Hacker News Reader

Comment by obilgic | Hacker News Reader

I operate 2 websites that are exact clones of each others, the only difference is the domain. This algorithm change literally shifted %70 of the traffic from one of them to the other one, in a matter of hours.

Edit:

    * Links to the websites, I would say almost identical
    * Same number of the pages are indexed by google, around ~5 million
    * Domains are almost same, no keyword difference
    * They both have same pagerank
    * Domains are registered together
    * Sites are hosted on different ips
    * Total traffic sites get is around ~40k/day unique
    * By this change, total unique increased by %10

View on HN

Sounds like the algorithm noticed the duplication, and picked a "winner". The new winner isn't the same as the old one.

While we are on this topic, does anyone know the current state of art for text duplication detection algos? I understand that Google used LSH but they must have made a lot of progress since.

LSH: http://en.wikipedia.org/wiki/Locality-sensitive_hashing

I'm not sure about the algorithm in use, but I what I hope is happening is that Google is now looking for the earliest publication of content when deduplicating. Most copycat sites have to copy their text from something existing, and if Google has already indexed that, they know that later versions of it are copies (and can presumably be knocked down in rank).

in a matter of hours

TFA mentions this update happened weeks ago. How are you able to pin this down to the hour and correlate your traffic changes to the algo change if we don't know when the algo change happened exactly?

Shift happened weeks ago, since then I am waiting for any kind of update from google about the algorithm change. Also I am not arguing what time they have deployed the new algorithm, I am just saying that the traffic shift happened in a matter of hours.

I think he was just hoping that you could pinpoint the change to a specific date.

I think I saw Hummingbird happen to one of my sites, but my traffic shift happened over the course of 5 weeks, with a steady ramp up in traffic. Very different from the near step-function you saw.

I doubt the incoming links are also identical.

Do you have any guesses as to why?

google used to care about matching domain and content, maybe it identified the 70% one as a more realistic name.

That's amazing. Do you have an idea what the difference is? Something like a hot keyword in the domain, or even a closely related word?

Reply on news.ycombinator.com