White House Youtube Dislike Manipulation

I’ve seen screenshots of YouTube modifying dislikes of White House videos. I decided I would do a thorough analysis myself. I wrote a script to check video stats every 80 seconds for 24 hours – for all videos on White House’s YouTube channel.

The collected data is archived here and here. The format is space-separated “CSV”, as follows:

VideoURL UnixTimestamp Date,Time Views Likes Dislikes

Here is a sample of the most egregious manipulation:

Some videos were delisted in minutes!:

https://www.youtube.com/watch?v=2bpSkdYUtNU 1611771197 01/27/2021,13:13:17      1227       437      2963
https://www.youtube.com/watch?v=2bpSkdYUtNU 1611771285 01/27/2021,13:14:45      1463       441      2999
https://www.youtube.com/watch?v=2bpSkdYUtNU 1611771372 01/27/2021,13:16:12      1763       449      3030
https://www.youtube.com/watch?v=2bpSkdYUtNU 1611771459 01/27/2021,13:17:39      2476       455      3060
https://www.youtube.com/watch?v=2bpSkdYUtNU 1611771546 01/27/2021,13:19:06      2640       459      3098
https://www.youtube.com/watch?v=2bpSkdYUtNU 1611771720 01/27/2021,13:22:00      3588       470      3183
https://www.youtube.com/watch?v=Fxo3OHKjfxs 1611699362 01/26/2021,17:16:02       918       405      4942
https://www.youtube.com/watch?v=Fxo3OHKjfxs 1611699448 01/26/2021,17:17:28      1202       412      4976
https://www.youtube.com/watch?v=Fxo3OHKjfxs 1611699534 01/26/2021,17:18:54      1375       415      5026
https://www.youtube.com/watch?v=juqHZYKzyx0 1611766646 01/27/2021,11:57:26       255       375      1771
https://www.youtube.com/watch?v=juqHZYKzyx0 1611766733 01/27/2021,11:58:53       455       380      1823
https://www.youtube.com/watch?v=juqHZYKzyx0 1611766819 01/27/2021,12:00:19       455       383      1852
https://www.youtube.com/watch?v=juqHZYKzyx0 1611766906 01/27/2021,12:01:46       819       387      1886
https://www.youtube.com/watch?v=juqHZYKzyx0 1611766992 01/27/2021,12:03:12      1148       393      1932
https://www.youtube.com/watch?v=juqHZYKzyx0 1611767079 01/27/2021,12:04:39      1462       397      1971
https://www.youtube.com/watch?v=juqHZYKzyx0 1611767166 01/27/2021,12:06:06      1830       398      2019
https://www.youtube.com/watch?v=ucvgAZG_IT4 1611770591 01/27/2021,13:03:11      1587        83      2040
https://www.youtube.com/watch?v=ucvgAZG_IT4 1611770764 01/27/2021,13:06:04      3014        95      2114

Likes+Dislikes was greater than views in some cases. Although that seems impossible, Youtube updates views slower, so they do not reflect real views at the time. For example:

https://www.youtube.com/watch?v=jw1_00uI02U 1611720090 01/26/2021,23:01:30     44404       924      8099
https://www.youtube.com/watch?v=jw1_00uI02U 1611720176 01/26/2021,23:02:56     44404       924      8118
https://www.youtube.com/watch?v=jw1_00uI02U 1611720260 01/26/2021,23:04:20     44404       925      8132
https://www.youtube.com/watch?v=jw1_00uI02U 1611720345 01/26/2021,23:05:45     44404       925      8151
https://www.youtube.com/watch?v=jw1_00uI02U 1611720429 01/26/2021,23:07:09     44404       925      8168
https://www.youtube.com/watch?v=jw1_00uI02U 1611720514 01/26/2021,23:08:34     44556       925      8184
https://www.youtube.com/watch?v=jw1_00uI02U 1611720599 01/26/2021,23:09:59     44556       925      8199
https://www.youtube.com/watch?v=jw1_00uI02U 1611720683 01/26/2021,23:11:23     44556       928      8219
https://www.youtube.com/watch?v=jw1_00uI02U 1611720768 01/26/2021,23:12:48     44556       928      8237

So it’s possible for likes and dislikes to accumulate while views stays the same. Eventually, views jumps up to better reflect reality.

