|
East Gorteau22261 Posts
Katowice 2015 | ViewershipsWe often theorize about what exactly influences viewer counts in esports. Breaks for commercials and analysis are often brought up as significant factors, while the influence of competitors' popularity on viewer counts is often disputed. For the layman, gathering the data necessary to draw any meaningful conclusion might seem like a daunting task, and solid data is rarely readily available elsewhere. Efe Karakus is a graduate student in Computer Science focusing on data mining and interactive visualizations. Efe is an avid esports fan and has developed an interest in visualizing things that often appear intangible to the average fan. Previously, Efe has broken down streamer revenue and the impact of subscriptions vs. ad revenue on a streamer's income. Now, as part of his work on a masters' thesis, he set out to analyze how viewer counts fluctuate during a broadcast, and to determine the factors that influence these fluctuations. Clicking the link below will take you to a dedicated page for his project.In the future, you can expect to find more interactive visualizations and stories about esports -- particularly League of Legends -- on Teamliquid and his website. If you like this project, consider following Efe on Twitter.
|
Sure is a lot of not Starcraft in here
|
On May 14 2015 16:18 SetGuitarsToKill wrote:
Sure is a lot of not Starcraft in here
I'm curious: do you think the results would be much different aside from viewer count sizes? You should probably read before commenting
|
On May 14 2015 16:20 Zealously wrote:Show nested quote +On May 14 2015 16:18 SetGuitarsToKill wrote:
Sure is a lot of not Starcraft in here I'm curious: do you think the results would be much different aside from viewer count sizes? You should probably read before commenting
No, I'm sure its similar. Obviously people watch the games all the way through and drop off during breaks. It's the same reason why the 2 hour swarm hosts games saw weird viewership rises for the streams despite being boring as hell. I didn't think we needed a pretty infographic to tell us this. Perhaps I'm just being an dick though.
|
On May 14 2015 16:26 SetGuitarsToKill wrote:Show nested quote +On May 14 2015 16:20 Zealously wrote:On May 14 2015 16:18 SetGuitarsToKill wrote:
Sure is a lot of not Starcraft in here I'm curious: do you think the results would be much different aside from viewer count sizes? You should probably read before commenting No, I'm sure its similar. Obviously people watch the games all the way through and drop off during breaks. It's the same reason why the 2 hour swarm hosts games saw weird viewership rises for the streams despite being boring as hell. I didn't think we needed a pretty infographic to tell us this. Perhaps I'm just being an dick though.
It's nice to think that some things are "obvious", it's better to confirm them with (f)actual analysis.
As for the analysis, well yeah ~20% in breaks is... a shame, let's put it this way.
|
On May 14 2015 16:26 SetGuitarsToKill wrote:Show nested quote +On May 14 2015 16:20 Zealously wrote:On May 14 2015 16:18 SetGuitarsToKill wrote:
Sure is a lot of not Starcraft in here I'm curious: do you think the results would be much different aside from viewer count sizes? You should probably read before commenting No, I'm sure its similar. Obviously people watch the games all the way through and drop off during breaks. It's the same reason why the 2 hour swarm hosts games saw weird viewership rises for the streams despite being boring as hell. I didn't think we needed a pretty infographic to tell us this. Perhaps I'm just being an dick though.
That's not really all it is, though. Quantifying the impact of popularity on viewer counts has, as far as I know, never been done before. At least not to this extent.
|
On May 14 2015 16:37 Zealously wrote:Show nested quote +On May 14 2015 16:26 SetGuitarsToKill wrote:On May 14 2015 16:20 Zealously wrote:On May 14 2015 16:18 SetGuitarsToKill wrote:
Sure is a lot of not Starcraft in here I'm curious: do you think the results would be much different aside from viewer count sizes? You should probably read before commenting No, I'm sure its similar. Obviously people watch the games all the way through and drop off during breaks. It's the same reason why the 2 hour swarm hosts games saw weird viewership rises for the streams despite being boring as hell. I didn't think we needed a pretty infographic to tell us this. Perhaps I'm just being an dick though. That's not really all it is, though. Quantifying the impact of popularity on viewer counts has, as far as I know, never been done before. At least not to this extent.
True, I suppose, though there's a lot more than popularity that goes into viewer counts, time of day/time zones, quality of games, etc, and it'd take a hell of a lot more detailed breakdown to determine all these variables. Still interesting I guess though
|
Katowice25012 Posts
On May 14 2015 16:31 OtherWorld wrote:Show nested quote +On May 14 2015 16:26 SetGuitarsToKill wrote:On May 14 2015 16:20 Zealously wrote:On May 14 2015 16:18 SetGuitarsToKill wrote:
Sure is a lot of not Starcraft in here I'm curious: do you think the results would be much different aside from viewer count sizes? You should probably read before commenting No, I'm sure its similar. Obviously people watch the games all the way through and drop off during breaks. It's the same reason why the 2 hour swarm hosts games saw weird viewership rises for the streams despite being boring as hell. I didn't think we needed a pretty infographic to tell us this. Perhaps I'm just being an dick though. It's nice to think that some things are "obvious", it's better to confirm them with (f)actual analysis. As for the analysis, well yeah ~20% in breaks is... a shame, let's put it this way.
