If I may, I would suggest removing those that were incorrectly recorded. Leaving them in inflates the sitewide download counts and unfairly devalues those creators whose downloads are normal.
I know that removing these points that had been recorded can cause some to complain, but for fairness to all, they should be removed.
Does MW have a reliable way to detect and prevent the same uploader creating two or more accounts and create multiple profiles under different account names to game the system? In other words, what if Designers A and B are actually the same person?
I am surprised by the low 10%. Based on the lengths that some uploaders have gone to game the system for points, I would think the percentage of invalid download counts to be much higher, could be up to 30% of the overall platform data (gut feeling only, no data to back this up ). These invalid downloads could come from 1) donwloard trading, i.e. groups of trading partners, each having multiiple accounts downloading each others’ big number of models to avoid having a few models getting too many downloads that may catch attention; 2) self-downloads, i.e. the same person having multiple accounts downloading their own models; 3) bot downloads, either DIY bots or using third-party bot download services.
Yes, we do have some limits and risk-control checks for individual users’ download counts, but we can’t share the exact threshold — otherwise we’d basically be handing out a playbook for gaming the system haha.
Based on the data we’ve reviewed, most cases above that threshold show clear abnormal patterns, while normal-use downloads make up only a very small share — otherwise this rule probably wouldn’t have passed internal review in the first place.
Believe me, we’d be more than happy to see a model blow up on social media haha. That kind of organic traffic is exactly what we want to see, not something we’d try to block.
The risk-control checks are mainly looking for abnormal behavior patterns, not simply “this model got a lot of downloads.”
We understand your concern about fairness. That said, going back and cleaning up old data has always been a headache for engineers. Download data doesn’t exist in isolation — it may have already affected recommendations, comments, and points distribution. Some of those points may also have already been used or redeemed.
If we simply delete or roll back historical data, it could create new problems and disputes. So at this stage, we’d rather take a careful approach with historical data haha. Instead, we’ll focus more on reducing its impact going forward and preventing abnormal activities from adding new download counts.
As for the detection ideas you mentioned, we’ve got the suggestions and are giving our risk-control team a friendly push to review them carefully and keep improving our detection. Thanks a lot for the feedback!
That’s funny. When I tried to bring this up 6 weeks ago and figure out what was going on, you accused me of trying to game the system. Apology accepted.
The fact that these unearned print and/or download counts give these models permanent long-term advantages, such as better chance in recommendations and better points distributions, is even more important reason to remove them. However, if engineering resources are limited, a forward-looking priority can be understandable.
Thank you for the continued commitment and efforts to make MW a fairer and better platform for creators.
I think you got it all wrong DWdesigns. Look at the pictures posted again. Print profile saying some models were downloaded for example 1000 times, where model says only 500. So what advantages those models get? Not more visibility on the page or chance in recommendations. They got less! Now ask yourself, is the traffic on any page so bad, that 50% of it is malicious? So who is responsible for it? Not me and not the people who raised this topic. And yet, 57% of downloads of one of my models are counted as malicious. I want to say it again. Print profile shows 57% more downloads that model was in fact awarded. So who is making it, again. Some bots scanning whole Makerworld and downloading everything from some category? Sounds weird to me. Now one more thing. There are tools like likes and collect, both can say you a lot about the model and its popularity. And definitely both are unique. You may only like or unlike the model, ad or remove it from your collection. Now compare like/collection ratio to actual downloads on Makerworld to other platforms. I’m not even saying 3D model platforms. What I see here is the most hamstering community in the world, throwing the most amount of likes to things they will never download. And it’s even more! Compare amount likes/collection to download on Makerworld now vs. same to Makerworld 1.5-2 years ago. Just check your older models. Drastic shift in community behaviour or something else?
Yes, that’s exactly how it works. Downloads, likes, collections, boosts, followers, everything is separate. My point was, their download counter between profile and model is dead accurate where global is not.
I won’t pretend to understand this in any detail whatsoever. But…
It is not uncommon for me to download the same model multiple times over the course of printing the project. One case recently was my print of the dual AMS Pro 2 rack. I wound up printing this project over the course of 2 weeks or so, so I know I wound up downloading the project numerous times. How does that fit into this? If that particular model is getting “credit” for all my downloads, that is neither fair nor accurate. Could this be, at least in part, the cause of the discrepencies?