The record of every time dislikes were removed is archived at https://pastebin.com/raw/F4ELDc4R

Grand Total         -130321

130 Thousand dislikes were removed in a 24hr period!

And this is for the most popular US President of all time!

Enjoy 🙂 -Zoe


Update

This research was featured in a youtube video.


Code wh.sh:

# Zoe Phin, 2021/01/26

require() { sudo apt-get install -y curl gnuplot; }

stats() {
    list=$(curl -s 'https://www.youtube.com/c/WhiteHouse/videos' | grep -o 'watch?v=[^"]*')
    for i in $list; do
        link="https://www.youtube.com/$i"
        date=$(date +"%s %x,%R:%S" | tr -d '\n')
        curl -s $link | tr -d ',' | tr '}' '\n' > new 
        grep -m 1 -o '[0-9,]* views' new > .views
        grep -m 1 -o '[0-9,]* likes' new > .likes
        grep -m 1 -o '[0-9,]* dislikes' new  > .dislikes

        paste .views .likes .dislikes | awk -vL=$link -vD="$date" '
            NF==6{printf "%s %s %9s %9s %9s\n", L, D, $1, $3, $5}'
    done
}

collect() {
    while true; do
        stats; sleep 75
    done | tee -a data.csv
}

dislikes() {
    list=$(cut -c1-44 data.csv | sort -u)

    for vid in $list; do	
        echo $vid
        grep ^$vid data.csv | awk '{
            DiffD=$6-D
            if (DiffD < 0) { 
                printf "%s %+7d\n", $3, DiffD 
                DLost+=DiffD
            }
            D=$6
        } END {
            printf "%-19s %7d\n", "Total", DLost
        }' 
        echo
    done | awk '{ print } $1=="Total" { GT+=$2 } 
        END { printf "%-17s %9d\n", "Grand Total", GT 
    }'
}

plot() {
    list=$(cut -c1-44 data.csv | sort -u)
    let n=0

    for vid in $list; do	
        let n++
        awk -vV=$vid '$1==V {print $2" "$4" "$5" "$6}' data.csv > plot.csv

        echo "set term png size 740,740
        set key top left
        set grid xtics ytics
        set title '$vid'
        set timefmt '%s'
        set xdata time
        set xtics format '%Hh'
        plot 'plot.csv' u 1:2 t 'Views'    w lines lc rgb 'black' lw 2,\
                     '' u 1:3 t 'Likes'    w lines lc rgb 'green' lw 2,\
                     '' u 1:4 t 'Dislikes' w lines lc rgb 'red'   lw 2
        " | gnuplot > example${n}.png 
    done
}

Run:

$ source wh.sh; require

Collect data:

$ collect

( press Ctrl-C when you're done )

Record of dislike drops:

$ dislikes

Generate charts:

$ plot

Version 2.0: Cleaner and grabs stats at random 60 to 120 second interval.

# Zoe Phin, v2.0 - 2021/02/20

require() { sudo apt-get install -y gnuplot; }

collect() {
    url="https://www.youtube.com"
    while true; do
        for vid in $(wget -qO- "$url/c/WhiteHouse/videos" | grep -o 'watch?v=[^"]*'); do
            wget -qO- $url/$vid | egrep -o '[0-9,]* (views|likes|dislikes)' |\
            sed -n 1~2p | tr -d '[:alpha:],\n' |\
            awk -vL=$url/$vid -vD="$(date +"%s %x,%R:%S" | tr -d '\n')" '
                NF==3 { printf "%s %s %9s %9s %9s\n", L, D, $1, $2, $3 }'
        done
        sleep $(seq 60 120 | shuf | head -1)
    done | tee -a data.csv
}

