"RokuJoel" wrote:
You have triggered a resort. We are switching it back to Algorithmic.
Well... <scratches head> that doesn't seem like an improvement at all.
It would be like saying using a triangular wheel in a bicycle is an improvement over the square wheel because it eliminates one bump per rotation.
[spoiler= Johnny Hart, BC:1ue034tu]
[/spoiler:1ue034tu]First thing to notice is that the sort order is unstable - i.e. re-loading the category changes the exact order. Compare these three shots (Roku3):
It's not that the order has naturally changed because s/o downloaded one of the channels - as timestamp shows, those are taken within couple of minutes from each other. And it does not seem like out-of-sync back-end servers returning different things, since there are more than two orderings.
And second - ad-infected games aside, there is no rhyme nor reason to the way games are ordered. Yes, i am scrutinizing the list because my apps (free and paid, new and "old") somehow reliably end up at the end both - but the order should make some sense, right?!
- We already know it is not by average rating - but to verify, cue "PartyTrivia" from the middle of the list has a 2.5-star rating (about the worst one can see on Roku), while "Hangman" at the end pulls a solid 4
- Maybe it's by total number of installs? No, because "Backgammon" with <1,000 installs is ahead of "Snake" and "Retaliate" which have >100,000 installs each
- Maybe it's by velocity, aka "how many are being installed lately". No, because "Agario" gets ~10-20x more installs per week than "Backgammon"
- Maybe paid vs free are unjustly weighed? No, because paid "Anna Montana" is ahead of the free "Anna Montana Demo" but the paid "Mazes Plus" is behind the free "Mazes"
To be sure, RokuCo does not have to disclose the exact ordering algorithm - it may be considered trade secret to avoid publishers strategically "gaming" it. Apple, Google, Amazon don't publsh their algorithms - but in addition theirs actually work - they don't produce bizarre ordering of the apps. And Roku's sample sizes are not that small to cause instability of statistics - something else is amiss?
If we pull back a little to the big picture - what is the purpose of maintaining "top charts" in general? The benefit is 3-fold: (a) for end-users, it lets them find the best content in some 3000-odd channels in the case of Roku; thus (b) for the platform owner (RokuCo), it improves engagement and goodwill to the platform, plus profit-sharing; and (c) for the app developer, when ordering is properly done, it aligns their interests with those of the platform owner, e.g. if charts are ordered by user ratings, that incentivize me to strive writing high-quality beloved app; if it were ordered by top grossing, that's extra incentive to squeeze more $$; if it were by number of installs, that's incentive to run viral marketing and so on. But as the ordering in Games is now, from all i see it's "none of the above" and fails to benefit viewers, Co and developers
