“TheEndless” wrote:
I don’t think we forget about the Roku 2 HD and the Roku LT 2400, we just overlook them. They had the shortest lifespans of any other Roku players, so the likelihood of there being a ton of them out there is low enough that it’s probably safe to overlook them.
Speaking of which, do you have recent statistics of model# usage in the wild that you can share with us?
Like from the past quarter/this year, percents of the pie (no raw numbers).
I wonder how things are - i can dig my own numbers but my sample is small and somewhat biased. The Co has the numbers for sure but likely can answer only after addressing the backlog (or never - whichever comes first).
“TheEndless” wrote:
I don’t think we forget about the Roku 2 HD and the Roku LT 2400, we just overlook them. They had the shortest lifespans of any other Roku players, so the likelihood of there being a ton of them out there is low enough that it’s probably safe to overlook them.
Speaking of which, do you have recent statistics of model# usage in the wild that you can share with us?
Like from the past quarter/this year, percents of the pie (no raw numbers).
I wonder how things are - i can dig my own numbers but my sample is small and somewhat biased. The Co has the numbers for sure but likely can answer only after addressing the backlog (or never - whichever comes first).
Probably shouldn’t derail this thread with this, but… Here’s what I’m seeing, since Jan 1st, for one of my more popular channels. I don’t know if this is indicative of all channels or not, and it’s also worth noting that this one isn’t available to legacy devices…
The most popular by a significant margin (as one might expect?) are the Roku 3 and the HDMI Streaming Stick…
*854x480 is how I record 16x9 SD, so I can gauge how many SD users are only using SD, because they most likely don’t have an HDMI cable to hook it up properly.
Most interesting i think is the observation that legacy players (firmware 3) have dropped to 10%, which likely was the cut-off for RokuCo un-supporting them. It took almost two years to go down from 20-25% (my estimate from 2013-08). I am guessing that happened mostly through dilution by new sales and less by retirement.
Also i am puzzled by the difference between the % of Roku3 players in TheEndless’s vs my data. He saw 39% of the post-legacy, me - 22% (20% corrected /90%). That’s a pretty big difference, almost twice. His sample is bigger but mine is big enough to rule out statistics random error too. So i think one (if not both) of us has systematic bias in the sample. In other words, there is a reason why his pool has unusually high % of Roku3 or mine has untypically low % of the same.
TheEndless - can you think of a reason why the channel from which your stats are might be unusually popular on Roku-3 players? (Say maybe it shows higher in Channel Store listings of Roku3; knowing that sort order differs per player)
Both of you should do the same measurements again, over a unique period of time, and see if your results are consistent. The power of random is great. :idea:
“EnTerr” wrote:
TheEndless - can you think of a reason why the channel from which your stats are might be unusually popular on Roku-3 players? (Say maybe it shows higher in Channel Store listings of Roku3; knowing that sort order differs per player)
Mine’s both a video and a 2D channel, which could potentially lead to some performance issues on lower end boxes, that might steer some users away… maybe.. ? It’s also had a lot of promotion that might be unevenly aimed towards the higher end boxes. Hard to say, but I’m sure the content plays a role.
“TheEndless” wrote:
Mine’s both a video and a 2D channel, which could potentially lead to some performance issues on lower end boxes, that might steer some users away… maybe.. ?
Was your data filtered per unique installs - or was the same install/device/user be counted more once if they used the channel multiple times in that time period? The latter will be prone to usage patterns affecting the distribution stats (i.e. app likely is more frequently used on boxes where it “runs better”). Mine counted only once each user/player (akin to SELECT DISTINCT or GROUP BY). And btw i am thinking it’s more likely my data be tainted, i just can’t put my finger on the rhyme or reason.
“Komag” wrote:
Both of you should do the same measurements again, over a unique period of time, and see if your results are consistent. The power of random is great. :idea:
Nah, not worth it. His data was for months 1-2-3-4, mine - 3-4-5; two months overlap, plus distribution does not change rapidly. A better thing will be if instead we see a 3rd data point - as in somebody else throws their numbers in the pot. Or, of course - best if the guys from the Co. share the stats, that would be ultimate answer.
