Rating System


#21

Hi antech, thanks for the “test” ahahah.

However, the formula you applied in my case is wrong, as it should probably be:
(1x10+4x8+2x5)/(1+4+2) = 7.428 ~= 7.4

It’s an arithmetic mean so it makes sense: you divide for the total number of votes obtained.

One could argue the introduction of the 2x5 initial votes that are initially applied to all of us is a bit misleading:
let’s for instance suppose I was having all 5 stars ratings (previously) that they got translated to “8”. An arithmetic mean where all the votes are “8” is exactly “8”, but this is no more true, of course, if you “stealthy” insert two votes that value “5” . I mean: in the comments section you cannot see these 2 votes rated at “5” so this probably misleads lots of people (me too actually, before this thread)

Of course it’s not a big issue, the computation is done correctly, at least in my case.

Actually, I cannot say the same for the example you posted on the 27 May (the wallpapers one).
They all have 2 votes rated at “8”. So by applying the formula we should have: (2x8 + 2x5)/(2+2) = 26/4 = 6.5 but they are all rated “7.5”.
So, I don’t know, maybe I’m missing something.

Finally, maybe I jumped on the wrong thread since you are all talking about ratings so maybe I should have started a new thread to keep things clean.
In my case, the issue is that I don’t know actually what’s the formula/algorithm used server-side to sort the products basing on the number of downloads.
I don’t know actually if it’s based on a “all time” downloads approach or on a “download/day” ratio. To me, since I have only the info that the products should be sorted by “most downloaded”, it’s seems really buggy these days. If someone knows something more please let me know.

Now, I re-checked to see if the issue continued to be there and my theme is back on the 7th position of the Most-Downloaded list, much more fair if you asked me. Again, today’s strange behavior is another symptom and indicates there’s an anomaly that triggers itself randomly on the server-side since it’s not the first time this happens.

Hoping opendesktop will help us to understand what’s happening.

Greetings to everyone!

EDIT: I was trying to see the difference between my products and the wallpapers with a “7.5” rating.
In their case they have a 10 votes rated @ 7.40

In my case, there are exactly 2 votes rated @ “5”, as you can clearly see from the screenshots:

So now the computation is correct: (2x8 + 10x7.4)/(12) = 7.5. One problem is being solved. Now, why these wallpaper are having these 10 votes rated @ 7.4 instead of the 2 votes rated @ “5” we are all having?


#22

Hi mrcuve0,

Count 10 Rating 7.4 come from old anonymous voting system. and with the same principle . (8x8 + 2x5)/(8+2) = 7.4.

please open a new thread for the other issue.

Regards,
Opendesktop Team


#23

Thank you maggiedong for your help, everything should be clear now. I’ll do as you suggested.


#24

After another period of testing, we adjusted the base of the algorithm that calculates the Rating for each product. This is increasing the importance to have a larger quantity of votes compared to a few very high rated votes. We think this is more in line with what users suggested as feedback in the comments.


#25

Great!!! When will this change start working?


#26

Hi Dani.
I think the fixes already hit production. As far as I can see, they doubled the number of default “5”-valued votes.

Before the changes we all had 2 votes rated “5”. Now we should all have 4 votes rated “5”.
I don’t know if there are other fixes “under the hood”, these may be the most evident ones.