Research talk:Portuguese Wikipedia trends and behavior/Overview Reverts and Deletions

Some comments:

• 3. The proportion of total editors who are reverters has increased in recent years.
  • The graph (% = proportion) shows the opposite, the bar is being reduced.
• 4. Since its introduction in 2006, the use of CAPTCHA for non-autoconfirmed users has not reduced the number of reverts as a percent of edits.
  • How do you know that it has not reduced reverts? CAPTCHA blocks edits, so perhaps, if you disable it, you will see a suddenly growth of the graph.

-- Emijrp 20:45, 1 January 2012 (UTC)

The sentence "the use of CAPTCHA for non-autoconfirmed users has not reduced the number of reverts as a percent of edits" is inequiocally what the graph shows. What I see you disagreeing (please correct me if I'm wrong) is that the conclusion text implies that CAPTCHA is not useful to hold back revertable editions, right?

Well, let me defend this point of view: The graph show that before CAPTCHA being introduced, there were already some cycles of ups and downs between 5% and 8%, since April 2006. After CAPTCHA (about January 2008), the tendency of cycles until nowadays remained basically the same, growing-reducing-growing-reducing, no major changes on that. We could expected a gradual growth or reduction, but not the same tendecy of cycles. So, on the long-term, my conclusion is CAPTCHA didn't have the effect originally expected. On the short-term, the same can be said: there was a "major up" in the three months right after its introduction. So, not even the impact of seeing a "new" barrier had stopped revertable editions.

Almost one year ago, I proposed some kind of test for about one or two months without CAPTCHA, for analysis on short-term and a later definite decision based on the data collected after that. But the pure fear was stronger (were used expressions like "be prepared for the flood of block requests", "if you guys want to fiddle while Rome burns", "my sadic, dark and destrutive side is interested to see this experience, by the same joy of seeing a hurricane, a sinking or a fire"). What a pity! =( Kleiner 21:14, 6 February 2012 (UTC)

What is a revert?

The page says "Revert: Any edit that returns text back to a prior version", but how was this determined in practice? Was the whole content of each revision compared to each of its previous revisions? Or the statistic it based on the edit summaries? Helder 18:05, 5 June 2012 (UTC)

Hi Helder! I checked with Fabian, who did this research, and he said "A revert is detected by comparing the MD5 checksum of the full revision text, using these scripts. Does that help? Jwild (talk) 16:06, 11 June 2012 (UTC)

Yep! I think that helps. Thanks for the information! Helder 12:05, 12 June 2012 (UTC)