It would be fun to get a few of these done so we can look at how different it is for different tournaments, there are definitely some that strive more for good filler content in between breaks. Lots of factors to consider of course.
|
We haven't had many tournaments with A and B streams lately, but I would have loved to see how viewership looked under that setup. I assume that B streams didn't get significant gains from A breaks or we'd still be seeing B streams. (*significant enough to warrant the resources of keeping a second stream running anyway)
I also think this might be a good reason to test out longer series (for sc2). Instead of almost every match being bo3, viewership might benefit from bo5 or even bo7 matches more frequently. I'd personally like to see longer series anyway...
|
For SC2 doing longer matches isn't so bad but it's really difficult in Dota/League/CSGO - you end up with series that take forever. Dota has a few bo7s through the year normally and inevitably one of them ends up being 8 hours or something crazy. There was that WPC finals at the beginning of 2014 that went close to 12 hours when all was said and done (there was a break in the middle for a passport issue or something though).
|
The thing that surprised me was that amount of viewers didn't always climb over the day but could actually decrease. I've always thought that streaming for multiple hours increases viewership (which it did most of the time), even if you have breaks in there.
To expand on this. The peak at the end of each game period should be higher than the peak at the previous one, at least that is the trend I tend to see in Dota events running multiple hours. It is also why it is important to run the stream a while before the actual game, so people don't skip it for something else.
|
err... and?
Must be dumb, totally missed the point of the whole article, it couldn't be just to say that viewers numbers were increasing during games and lowering during breaks? Oo
|
On May 14 2015 21:10 Gwavajuice wrote:
err... and?
Must be dumb, totally missed the point of the whole article, it couldn't be just to say that viewers numbers were increasing during games and lowering during breaks? Oo
Hi, I'm Efe the author of the article. The visualization started with three questions in mind:
1) Do viewerships increase depending on the teams that are played?
2) How many viewers do you lose when a break kicks in, or the analysis desk takes over?
3) How much time was spent on games vs. other stuff (analysis & breaks)?
Hopefully, we manage to give an answer on these problems. Yurie in the comment above had a very interesting observation:
On May 14 2015 19:11 Yurie wrote:
To expand on this. The peak at the end of each game period should be higher than the peak at the previous one, at least that is the trend I tend to see in Dota events running multiple hours. It is also why it is important to run the stream a while before the actual game, so people don't skip it for something else.
So hopefully, you'll be able to discover interesting insights as well! 
|
What I found interesting was that the League reddit bitched about long breaks etc but breaks were shorter than in cs:go and you had more game time.
|
This is quite correct.
I see no reason to watch analysis desk(they do nothing but to tell you what you just watched) or wait for the next match so I just close the tab asap. Also don't really enjoy watching a game if one of the teams is much stronger than the other. It's just one side getting rekt, no competition.
Who like breaks anyway? I just want constant action.
|
On May 15 2015 00:29 Invoker wrote:
This is quite correct.
I see no reason to watch analysis desk(they do nothing but to tell you what you just watched) or wait for the next match so I just close the tab asap. Also don't really enjoy watching a game if one of the teams is much stronger than the other. It's just one side getting rekt, no competition.
Who like breaks anyway? I just want constant action.
I like breaks!
But f**kin' esports breaks are longggggggggg........
|
This analysis is great and we need way more of this kind of stuff! But SetGuitarsToKill is right, where's the Starcraft in this post? I don't get why this is on the front page of the main site when it's LoL and CSGO focused...
|
East Gorteau22261 Posts
On May 15 2015 00:51 looknohands119 wrote:
This analysis is great and we need way more of this kind of stuff! But SetGuitarsToKill is right, where's the Starcraft in this post? I don't get why this is on the front page of the main site when it's LoL and CSGO focused...
TL has a CS team, regular tournament coverage, and a front page box dedicated to CS:GO. Further, this news post actually isn't tagged as Starcraft, but rather just "TL". It might be a bit confusing, but I tried to make as unambiguous as possible. But even then, although the analysis did not use data extrapolated from the Starcraft stream, the conclusions are very relevant to most esports titles (SC, Dota, HS) since I doubt things differ much between games.
|
Interesting stuff. Thanks for doing this!
|
On May 15 2015 02:34 Mag1c wrote:
Interesting stuff. Thanks for doing this!
