Calendar Clock日历时钟
SPDR S&P 500 ETF Trust SPY
Year-end strength here holds up after counting every window tried. Get ready.年末走强，在计入所有测试窗口后依然成立。做好准备。
- Oct 11 → Dec 1010月11日 → 12月10日
- Your window · exploratory自选窗口 · 探索性?A window you chose after seeing the chart spends testing budget that is not counted here. Presets are counted.你在看过图之后自己选的窗口，会消耗这里没有计入的测试预算。预设窗口是计入的。
- 25 years25 年
- 22 of 25 up25 年中 22 年上涨
- Its own pattern自身的规律 ?This symbol is the market benchmark, so there is no separate market leg to remove — its calendar is the market's by definition.该标的就是市场基准，没有可剥离的大盘部分——它的日历规律即为大盘的规律。
- Through Jul 31截至 7月31日
Seasonality data didn’t load. Reload the page.季节性数据未能加载，请重新载入页面。
We don’t cover that symbol yet.暂未覆盖该代码。 See what we cover →查看覆盖范围 →
This window这个窗口
typical year典型年份
Each thread is one year, starting from zero on Oct 11. 22 of 25 finished above the line.每条线是一年，从 10月11日 起算为零。25 年中 22 年收在零线之上。
We shuffled SPY’s history 2,000 times. A window this strong turned up by chance in 2% of them — rare enough that this one does stand out. 我们把 SPY 的历史打乱了 2,000 次。像这样强的窗口在其中 2% 里纯属偶然出现——出现得够罕见，因此这一个确实突出。 ?Family: 2645 windows on this symbol (365 start days × 8 lengths).Your window: |t| 7.25. Chance-alone 95th percentile of the best window in the family: |t| 6.65 (2,000 independent year-shift resamples).Method: joint maxT, Westfall–Young style, dependence preserved by shifting whole years.Unit of evidence: one complete year, not one day.窗口族：该代码共 2645 个窗口（365 个起点 × 8 种长度）。本窗口：|t| 7.25。纯属偶然情况下族内最佳窗口的 95 分位：|t| 6.65（2,000 次独立整年循环平移重抽样）。方法：联合 maxT，Westfall–Young 式，通过整年平移保留相关结构。证据单位：一个完整年份，而非一个交易日。
Nudging the window a few days either way loses it — the effect depends on these exact dates.把窗口前后挪几天就消失了——这个效应取决于这几个具体日期。
The years各年份
| Year年份 | Return回报 | Scale幅度 |
|---|---|---|
| 2001 | +3.98% | |
| 2002 | +7.77% | |
| 2003 | +2.07% | |
| 2004 | +5.94% | |
| 2005 | +6.67% | |
| 2006 | +4.67% | |
| 2007 | -2.18% | |
| 2008 | +1.82% | |
| 2009 | +3.15% | |
| 2010 | +6.71% | |
| 2011 | +5.30% | |
| 2012 | -0.62% | |
| 2013 | +6.16% | |
| 2014 | +6.62% | |
| 2015 | +2.26% | |
| 2016 | +6.13% | |
| 2017 | +4.11% | |
| 2018 | -2.98% | |
| 2019 | +5.82% | |
| 2020 | +5.73% | |
| 2021 | +8.29% | |
| 2022 | +9.93% | |
| 2023 | +5.47% | |
| 2024 | +4.01% | |
| 2025 | +5.29% |
More views更多视角
By month按月份 ?Bars: mean return per month, one complete year per observation.Strongest — N: mean +2.71%, median +2.92%, share up 80%, years counted 25.Weakest — S: mean -1.15%, median +0.59%, share up 56%, years counted 25.柱形：按月份的平均回报，每个完整年份计一次。最强 — 11月：均值 +2.71%，中位数 +2.92%，上涨占比 80%，计入年数 25。最弱 — 9月：均值 -1.15%，中位数 +0.59%，上涨占比 56%，计入年数 25。
Average calendar-month return, one point per year.各月平均回报，每年计一次。
By weekday按星期 ?Bars: mean return per weekday, one complete year per observation.Strongest — T: mean +3.12%, median +1.96%, share up 68%, years counted 25.Weakest — F: mean +0.60%, median -0.15%, share up 48%, years counted 25.柱形：按星期的平均回报，每个完整年份计一次。最强 — 二：均值 +3.12%，中位数 +1.96%，上涨占比 68%，计入年数 25。最弱 — 五：均值 +0.60%，中位数 -0.15%，上涨占比 48%，计入年数 25。
