JD.com · Streaks
History: 23.05.2014 — 02.10.2026 · 646 periods
JD: 6 weeks falling in a row
Candle colour compares close with open. Highlighting marks the close-to-close streak.
Since 2014: 4 earlier streaks of 6 or more · -12.22%
How long past streaks lasted
Completed episodes in the current direction; boundary-censored runs excluded.
Independent baseline
The baseline assumes independent movements with the instrument’s historical directional frequency. It describes expected episode counts, not the chance of a reversal or profit. Actual market movements need not be independent.
- Observed episodes
- 4
- Expected episodes
- 5.25
- Directional frequency
- 50.5%
What happened afterwards
Past episodes of at least this length and direction. Retrospective, not a forecast.
| Periods after | Cases | Median | Middle 50% | Min / max |
|---|---|---|---|---|
| +1 week | 4 Small sample | +4.76% | +2.82% / +8.78% | +0.59% / +17.22% |
| +5 weeks | 4 Small sample | -1.50% | -4.29% / +13.44% | -9.28% / +54.89% |
| +20 weeks | 4 Small sample | +3.03% | -1.16% / +11.49% | -12.32% / +35.45% |
Retrospective returns start at the last close of a completed streak. Its ending becomes known only after the next closed period. Every horizon has its own sample; missing futures are excluded. Overlapping windows are not independent trials.
Recent completed episodes
The end is confirmed by a later opposite or flat close.
Follow a streak
Use your existing email or linked Telegram subscription.
JD.com: 6 weeks falling in a row. Since 2014: 4 earlier streaks of 6 or more. History: 2014-05-23 — 2026-10-02
How streaks are measured
Counting rules
A rising period closes above the previous close; a falling period closes below it. Equal closes break a streak. Crypto uses UTC and Monday weeks. Session markets use their exchange calendar; forex uses 17:00 New York from hourly observations. A missing expected observation is a data gap, not a reversal. Each maximal completed run counts once. The current run and censored boundary runs are excluded from past-episode counts.
Independent baseline
The baseline assumes independent movements with the instrument’s historical directional frequency. It describes expected episode counts, not the chance of a reversal or profit. Actual market movements need not be independent.
What happened afterwards
Retrospective returns start at the last close of a completed streak. Its ending becomes known only after the next closed period. Every horizon has its own sample; missing futures are excluded. Overlapping windows are not independent trials.
Frequently asked questions
Does a long streak mean the market will reverse?
No. Length describes past closes. It does not predict the next move. The historical comparison and independent baseline are descriptive.
How are past streaks counted?
Each completed same-direction episode of at least the selected length counts once. A 12-period run is one episode, even when the threshold is nine.
Why do history and sample sizes differ?
Feeds have different start dates, calendars and missing observations. Longer outcome horizons also need more future observations. Small samples are marked explicitly.
Are these all-time records?
Records apply only to the named instrument, source and available verified history. Crypto venues are not stitched together. Commodity futures are labelled as proxies.
Historical observations are not forecasts or investment advice. Small samples and overlapping outcome windows limit interpretation. No buy or sell recommendation.