dislikes() {
    for vid in $(cut -c1-44 data.csv | sort -u); do	
        awk -vv=$vid 'BEGIN { print v } $1==v { 
            Diff=$6-Last
            if (Diff < 0) printf "%s %+7d\n", $3, Lost+=Diff 
            Last=$6
        } END {
            printf "%-19s %7d\n\n", "Total", Lost
        }' data.csv
    done | awk '{ print } $1=="Total" { GT+=$2 } 
        END { printf "%-17s %9d\n", "Grand Total", GT 
    }'
}

plot() { n=0
    for vid in $(cut -c1-44 data.csv | sort -u); do	let n++
        awk -vv=$vid '$1==v {print $2" "$4" "$5" "$6}' data.csv > plot.csv
        echo "set term png size 740,740
        set key top left
        set grid xtics ytics
        set title noenhanced '$vid'
        set xdata time
        set timefmt '%s'
        set xtics format '%Hh'
        plot 'plot.csv' u 1:2 t 'Views'    w lines lc rgb 'black' lw 2,\
                     '' u 1:3 t 'Likes'    w lines lc rgb 'green' lw 2,\
                     '' u 1:4 t 'Dislikes' w lines lc rgb 'red'   lw 2
        " | gnuplot > example${n}.png 
    done
}

Published by Zoe Phin

https://phzoe.com

66 thoughts on “White House Youtube Dislike Manipulation

  1. Zoe, as you know, you and I have often disagreed very strongly on scientific questions. Hey, that’s why we have horse races.

    But in this case, I can only totally agree, amidst laughter at what you’ve shown, and admiration for the excellent quality of your analysis.

    My very best to you as always, disagreements or not,

    w.

    Liked by 2 people

    1. It would be sad if they were also denied a legitimate means of expression also in the ability to vote for a president.

      People who feel they are right to manipulate data will feel they are right to manipulate data everywhere.

      Liked by 1 person

  2. elegantly simple, suitable to task, clean and clear coding => awesome again. They are belittling one stack by repeatedly interrupting the flow, excellent job, great to learn these tricks Zoe, thanks so much. Meanwhile, yt only requires login to vote, so watching the vid is irrelevant, we can vote ona whim, and yt keeps tabs on minutes watched, so perhaps two hundred people watching the first two minutes will count 200 ‘views’.

    Liked by 1 person

  3. Hey Zoe,

    This analysis is great. The thing that really struck me was how well you shell program. I’m wondering if you have any resource recommendations for getting better at shell programming specifically. I’m coming from an experienced software engineering background if that’s helpful.

    Nikhil

    Liked by 2 people

  4. Check out Liberal/Left Wing “Comedian” and News Channels like Trevor Noah and Stephen Colbert. They pump UP their numbers too. By about 5 to 10 fold.

    America is NOT divided 50/50.
    It’s more like 90/10. But the Left Control the Media.

    Liked by 1 person

  5. This is great work! I’ve been telling people for weeks that I suspected the same. Here’s the proof! I’ve been spreading this page like wildfire! Again, awesome work!

    Liked by 1 person

  6. Oh wow! Thank you Zoe for showing us how stupid those “genius” in Silicon Valley are. Only an imbecile modifies such data in bulk. At least even they know that getting them more likes than dislikes would expose their evil deeds.

    Like

  7. So very refreshing …. To see someone do great detective work … And great numbers crunching…. Thank you for the math …. Great work God speed to you …. Zoe ❤️

    Liked by 1 person

  8. The new president’s honeymoon usually lasts 100 days. Joe Biden managed to accelerate it to a week. Their interpretation of “business as usual” is “lying as usual”. The prognosis is not good.

    Zoe, a great idea, thanks.

    Liked by 1 person

  9. Excellent. Thanks. Just found your blog. Boy, are you a welcome relief from new world order science mantra. Just going over it. Sure if the NWO are right, fair enough, but they are not in many cases, and where have Biden’s voters gone to? Hmmmm. You sure that code is all right? I don’t have a clue btw 🙂

    Liked by 1 person

  10. Hey Zoe,

    Great work! While I imagine it’s not normal for the likes to be going down like this, I’d be curious to see what a typical YouTube video looks like. Do you have any links to examples of that that are similar to your work here to see a good side-by-side? Just curious if it’s normal for likes/dislikes to ever under any circumstance start going down only to rise again. Thanks!

