比較。
| 条件方法Conditional method條件方法 | 6M触达6M hits6M觸達 | 6M最高价/目标均值6M mean peak/target6M最高價/目標均值 | 6M峰值差6M peak gap6M峯值差 | 12M触达12M hits12M觸達 | 12M最高价/目标均值12M mean peak/target12M最高價/目標均值 | 12M峰值差12M peak gap12M峯值差 |
|---|---|---|---|---|---|---|
| 固定25.5/26倍Fixed 25.5 / 26 multiples固定25.5/26倍 | 31/39 · 79.49% | 109.35% | 9.91% | 34/37 · 91.89% | 122.24% | 17.02% |
| 利率调节Rate adjustment利率調節 | 32/39 · 82.05% | 111.86% | 10.99% | 35/37 · 94.59% | 124.64% | 18.11% |
| 时间倍数代理Time-multiple proxy時間倍數代理 | 28/39 · 71.79% | 107.59% | 9.04% | 35/37 · 94.59% | 122.97% | 17.66% |
| 50/50组合条件版50/50 conditional mixture50/50組合條件版 | 30/39 · 76.92% | 109.47% | 9.74% | 34/37 · 91.89% | 123.57% | 17.86% |
组合没有体现替换优势。文档引用的$408.61高点发生在2026-05-18,早于6月30日起点;起点之后截至10月6日已观察高点为$384.48,相关窗口尚未结束。补齐带预测日期的EPS和真实SOTP公式,再统一财报后起点与当前模型比较,才是可用于替换的研究。The mixture did not establish replacement value. The document’s $408.61 high occurred on May 18, 2026, before the June 30 origin; the observed post-origin peak through October 6 was $384.48, and those windows remain incomplete. Replacement research requires dated EPS snapshots, actual SOTP equations and origins aligned with the current model.組合沒有體現替換優勢。文檔引用的$408.61高點發生在2026-05-18,早於6月30日起點;起點之後截至10月6日已觀察高點為$384.48,相關窗口尚未結束。補齊帶預測日期的EPS和真實SOTP公式,再統一財報後起點與當前模型比較,才是可用於替換的研究。
验证边界与下一步Validation boundaries and next steps驗證邊界與下一步
原版本、拒绝候选和历史预测保留;这次网页更新发布方法与聚合结果,没有重新训练、改参数或追认最优模型。核算审计通过只验证计算、来源和时间边界,不证明模型假设正确、独立前瞻胜出或未来触达。Original versions, rejected candidates and saved forecasts remain preserved. This website update publishes methods and aggregate results without retraining, changing parameters or retroactively adopting a winner. Calculation audits verify arithmetic, inputs and timing; they do not prove assumptions, independent prospective superiority or future touches.原版本、拒絕候選和歷史預測保留;這次網頁更新發布方法與聚合結果,沒有重新訓練、改參數或追認最優模型。核算審計通過只驗證計算、來源和時間邊界,不證明模型假設正確、獨立前瞻勝出或未來觸達。
财报按当时披露可用性重建,但没有认证当时不可修订的历史PIT档案;分析师历史原始预测版本也未认证。价格是拆股调整常规时段日线,不是含股息总回报;日线触达只说明某天曾超过目标,不证明可成交时刻或独立公允价值。Filings are reconstructed by historical disclosure availability, without certification of immutable contemporaneous PIT archives; original analyst forecast vintages are also unverified. Prices are split-adjusted regular-session daily bars, not dividend-inclusive returns. A daily touch shows the target was exceeded on a day, not an executable timestamp or independent fair value.財報按當時披露可用性重建,但沒有認證當時不可修訂的歷史PIT檔案;分析師歷史原始預測版本也未認證。價格是拆股調整常規時段日線,不是含股息總回報;日線觸達只説明某天曾超過目標,不證明可成交時刻或獨立公允價值。
下一步:补齐G15历史共识与SOTP原始公式;冻结新的季度触达研究设计;在相同起点、相同完整窗口先比较触达,再约束与峰值的差距与低目标比例,同时封存之后每次财报的新预测等待兑现。Next: complete G15 historical consensus inputs and original SOTP equations; freeze a new quarterly touch study; compare touches first on identical origins and complete windows, then constrain peak gaps and low-target incidence. Archive new post-earnings forecasts before waiting for outcomes.下一步:補齊G15歷史共識與SOTP原始公式;凍結新的季度觸達研究設計;在相同起點、相同完整窗口先比較觸達,再約束與峯值的差距與低目標比例,同時封存之後每次財報的新預測等待兑現。
Alphabet 2026Q2 10-QAlphabet 2026Q2 10-QAlphabet 2026Q2 10-Q · 美国财政部10Y收益率源U.S. Treasury yield source美國財政部10Y收益率源 · 成长、再投资与终值关系Growth, reinvestment and terminal value成長、再投資與終值關係
返回Google个股研究模型价格Back to Google equity-research model prices返回Google個股研究模型價格