Topic: Ratings Aggregation

Quick question on the rating system out of curiosity - have read the guide on the site about weightings, median values etc - just want to check I'm interpreting it right?

Just to take a most recent example, the 'Yessingles' compilation was added by RushFan4 a few months ago who rated it as a 6. Today I've gone to it and added links to both Spotify and Progarchives, which has led to a PABot rating of 6.3. I have also now rated it as a 6.8.

Throughout all that, the overall aggregated rating has remained at RushFan4's original 6. Presumably the weighting given to RushFan4 is higher than the combined total of mine + the PABot, so that regardless of what I rate it as, the median value is always the rating given by RushFan4 (e.g. R-R-R-R-R-R-PA-ME-ME-ME-ME)

...until someone else also rates the album as higher?

Yes, that's basically what happens. First the system finds the tier group (low/mid/high) with the most votes (including the weighting by trust rank). In this case it's the mid range. Then it finds the predominant tier (60-69), and then within that it finds the predominant sub tier, which is 60-62 based on Rushfan4s rating.

When more people rate the release, whether their ratings affect the combined ratings depends on their ratings. When another user rates it as 6.8, this will become the result because combined with your rating it will overrule Rushfan4s rating. If it's in another tier, nothing will happen. If three users come along and rate it 7.0, this will be the new combined rating.

BTW: The paBot vote can carry a lot of weight, depending on how many users rated the release at PA. If it is only one user, paBot ratings have a trust rank of 1. But the trust rank increases with the number of PA ratings up to a value of 8 (for 50 ratings or more). A similar rule exists for the rymBot, which can go up to a trust rank of 10 for 500 ratings and more.

Mike, I will reiterate how much I disagree with giving the bots influence. It's pretty much well known that RYMers (being majority/plurality teenager) have pretty controversial, if not downright bad "taste," and I really want to see this site distanced from it. It's a different fanbase that uses these sites (although less so for the progarchives, albeit) and an aggregation from these sites really should not influence the ratings on this site. Same goes for genres - the PA and RYM are rather limited in their approach to genres, and thus a bot that draws from these sites cannot use Progfreak as efficiently as a person.

Much appreciated Mike!

I have about 1500 albums I've rated on Prog Archives and thought as some point I'd "transfer" them here so good to see it would make a difference in some cases!

Would also mean using the Discovery Mode in the chart page would much better for me beyond just the new releases!

Re Bardberic's comment - personally think its definitely necessary to use the bots in most cases as its a much smaller sample size on this site. There would be a huge amount of albums that are totally unreviewed or only by one or two people otherwise surely?

So at least with the bot ratings there's an early indicator of the broad opinion on an album that you can go by. And once a handful of people have rated it here themselves the influence of those bots isn't that much anyway as an album rated by 1000 people on RYM is still only weighted as much as two or three people using this site?

^ Yes, the bot ratings exist to put releases "on the map". I agree with bardberic about the biases at RYM, which is why I've really limited the influence (weight) there. Maybe 8/10 is still too high. I could reduce them to 4. Then a release with both pa and rym ratings would have at most 8 trust rank points from the bots.