How YouTube's Algorithm Really Works (theatlantic.com)

YouTube wants to recommend things people will like, and the clearest signal of that is whether other people liked them. From a report: Pew found that 64 percent of recommendations went to videos with more than a million views. The 50 videos that YouTube recommended most often had been viewed an average of 456 million times each. Popularity begets popularity, at least in the case of users (or bots, as here) that YouTube doesn't know much about. On the other hand, YouTube has said in previous work describing its algorithm that users like fresher content, all else being equal. But it takes time for a post to build huge numbers of views and signal to the algorithm that it's worth promoting. So, the challenge becomes how to recommend "new videos that users want to watch" when those videos are new to the system and low in views. (Finding fresh, potentially hot videos is important, YouTube researchers have written, for "propagating viral content.")

Pew's research reflects this: About 5 percent of the recommendations went to videos with fewer than 50,000 views. The system learns from a video's early performance, and if it does well, views can grow rapidly. In one case, a highly recommended kids' video went from 34,000 views when Pew first encountered it in July to 30 million in August. The behavior of the system was explicable in a few other ways, too, especially as it adapted to making more clicks inside YouTube's system. First, as Pew's software made choices, the system selected longer videos. It's as if the software recognizes that the user is going to be around for a while, and starts to serve up longer fare. Second, it also began to recommend more popular videos regardless of how popular the starting video was.

Comments

No not really

[Anonymous Coward]

I look up one video on how to fix something and all i get now are howto videos. I just needed one video to fix it. Its fixed stop sending me fixit videos and let me get back to my cat videos. Right meow

Re:No not really

[Z00L00K]

That's not the worst with the YT algorithm - the worst is that it suggests videos I already watched like a year ago. Only a rare few of them are worth watching again.

How YouTube's Algorithm Really Works

[taustin]

Very, very poorly.

Their criteria is what they think they'll make the most money on, and to hell with user preferences. I know this because I can tell it I'm not interested in a particular video, or entire category, then refresh the page, and that video or category will be back. Again and again and again, for months, even though I tell it I'm not interested every time.

not interested broken

[slothman32]

This is as good a place as any.
For the last month or so when I clicked the "not interested" box the row collapsed but when I closed then re-opened YT it was there again.
I don't like seeing the "live gaming" and other things but they keep coming back.

Re:not interested broken

[TechyImmigrant]

>I don't like seeing the "live gaming" and other things but they keep coming back.

I don't like anime, but when you do like a few Japanese bands, the algorithm has only one mapping in its hash table. Japanese --> anime.

It's a shit show right now

[SmaryJerry]

YouTube took a wrong turn somewhere with it's algorithm. Recommendations have less variety than basic on air tv channels. The algorithm always assumes you want something similar to what you've previously watched which is a major mistake. Yes I watched a clip of a late night talk show. But I don't need late night talk shows recommended to me constantly. Google has the same issue with advertising. Yes I bought a product but I don't need to buy multiple of them so stop advertising it to me constantly for a month afterwards.

I forget who

[rsilvergun]

but somebody once observed that some of the brightest minds of our generation are figuring out how to get me to watch cat videos (and the 15 second ad spot before them) instead of curing cancer. To say nothing of all the math wiz's working on quicker High Frequency Transactions for Wall Street.