Thanks for liking it :D
|
GRAND OLD AMERICA16375 Posts
|
As someone who loves to see this kind of stuff. It was well written for the general public to understand. Kick *&^!^ bro!?
|
On May 14 2015 21:36 malahay wrote:Show nested quote +On May 14 2015 21:10 Gwavajuice wrote:
err... and?
Must be dumb, totally missed the point of the whole article, it couldn't be just to say that viewers numbers were increasing during games and lowering during breaks? Oo Hi, I'm Efe the author of the article. The visualization started with three questions in mind: 1) Do viewerships increase depending on the teams that are played? 2) How many viewers do you lose when a break kicks in, or the analysis desk takes over? 3) How much time was spent on games vs. other stuff (analysis & breaks)? Hopefully, we manage to give an answer on these problems. Yurie in the comment above had a very interesting observation: Show nested quote +On May 14 2015 19:11 Yurie wrote:
To expand on this. The peak at the end of each game period should be higher than the peak at the previous one, at least that is the trend I tend to see in Dota events running multiple hours. It is also why it is important to run the stream a while before the actual game, so people don't skip it for something else. So hopefully, you'll be able to discover interesting insights as well!
1) yeah but it you only give "much more viewers" as a result. did you effectively a model with teams as factors and viewers as outcome, did you analyse each teams impact (like in a tukey test for example) what is the IC you obtained, did you run a fisher test or another method? did the model take time of the day as a factor?
2) did you try to estimate a percentage of loss? what did you obtain? with what IC? was it dependant on team/time/stage of competition?
3) Ok with this one, not totally sure with what to do with this information though
Edit : I'm not saying you're work is not good, just curious to see what happens when you go beyond the raw data...
|
Is Mr. Karakus involved with the upcoming ESL movie in any way? Just wondering.
|
I find the analysis desk boring with (usually) apollo, artosis, (insert random player/caster) etc.
If you want pure viewership and fun enjoyable environment you want casters/players like MC/Dragon/Parting.
And the "serious" tasteless/artosis @ non-GSL events is not nearly as fun/funny as the GSL tastetosis.
Just my opinion as a long time sc2 viewer anyway.
|
On May 15 2015 10:34 Circumstance wrote:
Is Mr. Karakus involved with the upcoming ESL movie in any way? Just wondering.
haha no I am not  I am about to graduate from grad school. I just built this interactive because I thought it could be a fun project.
|
well this is awesome... an actual analysis... with actual data... about esports =)
On May 14 2015 16:13 Zealously wrote: Now, as part of his work on a masters' thesis, he set out to analyze how viewer counts fluctuate during a broadcast, and to determine the factors that influence these fluctuations. Clicking the link below will take you to a dedicated page for his project.
Wait, what? You actually managed to convince, what I assume must be, Computer Science Professor to let you do master thesis about esports. I certainly hope you never try sell me your car  .
Joking aside I'm seriously interested. I assume you had to write a proposal for this. Was all that really mattered that you had a way to gather enough (viewership) data, a difference in content (games, breaks/ads, analysis) and the questions to ask about it? Or did you have to obfuscate what you would be analyzing with something like 'effects of breaks and ads on streamed content'?
|
On May 15 2015 17:46 nachtkap wrote:well this is awesome... an actual analysis... with actual data... about esports =) Show nested quote +On May 14 2015 16:13 Zealously wrote: Now, as part of his work on a masters' thesis, he set out to analyze how viewer counts fluctuate during a broadcast, and to determine the factors that influence these fluctuations. Clicking the link below will take you to a dedicated page for his project. Wait, what? You actually managed to convince, what I assume must be, Computer Science Professor to let you do master thesis about esports. I certainly hope you never try sell me your car . Joking aside I'm seriously interested. I assume you had to write a proposal for this. Was all that really mattered that you had a way to gather enough (viewership) data, a difference in content (games, breaks/ads, analysis) and the questions to ask about it? Or did you have to obfuscate what you would be analyzing with something like 'effects of breaks and ads on streamed content'?
Well in my days, you could take any subject that had data qualified enough to let you use enough analysis and visualition tecniques. It didn't matter much if the subjects were customers, cars, crops, hair styles or bacteriae
|
This is some good stuff even if it isn't directly looking at SC2.
I was wondering how accurate and reliable things are and if this trend is seen in other events around the world or even more regular tournaments like proleague or GSL. You have stream numbers here but as with all online content it is hard to show how many people are 'truly' watching. I know I'm guilty of wandering off/away from the PC while breaks are on, coming back for the last of the analysis or even the beginning of the game as well factoring multiple viewers (ie barcrafts/friends). I am unsure as to how commercial television obtains it's numbers perhaps there would be a similar equation that could be applied here.