“EnTerr” wrote:
Was your data filtered per unique installs - or was the same install/device/user be counted more once if they used the channel multiple times in that time period? The latter will be prone to usage patterns affecting the distribution stats (i.e. app likely is more frequently used on boxes where it “runs better”). Mine counted only once each user/player (akin to SELECT DISTINCT or GROUP BY). And btw i am thinking it’s more likely my data be tainted, i just can’t put my finger on the rhyme or reason.
I filtered on “New Users” but I don’t know how accurate that is.
“EnTerr” wrote:
Was your data filtered per unique installs - …
I filtered on “New Users” but I don’t know how accurate that is.
Aha!
That explains some things. So they are unique users indeed but it counts only the new users to the channel. Which hmm, is interesting but not a good representation for the players in use as a whole. There will be higher % of new player owners that become new channel users (only a few of long-time owners will be discovering that channel now - eager-beavers have already found it and the non-curious ones remain non-curious). So in that light TheEndless numbers highlight where the user base is heading (more 4xxx and 5xxx models) vs mine more status quo.
We do request 6.1 of the firmware to install, which could affect these numbers some
Also, in case it helps, video modes stats:
1080p … 48.2%
720p … 43.7%
480i … 8.1%
A bit of additional info, these are all new users as the channel launched less than a week ago, and these counts are individual boxes, not installs (multiple boxes per account) or IPs (multiple boxes per household or multiple IPs per box via proxy/spoofing).
Thank you, very interesting. It looks like a very detailed breakdown of all the non-classic models.
I see “everything” except 4210 and 4230 (newest Roku 2 and 3), which surprises me a bit that they’re not at least some tiny percentage. Is it possible they are somehow getting lumped into the 4200s?
“Komag” wrote:
I see “everything” except 4210 and 4230 (newest Roku 2 and 3), which surprises me a bit that they’re not at least some tiny percentage. Is it possible they are somehow getting lumped into the 4200s?
Might be, as I haven’t received a hit on any of those, when my channel loads I get a tracking packet of FlixRaider details such as load time and version, and roku details like screen, model, firmware, and timezone straight from the various rodevice and roapp functions. Since I am listing this from thousands of installs, and I haven’t seen those model numbers, it just might be that they’re bundled up by the Roku functions.
The only ones I didn’t list were the 4500sk and 4400x since the counts, although they do have values, are so small they listed as 0.0%
EDIT: Am amending my answer here, TheEndless’s post listed a percentage on the 4210x, so maybe they do identify themselves and I simply haven’t seen them yet.
“Komag” wrote:
I see “everything” except 4210 and 4230 (newest Roku 2 and 3), which surprises me a bit that they’re not at least some tiny percentage.
Maybe 4210 and 4230 are reported as “4200” by the method @sjb64 uses. As a reminder, we have the precedent of all RokuTV, where roDeviceInfo.getModel() returns a generic “5000X”. And maybe @TheEndless uses roDeviceInfo.getModelDetails().ModelNumber to see the real model#.
“EnTerr” wrote:
Maybe 4210 and 4230 are reported as “4200” by the method @sjb64 uses. As a reminder, we have the precedent of all RokuTV, where roDeviceInfo.getModel() returns a generic “5000X”. And maybe @TheEndless uses roDeviceInfo.getModelDetails().ModelNumber to see the real model#.
“EnTerr” wrote:
Maybe 4210 and 4230 are reported as “4200” by the method @sjb64 uses. As a reminder, we have the precedent of all RokuTV, where roDeviceInfo.getModel() returns a generic “5000X”. And maybe @TheEndless uses roDeviceInfo.getModelDetails().ModelNumber to see the real model#.
Correct.
To be clear, my model numbers came from roDeviceInfo.GetModel(). If it no longer reports 4210 or 4230, it was changed at some point recently.
Now I’m confused by the result being potentially vague to begin with, the result from GetModel being maybe suddenly vague, the answers using the two calls being different to the same question from the same box, and the fact that everybody seems unsure. A lot of confusion around such a simple function (or two).
