Yup, it seems like BL wants to force people away from simple, robust, functional designs and towards more complex models. I get that they look cool and they want their brand to be associated with high-end, intricate designs, but the numbers speak for themselves.
2 Likes
well it turns out that to please the gods of the algorithm you have to stop Simplify Model and throw in super retail models with 1gb capacity. For me it’s not a problem XD I wonder if the server will accept it.
I do not believe ‘complex’ relates to polygon count. This would only elevate AI generated models as they typically have a high polygon count.
I suspect complex means the following:
- over a minimum amount of filament usage
- maker supply parts
- detailed description
- assembly guide
- unique from other models on the site*
- low chance of copyright strikes*
* unsure how these would be determined automatically
1 Like
So far, I don’t see any difference among the models I own. The milestones and rewards seems to be all still there, just redeemed 24~48h later.
Feels like a preparatory change.
You might have mixed originality in with complexity.
One way to assess the complexity of a model is to use statistical modeling (regression analysis) to find a set of predictors. Model comlexity is unlikely assessed as binary, ie, either complex or not complex, but rather various degrees of complexity.
Originality assessment will likely involve some AI. Images or even the model file itself may be compared to what are out there to assess the degree of originality.
Attempting to artificially increase model complexity will most likely backfire. Doing so will be like some overloading their articles with keywords and end up with an article that is hard to read and also detected and punished by Google algorithm.
The best way is to ignore the algorithm. If the idea calls for a simple model, then make it so and try to make it attractive to users to get many downloads/prints. If an idea naturally leads to a more complex model then do so.
The difference the new upgrad makes is that, if it works, designers don’t have to discard ideas for complex models, knowing that if done well they will get the same reward (even with less prints and downloads) as those simpler and more popular models.
1 Like
Mi dispiace x te ma l’ho creato con AI tripo3d.ai
Poi lavorato con nomad sculpt
To put it neutrally: proof or measurement would still be needed to confirm that this is really the case. Are more points really being awarded overall? Or are fewer points being handed out?
According to the theory you describe, downloads and prints would have different values. It would be interesting to know how this is calculated. Would one print be worth 1 point and another 5 points?
I can’t imagine that users would be happy if certain models suddenly became five times more valuable, respectively hidden multiplier. If that becomes public knowledge, then we can be sure that spammers and creators alike will rush to come up with ideas and implementations that are more valuable, having a hidden multiplier thanks to the algorithm.
2 Likes
On the two extremes of the models here on MW, I can see one print of some well-designed original models is worth 20-50 times that of one print of some low-effort mass-produced models.
MW is the one who decides what value they put on the models though. It’s their money after all.
1 Like
On the Reputation element, I hope it’s not based on ambiguous criteria and lead to another layer of playing favorites.
I would suggest the following be added to the reputation score. These should be counted and be a permanent part of the recoreds for accounts. The reputation score will affect all models under the account, not just the offending model(s).
- Any attempts to artificially increasing complexity of the model
- Engaging in fraudulent activities, such as trading of print, download, and boost, fake engagement, using multiple accounts to print and download their own models, etc.
- Violations of rules, such as lack of real photo in print profile, using AI image as cover, etc.
Reputation score can also be set up like that of a driver’s driving safety score. All start with the highest value, then deducted from that value after each offending incident, much like car accidents resulting in point deductions.
Based on the information given in the post, it seems like originality and hard work will be rewarded more than low effort Ctrl + C Ctrl + V content that has dominated the treding page on maker world for past few months now.
I am just hoping this will not be the same kind of algorithm that rewards trending over content.
A few weeks passed and I am wondering how all of you see the the ‘upgrade’.
Has it affected you at all? Was it positive or negative?
Or is it too early to say in your opinion?
Inna separate note
It seems that there are tons of new junk uploads every day and some of those seem to be doing quite well. It does not seem we are getting significantly more complex/ interesting models.
One can use the number and intensity of complaints as the indicator. So far none, so the upgrade is not working yet.
1 Like
I’m down 300 points and like 70 boosts in the last 30 days. I will say all of my numbers are down so it’s likely just that. The boost thing concerns me though.
Hey all here’s some data I’ve been taking by Hand recently, screenshots are below from a google sheet I’ve been using.
Some notes:
- Boosts come through straight away
- profile and milestone rewards come through with a 1-2 day delay
- Some models have their profile rewards come through after 1 day, and their milestone rewards come through after 2 days.
- On two occasions, I had milestone rewards come through 3-4 days later
- if a single model receives multiple of the same profile or milestone reward in the same period, they are grouped together (instead of multiple 3.75 rewards, they come as a 7 or 11.25)
- One model that was usually receiving milestone rewards of 18.75, received 21, I hadn’t seen any other cases where an 18.75 and 3.75 were grouped together like this.
- in general it seems like my boosts have increased and my milestone/profile rewards have decreased
- I have taken a mini hiatus over the last month and a bit due to personal reasons, and I’m sure this has affected me algorithmically. The fact that I’m staying steady despite this could imply a reward increase while my activity has decreased. It’s not concrete though
Screenshots below, with dollar values in AUD
1 Like
I didn’t spot any difference. Still getting the milestone reward I use to get before, but delayed by 24~48 hours.
The only reward I’m not sure about, is the 10 points for profiles I get time to time on a single model, but maybe it’s just the regular milestone for the >1000 download of a profile? which I neved had before.
It does not seem like an upgrade to me. It is impossible to check what gets me points, how many and for what reason.
I do have a drop in points (i get ca 75% now)
As for the second part, yes i also see a lot of junk uploads, badly executed models, a lot of them created with ai and poorly painted in bambu studio, sometimes even the avatar is AI generated. Many of them have no regard for good print settings (yes i sometimes print 4-6 versions to get it right)
I also took my models off the china site because some of them were mass sold on temu. I did not get many points from the china site anyway. The question is: why do they have a china site when they can translate the global site into all kinds of different languages? Are the terms and conditions different there?
I know Bambu does not owe me anything, but this goes both ways.
1 Like
While you don’t have that info in “real time” as previously, plus it’s additionally hidden behind sums and whatever adjustment ratio’s there, there’s analytics on the platform: https://makerworld.com/en/my/data-overview/model
There’s a bug there with analytics of print profile (https://makerworld.com/en/my/data-overview/printProfile) claiming there’s no points since 17th june (@MakerWorld ):
Yes. China’s site is hosted on different servers, behind “great firewall”, there are “protections” in place and (additional) censorship making it necessary to host dedicated china-only site.
I am aware of the statistics in the creator center, however it is impossible for me to check the point rewards from 2 days ago, since it is even less transparent than it was i just have to trust all is okay i guess. Looking at the statistics i see a rough 25% decline in points from the day the new system was put in place and cannot pinpoint how they calculate this. And no it is not because i pulled my models from the china site because i did this this weekend and not before.
So my questions remain: what gets me points, how many and for what reason?
@jaayjeee That is some serious approach. I have not been that thorough/precise, only compared week to week in Creator center and my points are down a bit. Luckily its not -80% as some people complain elsewhere. I kind of stopped with new uploads so I don’t really benefit from early milestones (I think I did one since new rules came in and it only has 3 downloads)
What I gather and
@user_4059763918 I did not make up my mind about it yet but I also don’t like the secret rewards formula. Looks like it might not have changed much for now and it is still early to give definitive answer.
I don’t know how to feel about the China site. It does not seem my models are NOT very popular there but I do get occasional boost or few points from China so i’m happy.
More and more people (and I’m kind of leaning toward this myself) are thinking that a secret to the success on MW is to design so model relies on Bambu technology and supplies/materials.
Bambu filaments, Makers supply parts, hardware, electronics, recently Cyber brick components. Seems like they are taking the lessons from the HP book. Selling relatively cheap plug and play printers with really low entry point in the form of A1mini and nudging people to their “consumables” Someone who I gifted A1mini already spend almost £1000 on filaments and kits from makers supply. They got the lamp kit with the a1 mini and they got hooked really quickly.
My most popular models would definitely fall into the “complex” category but it doesn’t seem like they’ve benefitted from the changes. Here’s the last five weeks of user activity with June 18th representing the first day the new point system started working. The recent spike in activity is due to a new model but prior to that my last upload was in March so the numbers had been fairly stable.
