Analysis archive

A Top 10 Is Not the Same Object Everywhere: Measuring Ranking Depth

Eight feeds, eight list shapes. This Q1 2026 audit shows why cross-platform comparisons need a declared rank window.

TG
TrendGoing Editorial
Editorial standards

A “trend list” sounds like a standard object. It is not. In our Q1 archive, the typical saved list ranged from 19 items on Zhihu to 50 on Weibo. That difference changes almost every comparison a researcher might make.

If one feed exposes 50 positions and another exposes 10, counting raw overlap rewards the deeper list. Counting all appearances rewards sources with more available slots. Even a simple claim such as “this topic lasted three days” depends on whether we watched the top 10, top 20, or the full captured list.

Median number of items in a saved ranking

Weibo50 items
Toutiao50 items
Baidu50 items
Hacker News Top50 items
Hacker News Best50 items
V2EX38 items
Tieba30 items
Zhihu19 items
Median list depth across every non-empty Q1 2026 snapshot.

The practical fix: choose a common rank window

For comparisons in this series, we use the first 20 positions when both lists contain at least 20 items. The cutoff is not sacred. It is a declared compromise that preserves more information than a top-five comparison while avoiding the deepest tail of longer feeds. A different study could reasonably choose 10 or 30, but it should choose before looking at the outcome.

Source10th percentileMedian90th percentile
Weibo505050
Toutiao505050
Baidu505050
Hacker News Top505050
Hacker News Best505050
V2EX213842
Tieba293030
Zhihu181920
Table 1. Observed list lengths; percentiles describe captured output, not an official platform limit.

The percentile columns are a data-quality check. A stable 10th-to-90th-percentile range means the collector usually received a consistent list shape. A wide range can signal a changing upstream response, a partial fetch, or a source whose output is genuinely variable. We do not silently pad short lists with blanks or assume missing ranks were unchanged.

Depth is part of the product

List depth affects what a visitor experiences. A short list emphasizes consensus and speed. A longer list preserves niche or slow-moving items that would disappear from a top-ten view. Neither is inherently better. The trouble begins when analysts compare them without acknowledging that one source offered five times as many chances to observe an item.

Official descriptions reinforce that the feeds are designed differently. Zhihu describes a top-30 list of high-activity questions. Weibo's published management rules describe a top-50 display. The Hacker News API can return up to 500 IDs for top and new stories, while a website or archive may intentionally retain a smaller slice. The number stored by TrendGoing is therefore an observation of our collection, not a claim about a universal platform standard.

A small example

Suppose Topic A appears at rank 18 on a 50-item feed and is absent from a 10-item feed. The data does not show that the second community ignored Topic A; it shows only that Topic A was not visible in the ten positions we observed. A fair conclusion would compare the common top 10, or explicitly say that the second feed had insufficient depth for the question.

This is why our charts include the denominator. “Six shared items” is ambiguous. “Six of the first 20 positions” is inspectable. Readers can disagree with the chosen window without having to guess what was measured.

Primary references

About this analysis

TrendGoing uses public announcements and observable online signals as source material, then adds original interpretation. Analysis reflects the information available on the publication date and does not constitute financial, legal, or medical advice.

Spot an error or missing context? Email [email protected] with this page's URL.

Keep reading

More original analysis