shahvanicaفارسی
About

About this site

An independent, bilingual look at one Persian-language story site: what was collected, how the numbers were made and what they cannot tell you.

This is an independent analysis. It is not affiliated with shahvani.com, and it shows no logo, photo or audio from that site.

Data and provenance

Everything on the site comes from one database snapshot. The collection process is not documented, so a missing row is not evidence that something was deleted on the original site. No new data was collected for this analysis.

What the snapshot holds
TableRowsPeriod
Stories37,8372009-10-21– 2026-11-15
Comments536,9192010-04-02– 2026-11-16
Member profiles234,0772015-12-01– 2026-11-16
Forum topics303,9772009-11-04– 2026-11-16
Photo records63,8852010-09-07– 2026-11-15
Tags710no dates

Not in the data: story authors, comment threading, forum post bodies and authors, and the photo and audio files themselves. Dates are server time. Account sign-up dates start part-way through the record; older accounts are undated.

The province map is drawn from outlines © OpenStreetMap contributors (Open Database License), simplified for this site. The only figures that do not come from the corpus are the provincial populations used for per-capita rates. openstreetmap.org/copyright

Method

How the numbers were made. Every threshold below is a setting in the analysis, not a number chosen per chart.

Source
One SQLite snapshot. Counts such as views and likes are totals as of collection, with no history.
Cleaning
Negative like counts are clipped to zero. Comments dated before their story (254) are left out of delay figures only. HTML is stripped from bodies before tokenising.
Tokens
Persian letters only, with Arabic ⁨ي ك ى⁩ folded to ⁨ی ک⁩ respectively, zero-width joiners treated as spaces, diacritics and tatweel dropped, and runs of 3+ repeated letters collapsed. Words under 2 letters are ignored. Latin tokens of 3+ letters are counted apart.
Spam
A comment is flagged when its normalised text repeats at least 5 times with at least 6 words, or contains a link marker. 2,818 comments were flagged. They stay in volume counts and are left out of word statistics.
Hours
Hour of day uses one fixed offset of ⁦UTC+3:30⁩ for every year.
Thresholds
Tag statistics need 150 stories per tag; tag pairs need 60 shared stories; title words need 150 titles.
Profiles
Gender, province and city are taken as stated. Provincial rates divide profile counts by population figures from Statistical Centre of Iran, 2016 census (rounded); this is the only external figure used.
Forum places
Places are the cities named by at least 150 profiles plus every province, matched as words in topic titles; a trailing ⁨ی⁩ counts as a match.
Views per day
Age is measured to the end of the record; stories under 30 days old are not rated. The last-activity stamp is not used.
Retention
Accounts are followed through clean comments from their first to their last. Survival counts only accounts old enough to be observed for the age asked. Burstiness needs 10+ comments.
Series
A series is a set of stories whose titles share a stem and end in part numbers (a bracketed number, a marker word and a number, or an ordinal word). Stems where two stories share a part number are skipped. Authors are unknown, so two unrelated stories with the same stem and different numbers can be joined by mistake.
Style
Style features use stories with at least 300 Persian words. Pronoun, spoken and formal markers are exact-word lists, so they are indicators, not a grammar analysis.
Reply mentions
There is no parent-comment id, and only 68 comments name an earlier commenter with an @ mention, so no reply graph was built.
Outside events
Dated events from the public record (messenger bans and shutdowns, Covid-19, protests, war) are marked on date-axis charts and listed below. They are not in the database. They show timing only, not cause.
Sensitive content
Explicit terms are shown as they appear in the corpus. Titles that mention minors are hidden from every listing and tags of that kind are excluded. User names appear only in the top-commenter list. No profile, bio, listing, link, photo or audio is shown.

Caveats

Read the figures with these in mind.

  • This is one snapshot, and the original site keeps growing: figures change when the database is replaced.
  • 37,287 of 65,665 commenting accounts have no profile row, so profile-based comparisons cover only the matched part.
  • 10 profile columns are empty in this snapshot and are not analysed.
  • Views, likes and dislikes are cumulative totals at collection time, so recent stories have had less time to gather them. Percentiles within the publication year reduce this bias but do not remove it.
  • The forum table has no author column, so forum figures count topics, not people. Photo uploads are dominated by very few accounts.
  • Correlations describe association only. Popular stories draw both likes and dislikes because both need a viewer.
  • Word rankings are shaped by spelling variants, names and series; the function-word list is hand-built and imperfect.

Outside events

These dates come from the public record, not from the data. They are drawn on the time charts so that timing can be read against them; they are not explanations.

Dated outside events
DateEvent
Apr 30, 2018Telegram blocked in Iran (judiciary order)
Jul 17, 2018Yahoo Messenger shut down
Nov 16, 2019Nationwide internet shutdown during the November protests
Feb 19, 2020First confirmed Covid-19 cases in Iran
Sep 16, 2022Death of Mahsa Amini and the protests that followed
Dec 24, 2024WhatsApp and Google Play unblocked
Jun 13, 2025Israel-Iran war begins (internet restrictions)

Downloads

Aggregate tables behind the charts, as plain files. They carry the same filters as the site: no profile rows, no listing details, no minor-adjacent titles.

Glossary

Terms used on the site.

14 terms

View percentile
Rank of a story's views among stories published in the same year (0 to 100), which removes the advantage of age.
Like ratio
Likes divided by likes plus dislikes, with negative counts clipped to zero.
Gini coefficient
Inequality of views across stories: 0 means equal, 1 means one story has everything.
Spearman correlation
Rank correlation between two measures; 0 means unrelated, 1 means the same order.
Lift
How much more often something occurs inside a group than in the whole corpus; 1 means no difference.
Jaccard
Shared stories divided by stories carrying either tag, for a pair of tags.
Per million words
Word count divided by the year's Persian word total, times one million, so years of different size compare.
Function words
Pronouns, particles, clitics and auxiliary verb forms removed before ranking vocabulary.
Clean comments
Comments left after removing repeated copy-paste texts and link spam; used for word and style statistics.
Heavy commenter
A commenting account with at least 100 comments.
Views per day
A story's total views divided by the days between publication and the end of the record; only stories above a minimum age are rated.
Burstiness
Measure of how uneven the gaps between one account's comments are: -1 is perfectly regular, 0 random, +1 comes in bursts.
Type-token ratio
Distinct words divided by words in a fixed-length stretch of text (here the first words of each story); higher means more varied vocabulary.
Zipf slope
Slope of word frequency against rank on log-log axes; about -1 is typical of natural language.

Explore

Search