Upcoming Twitch embeds experiment

This would work fine for e.g. YouTube where most embeds are just a short clip, and then the user moves on to the next, but a single Twitch embed is in most cases meant to be watched for more than 1 minute.

If this change really goes through, you might as well just remove the embed altogether. I for sure will remove it from my site, as it will be completely pointless to use.

While I don’t have a direct use-case for input here, I am curious from a developer standpoint why Twitch thought implementing an experimental feature across all users without an ability to opt-out was the best choice?

Given that it directly impacts all embeds and breaks many of the above use cases, shouldn’t this have been something that’s tested for those opted into beta testing of features? This is very common practice, even your parent company utilizes this practice, along with plenty of others as it’s standard practice to do so with features that directly impact & change the required interactions with your main media player.

This is not to say there’s never a case to A/B test on a live site / experiment with minor changes in production such as style, layout, CTA, etc., but major functionality changes don’t fall in this group.

1 Like

You just compared what a e-commerce store does to a video stie. Sellers on Amazon will want to paticipate in experiements as it may increase sales, and give them an edge over other sellers on the platform.

Most of the people on this thread vary in size of traffic and use case, but uniformly no one would turn it on. Most want a option to disable it.

So, if it was opt-in, no one on this thread would opt in and then Twitch would have no data.

What you link to (loosly) refers to allowing merchants to run A/B split tests on their products on the Amazon Seller Market place.

Do I use this description/that description, do I offer free shipping here, do I move the buy button somewhere else. Amazon SellersCentral is giving choice to the merchants that use Amazon to sell their own products.

It’s a unfair comparsion as you are comparing two completely differeht things.

So you’d expect sites like multitwitch, kadger, raredrop, juked etc (some of the high traffic big guys) to opt in to these tests? Where the end result would be Reddit going “oh such and such a site is getting these popups, anyone know a site that isn’t getting popups” so Twitch has to go “do it to everyone” in order to get any test data.

Personally I agree we need a switch/option so that developers have choice, but Twitch needs to collect test data/measure effectiveness that doesn’t involve the end user, leaving site a (look on reddit for a fix) for site b instead of “paticipating” in the test by closing the dialog (if you get that test version) or moving from the embed to Twitch.

How and what to test is tough to decide, hence the test is across all embeds in order to collect enough statistically significant data.

Some of the “use cases” poeple highlight, they never want you to leave the site the embed is on, hence this is disruptive to their business.

So that, in my opinion, is why it’s not opt out.

It’s hard to test something “disruptive” on a opt-in basis.


That’s not entirely fair… only the people who are disrupted by this experiment come to this topic to complain. It is entirely possible (albeit unlikely) that the majority of embedded player users would not opt out. Also, ‘the percentage of people who opt out’ would probably also be a useful metric.

Yes I should of noted this in my reply. Generally it’s only the disrupted that will post as those running embeds that the experiment is not disruptive to won’t notice as the users won’t complain, the viewer will either move over to Twitch, or close the dialog or leave the site.

My “uniformly” refers to the people on this thread

Now, the counter point here is that if Twitch published a opt in experiement, there is no way to publicise it to get people to turn it on, since you don’t need a clientID to generate an embed.

So Twitch doesn’t have email address’s for every developer that is utilising Embeds. So there is no way to get sufficent penetration of the Developer Market to get people to turn it on.

If TwitchDev Tweeted about it on Twitter, how many developers will even see the Tweet? With some people I’ve been talking to about this experiement, turns out the don’t even follow TwitchDev on Twitter. (And yes the Tweet about this experiement was sent after FishStix’s tweet, and his post generated more activity, since he linked to this thread directly)

So the test would fail due to “lack of data” as no one would know about the experiement to turn it on. Which would then result in a bad business decision due to lack of data (whichever way that decision went)

Edit: 'course this is me taking the unpopular opinion route and trying to look at everything from all sides and all angles. And trying to land right down the middle


so Twitch has to go “do it to everyone” in order to get any test data.

I don’t know if this is accurate, our users are continuously pointing towards a site who are not currently having this happen, I think this roll out is happening on multiple levels (site and pop up configuration), which is honestly the next logical step of Twitch Embedded Player Migration - Timeline Update

We don’t know the parameters of the test, so Twitch may have exlucded that site to not have them as part of the test (as a control group) or other agreement in place between Twitch and that site.

If the former is valid test data, if the latter, who knows.

Again, trying to look at both sides here.

Have you had/noticed this happen on https://www.cohhilition.com/ ?

I’m wondering if this is another site that received an exception, which would lead me to think that streamers with personal websites are being exempt.

As I said, it I received a pop up. My apologies. I just wish this was a more clear process as it feels very aggressive.

Haven’t tested, and technically that embed is below the minimum width that the embeds support (it’s 340 now and that one is 300). And I need to update the embed to meet this change


Course at this dimension you can’t really watch anyway and the whole point of the embed there is a easy way for viewers to know if cohh is live and follow the CTA (that is below the title)

Gotcha, this is a bit different than from a site like https://www.destiny.gg/bigscreen

I haven’t heard of anyone reporting one from this site, and it would be hard for me to believe they came to an agreement with an un-partnered streamer, which is why I think the targeting is happening on a few levels.

Edit: we specifically had a few users report that they could watch the stream there but not on our site, thus leaving our site.

So, this has a TOTALLY different experience when you are not LOGGED IN, which I didnt test out the other time.

First - In some websites, IT KILLS the stream after a period of time with ZERO way to remove the “WATCH ON TWITCH”

This is kinda turning a EMBED into a UNPAID Advertisement, If this continues, people like myself will just Switch to another Video Platform.

Some embeds have the PURPLE Screen issue, Some Show the CONTINUE TO WATCH, some freeze the screen, this test is all over the place…

Can’t wait until its over.

1 Like

As noted in this reply in this thread here

There is more than one variation that appears to be part of this test. So would seem that Twitch is also A/B split testing different varations as part of this test

At last count there are four known variations, the three in the linked reply and the one from my own (in a really small embed - Upcoming Twitch embeds experiment - #57 by BarryCarlyon)


I’ve tried to read through all posts here, but I couldn’t find an answer how long until the CTA appears.

If it is immediately, I am wondering how this will impact Twitter Player Cards and Twitter’s upcoming Twitter Media Studio features, where large companies and organizations will be able to embed video streams directly into Tweets.

Twitch should reach out to large events and companies such as Riot Games that has millions of views on their streams and ask them if they would appreciate (data on conversion) a CTA button on their embedded streams.

The timings for how long till it shows, and/or/if auto hides afterwards, is probably part of what is being tested.

For all we know, they already did

1 Like

Such a popup actually makes me stop watching a stream.
You completely break sites like speedrunslive…

1 Like

How are we meant to host tournaments with multiple twitch embeds? The purple screen appears way too often and makes the embeds completely unreliable.

This experiment ruined a lot of things and we are now redirecting our players on our own RTMP server to avoid these purple screens of death.

What the fuck Twitch?


That is the way you should run tournaments were you want to include other peoples video, as then you get the most flexibility to handle how you process and consume the content, without the twitch embed player controls getting in the way (lovely) or anything else occuring such as this experiement or other issues beyond your control that the platform may enforce.

A custom RTMP server (that may forward the ingest to the streamer/paritipants own channel or they send game only to the custom ingest and send the full shebang (overlays etc) to their channel), is potentially the more professional and maximum flexibility for a tournament or other eveny where you combine feeds, without having to rely on someone elses video pipeline, it can also reduce delay on your combined production on consumption of the content from the particpants, reducing desync when you cut between two players in the same game. For example.

Yes this is distruptive to your use case, but there is an argument to say that your use case shouldn’t be using embeds.

And please keep it civil on the forums. There is no need to throw swear words around please!

1 Like

We always used Twitch in the past for that and it was really smooth in every way.
Setting up a RTMP server can be really hard and sometimes expensive. Our community is relatively small and we just cannot invest that much for every tournament we host.

Controlling the Twitch Embed with Vue.js to make it show on a Browser Source is painless and really easy and now, I have to use a HTML5 player. It would’ve been 10x simpler if your experiment didn’t affect the embeds like right now.

I understand that it is a big market and a CTA can sometimes solve some issues but in this case, it ruins a lot of things especially for small communities willing to host tournaments.

I really hope that this option will be available to be turned off on the creators’ channel settings as it would make you all look bad in my opinion.


Hi there, thanks for your additional feedback on this particular use case. For additional clarity, was the purple screen considered to be the more invasive part of the experience for your particular tournament? Was it clear that it would end after a certain period of time? There is additional information earlier in the thread, but that is an ad-experience solution that is different from but tangentially related to this experiment.

It is still our aim to use the info from this test to allow smaller communities to benefit from the ease of embed player implementation while also increasing more monetization and discovery opportunities for our creators. Again, thanks for your feedback and perspective.

1 Like

Yep, the purple screen was the most invasive part of the experience. It would clear after a short period of time but would return after a while.

Thanks for hearing me out. :slight_smile: