Little big change in scoring system
In recent months, we have been receiving more feedback from you about our scoring system. Increasingly, we are seeing the problem where a single vote can put an asset's score into a different rating level.
We have therefore come up with a simple solution that we hope will help a lot to improve the distribution of ratings. Since Friday 10.9.2021, we are switching the complexity calculation in our score equation from average to median.
Average quality * average complexity * 10
Average quality * median complexity * 10
How does this change help?
Since most users rate reasonably according to our statistics, the median is much closer to the most commonly held opinion than the average. Let's look at this visually:
You can see that among the 7 ratings (color circles), there is one rating far away from the other ratings, and this single rating can shift the final score significantly from the other opinions.
The median ensures that the result is closest to the most represented opinion among users.
In this case, the final score moves lower in the result, but we have also observed the opposite effect, where for very high quality models the score is lowered in a similar way for the original algorithm and raised in the opposite way. For now, however, it is likely that for a lot of models the complexity rating is reduced - the algorithm and recalculation is the same for all creators. Because the redistribution system is relativistic, this reduction does not mean that your assets will start earning less.
We believe this change will help to make the rating of models and other types of assets more fair, where the community will prevail over the opinions of individuals and random mistakes.
Why not do the same for quality?
Given that many users rate either 10 stars if they like something, or conversely often only 1 star if they dislike something, the median value would often end up at these values.
We are working on a better rating system that will be linked to a privilege system. For example, we've found that most of the excessive ratings come from users who have been with us less than 14 days. Therefore, only users who have been with us for a longer period of time and have downloaded a certain number of assets will be able to rate. We are also considering increasing the weight of ratings from creators or paying users.
See next articles
We are glad to introduce the Texture Resolutions feature in Blender 2.92. If you have a small GPU memory or want to create much larger scenes, this is the feature you’ve been waiting for.
We created the first of the BlenderKit artists interviews with Julio Sillet, who has uploaded many great materials to our website.
Literally every area you step in Blender requires you to learn a lot of new tools and information, and that's why we implemented several 'shortcuts' for convenience and speed.
Stones, rocks, pieces of wood can be sometimes photoscanned. But plants are still a domain of hard-core modeling. This really takes time, skill and precision, and a lot of observation what "natural" really means.
We are trying to help you, blenderheads, to save your time while achieving best results possible. Wallpaper textures can really boost interior renders, but do you already have those on your hard drive?
As a late Christmas gift, we are releasing our BlenderKit mascot kitten for free. Now, you can make any scene insta-cute! The kitten is poseable and can be animated (by now 2.79 only, because of a hair-import bug in 2.8 beta).
BlenderKit introduces add-on for Blender 2.8. While there are still many changes in Blender happening, which may cause the add-on will work today and not tomorrow, the add-on should be mostly working fine, and we try to keep the new version updated with the latest Blender builds.
We are happy to announce the release of something we've been working on for a long time.
BlenderKit is a new online platform made for Blender users, creators, and developers. We hope to bring a new momentum into the open source world by introducing several innovations we've been working on.