Let's see if there are smart people here. Is the YouTube algorithm biased? Can it be wrong about your video(s)? The answer is yes. But why? The video attached is a visual representation of that using an ant simulation (which is very close to how the algorithm actually works).... - Ants represent viewers - Green squares (food) represent videos I created a path bias purposefully to help you out. Take the time to analyze it, the first person who finds out will be gifted 3 months free in my private Discord community (worth ~$200). A detailed post to explain why and how to "fight" against these biases is coming soon.show more

Looks fun and abstract. I’ll give it a shot at interpretation, not necessarily because I’ll be correct, just because I think it’s interesting and I’m woefully baited by this kind of thing. - This is essentially a kind of physics simulation with two magnets. The left magnet is “source of views”. - The right magnets are defined as videos, but their x-axis represents time. - The initial path is curved downwards, which simulates more demand for the lower videos on the y-axis, as opposed to the ones higher up on the y-axis. This demand may be represented by quality of the video, or it may be because of video topic. - The first phase of the simulation is biased towards the top row, with lots of videos posted in quick succession. This might be when someone’s discovered a new niche and posts quantity > quality. There’s more magnets (videos) there, so videos don’t really have all much choice apart from to go to those videos which exist. - As time goes on, the path biases towards the direction of least resistance (highest quality or most in-demand topic), and then the shared views are subtracted from the cruddier “quantity vs quality” channel. So really, this simulation represents gradients in physics, which applies to people and views, and the arbitrary effect of magnets which represents videos in space/time. The ultimate winner of views is simply the one that ends up being on the path of least resistance (highest demand-supply asymmetry, because that’s actually a gradient too). Lastly, I believe super nerds might call this a variant of a multi-armed bandit called “knapsack bandit”. But that’s getting towards the edge of my knowledge lol as I’m not a statistician nor a computer scientist by trade. (Opinions on above thoughts held loosely, this is a post-drinks midnight write-up since an espresso martini put me in the mood for it)

I think it's if you upload a video onto a channel with fewer videos than another channel , The algorithm Is more biased towards the channel with more videos because of returning viewers, watch-time Etc because it keeps viewers on the platform for longer , even though the channel with fewer videos uploads a video that shows positive signals they can't keep viewers on the platform as long as the channel with more videos even if the channel with viewer videos shows stronger positive signals to the algorithm. So the algorithm is more biased to the channel with more video purely because of watch time , returning viewers and can retain viewers on the platform for longer .

I'm assuming each line on the right represents a different channel, and how close the green squares are to each other represents how similar they are in terms of niche and who they could interest The first square for each line is a successful video, and if that is similar enough to other videos on the channel, those are gonna get a lot of watchtime and manipulate the algorithm that way Meanwhile if the successful video differs from the other videos (example, last line where the other squares are very far from it ), it having big views doesnt immediately translate in big channel growth ( but, since on the list line all the other squares are very close, once the 2nd video of that section picks up, it skyrockets in visibility and starts getting favored by the algorithm at a faster rate)

my interpretation is that the closest, most available source will be prioritized, the more it's prioritized the more it gets prioritized, because ants leave a path for other ants, in terms of videos, that is data. The more data for a video there is that it's good, the more YouTube will recommend it. (I think one of the reason for that is that the algorithm always makes a bet, and it would rather make a safe bet than risk testing another video that may not give the same results) In other words, you're more likely to be recommended something that majority of others(like you) watch. At times you may wander and find another source (video) but that is less likely to happen and random.

YouTube's recommendation system can unintentionally create a feedback loop of perpetual bias by prioritizing videos from channels with proven track records over others. This happens regardless of whether those videos are the best choices for viewers, potentially limiting exposure to more satisfying content. This raises the question of how effective satisfaction surveys are in mitigating this bias.

First it goes to the "path of the least resistance", so closest to what it can find. Then it continues to try to look for "food" in these paths, but because it's not "constant", they move on to the next paths that have the most constant "source of food". They still continue to revisit previous paths to "food" but more as a "scouting purpose". And, what's interesting is that they will get "tired" of the same "repetitive" source of food, and will start to look for more "variety", while prioritizing sources that have slight "pauses".

The single green square and then air? They clear the first one before reaching the last ones.

From what I could deduce: YouTube recommends longer videos first, but the ones having super frequent uploads get more views. However the one making decently long videos and posting a bit infrequently gets more views overtime

The space between green boxes represents video upload frequency. Then if you upload more consistently, YouTube algorithm will have less margin of error. That's what video says i think.

Early traffic is highly concentrated on a few paths due to early viewer engagement. Over time, ants begin to explore more path.Paths with higher video density start to attract more traffic. Paths with more videos consistently attract more traffic.