Hi to all. I was hesitant to participate in this discussion but I would like to present my download / print history for the last year. You can see how print number started exceeding the download number, which is an interesting phenomenon.
I’m posting this as another input to the conversation and possible analysis. One of the possible explanations I came up with is that one of the good performing models is printed mostly from mobile (which counts only as a print). However, also on other models I see lower number of downloads compared to before.
Another user pointed this out - was this answered?
Actually, that shouldn’t be possible, since prints only count when they go through the cloud - for example, when using the app or printing directly via MakerWorld. This points to a different explanation: the downloads weren’t counted because they were flagged as bot activity.
So this sentence needs to be reworded:
It’s not that the print count exceeded the download count; rather, your downloads simply weren’t registered / counted. They were retroactively invalidated.
You are clearly affected by the problem described here: Sometimes more, sometimes less. In the first example, 100 downloads weren’t counted; in the other, there are only a few.
In a nutshell: That’s exactly what was written in the original post, you are affected. Those downloads weren’t counted, so you’re missing those downloads and points.
Then you would expect to see a disproportionate increase in the gap between downloads and prints at one point, followed by an equal closing of the gap later.
Although I really struggle to think of a scenario that would lead to what you’re describing. Normal user behavior is pretty…well…normal. What you’re describing would require two very atypical behaviors, which each have different causes.
That can be ruled out. Because it doesn’t explain the different numbers. That’s exactly what was written in the original post and what Bambu also confirmed; see the attached solution.
I marked it specifically with red arrows. 495 / 393 - 102 downloads weren’t counted because Bambu flags them as alleged bot activity. The numbers should be the same at all times.
And that’s just one 3D object - it runs through Slava_Jim’s entire portfolio.
Hmm… This is really puzzling. So this “mysterious” bots infesting Makerworld following trends! It’s not that they bumping numbers for users who try to work around the system. Download spikes exists the same time as print spikes. Mind Makerworld not called them bots but “malicious download” and estimated numbers at around 10%. Well, Ogi3D’s (whos been removed from this topic) most popular model - Druid helmet, has this malicious download at 54% (and I’m just counting one print profile!), even MaKim’s models are affected big time. One of his newer models - Exploding tower, has this numbers at 47%! Do not show stats for old models RetroSharky, check the models from last 3 months so the numbers will be a lot bigger. One of my newest model from September shows 7 print profile downloads and just ONE for the model. This is madness. Bots? And what bots will do with full size cosplay helmet, crazy complicated and time consuming to build dice tower or a stand for studio light? That’s not flexi dragons or fidget spinners. Bots are limited by captcha and then download limit, how were close to 50% on newer models, where Bambu saying 10%? I got my theory. If bots are roughly at 10%, the rest are legit people making legit downloads, having Bambu account but not registered Bambu printers! QIDI, Flashforge (and others who want access to biggest 3D models repository) and especially Snapmaker users. Bambu decided to not kick them from the site (as potential customers), but also to not pay for that kind of downloads so we are punished for that. That’s what it is in my opinion.
I’ll post stats from people who want to participate voluntarily; I don’t want to drag anyone into this without their consent.
The people whose numbers were posted here have given their approval.
Of course, everyone is free to post their own current figures here on a voluntary basis.
I wrote the word “bots” for simplicity’s sake, because I didn’t want to use the word from the Bambu Statement every time: “Malicious download data”. You’d have to ask Bambu what all that involves; her statement already says quite a bit.
Otherwise, of course, thank you very much for your comment.
I also noticed that some of my more recent models have a disparity between profile and model downloads/prints, with many not being counted, or prints being higher than downloads, which doesn’t make any sense at all. I sent a support message and got the same copy pasta as you.
I don’t really get how prints can happen from thin air
). These invalid downloads could come from 1) donwloard trading, i.e. groups of trading partners, each having multiiple accounts downloading each others’ big number of models to avoid having a few models getting too many downloads that may catch attention; 2) self-downloads, i.e. the same person having multiple accounts downloading their own models; 3) bot downloads, either DIY bots or using third-party bot download services.