Average day-of-week return, one point per year.各星期平均回报，每年计一次。
By trading day按月内交易日 ?Bars: mean return per trading day, one complete year per observation.Strongest — 1: mean +1.81%, median +2.55%, share up 72%, years counted 25.Weakest — 14: mean -2.01%, median -1.89%, share up 24%, years counted 25.柱形：按交易日的平均回报，每个完整年份计一次。最强 — 1：均值 +1.81%，中位数 +2.55%，上涨占比 72%，计入年数 25。最弱 — 14：均值 -2.01%，中位数 -1.89%，上涨占比 24%，计入年数 25。
Average return by trading day of the month.按月内第几个交易日的平均回报。
How this is computed计算方式
- What a year means here这里的「一年」是什么
- One complete year is one piece of evidence — not one trading day. 25 years of history is 25 pieces.一个完整年份 = 一条证据，而不是一个交易日。25 年历史 = 25 条证据。
- Prices价格
- Split- and dividend-adjusted closing prices from our daily vendor feed.来自我们每日数据源的复权（拆股与分红）收盘价。?Adjustment is the vendor’s current vintage, re-applied to all history — it is not a frozen point-in-time adjustment, so a very old year can read slightly differently than it did at the time.复权采用数据源当前版本并回溯应用于全部历史，并非冻结的时点复权，因此很久以前的年份读数可能与当时略有不同。
- Which names覆盖哪些标的
- We cover 220 symbols with enough complete years to draw.我们覆盖 220 个拥有足够完整年份的代码。?The list is built from today’s index membership, so names that were delisted or acquired are not in it. That makes the covered set look healthier than the real historical market. This page never ranks names against each other, so that bias does not enter any number on it.名单基于今天的指数成分构建，已退市或被收购的标的不在其中，因此覆盖范围看起来会比真实历史市场更健康。本页从不对标的互相排名，所以这一偏差不会进入页面上的任何数字。
- Browsing many symbols浏览多个代码时
- Look at enough symbols and a few clear the bar by luck alone — about 1 in 20. That is why this page answers one symbol at a time and never ranks them.看得足够多，总有少数标的纯靠运气过关 — 大约二十分之一。这正是本页每次只回答一个代码、且从不排名的原因。 Across the 220 symbols we cover, 34 clear their own bar — about 15 in 100.在我们覆盖的 220 个代码中，有 34 个通过了各自的门槛 — 约百分之 15。?Browsing N symbols runs N familywise tests, so roughly 5% clear by chance regardless of the per-symbol correction. The program-level rate against that 5% expectation is published with the universe build; a Benjamini–Yekutieli step across symbols is not used here because it treats heavily correlated equities as independent hypotheses and is far too conservative.浏览 N 个代码即等于做了 N 次族级检验，因此无论单代码校正如何，约 5% 会凭偶然过关。全域实际比率将随universe构建一同发布；此处不采用跨代码的 Benjamini–Yekutieli 步骤，因为它把高度相关的股票当作独立假设，过于保守。
- No screener yet暂无筛选器
- We don’t rank symbols against each other yet. When we do, it will carry its own selection accounting.我们暂不对个股互相排名。推出时会附带独立的选择校正。
- Windows you drag你自己拖出的窗口
- Presets are counted in the search budget. A window you draw after seeing the chart is not — it’s marked exploratory.预设窗口已计入搜索预算。你在看过图之后自己拖出的窗口没有计入 — 会标记为探索性。
Full method (JSON) →完整方法（JSON）→ · Generated生成于 2026-08-03 10:39 UTC