Another interesting little number would be being able to see where viewers are located across multiple regions and multiple events. Oh the possibilities are endless. Thanks for the info. The twitch revenue was particularly eye opening.
|
On May 15 2015 17:46 nachtkap wrote:well this is awesome... an actual analysis... with actual data... about esports =) Show nested quote +On May 14 2015 16:13 Zealously wrote: Now, as part of his work on a masters' thesis, he set out to analyze how viewer counts fluctuate during a broadcast, and to determine the factors that influence these fluctuations. Clicking the link below will take you to a dedicated page for his project. Wait, what? You actually managed to convince, what I assume must be, Computer Science Professor to let you do master thesis about esports. I certainly hope you never try sell me your car . Joking aside I'm seriously interested. I assume you had to write a proposal for this. Was all that really mattered that you had a way to gather enough (viewership) data, a difference in content (games, breaks/ads, analysis) and the questions to ask about it? Or did you have to obfuscate what you would be analyzing with something like 'effects of breaks and ads on streamed content'?
Hi Haha I think there was a misunderstanding, I did my thesis on certain aspects of esports but this project was just purely for fun. So no proposal or IRB approval was necessary for the project. The data collected didn't need any sort of obfuscation because it's public. You can make requests to the Twitch API (https://github.com/justintv/Twitch-API) to get viewership numbers. Hope this helps!
On May 16 2015 09:58 Japetto wrote:
This is some good stuff even if it isn't directly looking at SC2.
I was wondering how accurate and reliable things are and if this trend is seen in other events around the world or even more regular tournaments like proleague or GSL. You have stream numbers here but as with all online content it is hard to show how many people are 'truly' watching. I know I'm guilty of wandering off/away from the PC while breaks are on, coming back for the last of the analysis or even the beginning of the game as well factoring multiple viewers (ie barcrafts/friends). I am unsure as to how commercial television obtains it's numbers perhaps there would be a similar equation that could be applied here.
Another interesting little number would be being able to see where viewers are located across multiple regions and multiple events. Oh the possibilities are endless. Thanks for the info. The twitch revenue was particularly eye opening.
I think all these questions are right on point. Unfortunately, there is no real way of measuring if someone is "really" watching  or just have the stream on in a tab or just listening. Perhaps if you work at Twitch you might be able to see if the stream is right now on full screen and judge that the person is really watching that way.
The way TVs measure things are even more inaccurate as far as I know (http://en.wikipedia.org/wiki/Nielsen_ratings)
The location would again be definitely fun to see, but that data is private and not provided by the API for now
Thanks for the great thoughts everyone!
|
On May 16 2015 13:13 malahay wrote:Hi Haha I think there was a misunderstanding, I did my thesis on certain aspects of esports but this project was just purely for fun. So no proposal or IRB approval was necessary for the project. The data collected didn't need any sort of obfuscation because it's public. You can make requests to the Twitch API (https://github.com/justintv/Twitch-API) to get viewership numbers. Hope this helps!
I did mean the thesis (proposal). It was a tongue in cheek question. I wouldnt have thought that a thesis proposal covering esport would get approved. What is the thesis about and is available anywhere (arxiv, pubmed, or something similar)?
|
On May 17 2015 23:04 nachtkap wrote:Show nested quote +On May 16 2015 13:13 malahay wrote:Hi Haha I think there was a misunderstanding, I did my thesis on certain aspects of esports but this project was just purely for fun. So no proposal or IRB approval was necessary for the project. The data collected didn't need any sort of obfuscation because it's public. You can make requests to the Twitch API (https://github.com/justintv/Twitch-API) to get viewership numbers. Hope this helps! I did mean the thesis (proposal). It was a tongue in cheek question. I wouldnt have thought that a thesis proposal covering esport would get approved. What is the thesis about and is available anywhere (arxiv, pubmed, or something similar)?
It's not out yet unfortunately. I think the university makes it public two years after submission I was looking into comparing certain aspects of esports and traditional sports. The two big visualizations that were part of my thesis were:
* LCS Fandom Shift: http://efekarakus.github.io/esports/fandom/ (for LoL)
* Twitch revenue: http://efekarakus.github.io/twitch-analytics/#/revenue
And some other things like what retired professionals do that are just static images and not interactive on the web.
|
I don't think the drop in breaks tells the whole story. We do not watch breaks usually, but we just mute the stream and put it into the corner of a secondary screen, not close it, so that we catch the next game - I would guess many people behave similarly, so the real viewership outside the games could be much lower than indicated.
|
|
|
|
|
|
EPT