    Like

  11. You point out the “drop” at 1am Eastern – but did you also notice the smaller but equivalent %age drop 8 hours earlier?

    It looks to me like shift-change block censorship – switching to auto-censorship after the night shift come on…

    Like

  12. Hi
    First of all I should applaud you for your effort to denounce data manipulation on Youtube. This is very interesting!
    However, it seems to me that the implicit conclusion from this analysis, that the manipulation is politically motivated, is FLAWED!!! I’m not saying it isn’t true, only that you haven’t shown enough data to conclude that.

    It is possible that YT is employing a general algorithm to prevent the misleading effects of “dislike mobs”, as discussed here: https://www.theverge.com/2019/2/1/18207189/youtube-dislike-attack-mob-review-bomb-creator-insider
    Don’t get me wrong, that’s still condemnable if it’s done without transparency! What I’m trying to say is that before rushing into conclusions that these actions are politically motivated, we should do a “control experiment” with non-political YT videos, which could be targets of “dislike mobs” as well, like product reviews, and test if the same effect is observed.

    Also, even though a lot of people could be in bed at 1 am EST, on the West coast the time would be 10pm. Not to mention that there’s a considerable cut to ~50% of the dislikes happening at 6pm EST on the second plot. Can you argue that most people are asleep then? If they wanted to hide, why not something 5am EST/2am PST? And that would be effective in the US only.

    Finally, I want to emphasize that I do think that the analysis you have done is extremely important, especially if it ends up demonstrating clear political bias in a huge platform as YT. But we should all make an effort not to look at data from a biased perspective too.
    Regards

    Like

    1. Fair points, Joao.

      But remember,

      “YouTube wants ‘dislike mobs’ to stop weaponizing the dislike button”

      Is also a political position. Not unlike banning channels, 95+% of which has certain political views.

      Alphabet, Inc seems to direct which way cancel culture works.

      It’s pretty obvious that they made enormous in-kind donations to political parties that handed them critical gov posts.

      It’s all rational from their point of view, but is it good for society?

      Liked by 1 person

      1. Hi Joao!

        At last a critical view! What’s the point of analysing data you got from a source you don’t trust. And then even try to infer anything from that analysis. In spite of that they should test their hypothesis and not just take one particular channel and then claim that these “manipulations” are political. Also, what about the manipulations of the likes? Although he proportion of removed likes is much smaller than the the proportion of removed dislikes, why would they do that if their goal is to present the channel in a better light.

        I is awesome that Zoe shows us all these data and even giving us the tools to do our own research. But these data itself are by far not enough to conclude anything. Does youtube also remove likes/dislikes on other videos/channels? Is the drop at 1 am EST seen on other channels? Under which circumstances does that happen? If you correct the dislike data by adding the removed dislikes, why does the data not reflect US day cycles?

        Regards

        Like

    2. Hi Joao!

      At last a critical view! What’s the point of analysing data you got from a source you don’t trust. And then even try to infer anything from that analysis. In spite of that they should test their hypothesis and not just take one particular channel and then claim that these “manipulations” are political. Also, what about the manipulations of the likes? Although he proportion of removed likes is much smaller than the the proportion of removed dislikes, why would they do that if their goal is to present the channel in a better light.

      I is awesome that Zoe shows us all these data and even giving us the tools to do our own research. But these data itself are by far not enough to conclude anything. Does youtube also remove likes/dislikes on other videos/channels? Is the drop at 1 am EST seen on other channels? Under which circumstances does that happen? If you correct the dislike data by adding the removed dislikes, why does the data not reflect US day cycles?

      Regards

      Like

  13. “Did you notice the huge drop at 06 hour (1 AM US Eastern)? Most working people go to sleep by that time. Coincidence? I think not.”

