Digg’s Algorithmic Mystery Tour
Hey all -
Nothing exciting to report here unless you consider testing out a new Top News algorithim to be of note. We’re rolling out an alternative algo to a small percentage of our users today (it’s a random percentage; we have no way of assigning certain users or groups to a specific algo) and there’s a chance you might notice some weirdness on Top News during this test.
One of the side effects of running two algos side by side is that you may notice the time stamps on some Top News stories are out of order. This is because a story might be promoted at, say, noon on one algo but not until 1PM on the other. When the story is promoted at 1PM, the promotion time stamp on that story will update to the later promotion time.
Now on to the good stuff. Of course, I can’t give you any hardcore details about the algo we’re testing, but I can give some high level information:
- You won’t know which algo you’re using during the testing phase
- The new algo incorporates signals from both the logged-in Digg community and Digg’s unlogged-in visitors
- We’re going to be analyzing story performance across permalinks, My News, Top News and Upcoming
- We’re going to be removing stories from Top News that aren’t performing well
- For the duration of the test, Digg will have two (or more) versions of Top News and what you see will be dependant on which algo bucket you’re in though you won’t be able to view the other version.
We’re going to continue testing and tweaking the Digg algos, so expect more experimentation from us.
Please continue to send us your feedback - we’ll be especially curious to hear if you’re seeing an improved Top News experience.
Special thanks to folks who helped design and build the new algo: Dan Huard, Scott Reynolds, Ilya Gluhovsky, Weidong Zhang and Dash Gopinath.
Thanks,
Jen
jen@digg.com