    It’s Politics, Zoe; in Politics, there ARE no Coincidences..

    Like

  14. That red line shift is remarkably similar to the red vote count shift in several Philadelphia precincts at 1:00 am. I suspect some kind of red gravitational shift worldwide.

    Liked by 1 person

  15. Not that it’d be odd for YouTube to adjust the like/dislike ratio to favor biden, just that they appear to do a poor job of it?
    Checking the cpl dozen WhiteHouse vids since the 20th and they all show a 3/1, 4/1, 7/1 dislike/like ratio. Unlike President Trump’s numbers (Trump White House Archive)

    Like

  16. There was once a dream called Liberty. Elusive and easily killed; but if nurtured and believed in could free all of humanity and unite the world.
    This is not it. This is not it.

    Liked by 1 person

    1. If Youtube is indeed doing data manipulation for political purposes then we should ask ourselves whether that is fundamentally wrong in a free market system. Isn’t that what Liberty is all about? I mean, YT costumers are free to leave the platform if they don’t agree with something. Unless people are defending more government involvement in regulating what private companies can or cannot do, which I’d actually be in favor of. But of course, that would raise many red flags for people at the right, wouldn’t it?

      Like

      1. Joao,
        Stock exchanges are free to drop any company they don’t like, right? If companies manipulate their data, they are fraudulent, no?

        If corporations work with political parties to stifle their competition, they are regulating, aren’t they? If they lie about being open and fair, then they are committing fraud, no?

        Like

  17. Here is ‘stats’ that works with zsh on a Mac…

    stats() {
    list=$(curl -s ‘https://www.youtube.com/c/WhiteHouse/videos’ | grep -o ‘watch?v=[^”]*’ | tr ‘\n’ ‘ ‘)
    for i in $(echo $list); do
    link=”https://www.youtube.com/$i”
    date=$(date +”%s %x,%R:%S” | tr -d ‘\n’)
    curl -s $link | tr -d ‘,’ | tr ‘}’ ‘\n’ > new
    grep -m 1 -o ‘[0-9,]* views’ new > .views
    grep -m 1 -o ‘[0-9,]* likes’ new > .likes
    grep -m 1 -o ‘[0-9,]* dislikes’ new > .dislikes

    paste .views .likes .dislikes | awk -v L=$link -v D=”$date” ‘
    NF==6{printf “%s %s %9s %9s %9s\n”, L, D, $1, $3, $5}’
    done
    }

    Liked by 1 person

  18. using Searx to compare DDG/Google we see how google cooks search results to hide political stuff. Example search for ruby freeman and oops google removes stuff related to her election fraud while DDG shows it

    Like

  19. Thank you very much for your incredible work. I have found you from someone mentioning you on Youtube. He replied to my comment on the White House page regarding the counts, I think…but when I clicked on the notice to read it, the comment section was turned off. Luckily I was able to read what he said under the notification so I could do some research, then found a Youtube channel talked about you and found this link from there.

    I have put the numbers of likes and dislikes with date and time and observed them dropping so I commented again a day later with the like/dislike numbers on the comment section of the White House cannel. Now the whole comment section has been removed. Today, I took screen shots of all of the comments (more than 100) which were all against Biden. I mean…not a single comment there supported him. I knew the comment section would be turned off as I have witnessed it happened before and sure enough, while I was taking the screenshots, the comment section was turned off. I had managed to take all the screenshots as I did not refresh my screen, though. I am sick and tired of them controlling our real voices.

    Liked by 1 person

  20. Nice job, symbollikely (sic) showing the reverse proportionality between voters and general approval…en français….le rapport inversement proportionnel entre le nombre d’électeurs et le taux d’approbation de la population….

    Like

    1. Soldier, if you don’t use quotes around “white house youtube dislike manipulation”, I get 982,000 hits, with Zoe in fourth place, and a couple of places mentioning her post not far below that.

      w.

      Like

Leave a Reply to Anubis de la Vallée. Cancel reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

%d bloggers like this: