I might have found a quite serious bug in Bambu Lab’s internal analysis tools, which cause innocent people to be banned. Fortunitely it wont happen that often, but it might cause a lot of problems for innocent users. This might sound crazy since I don’t have access too the tools, but I have access to the front end analysis data and the words of the support. Combining these its possible to figure out that either A) There is a bug in the internal analysis tools, or B) The support is lying. Now I don’t think the support is lying, so I assume option A.
I will go through the data in this post to try prove my point. Bear with me, this will be a quite long and possible quite schizo posting, a bit of boring excel math, some transparency regarding my rewards and downloads, some openness about my discussions with the support, I will have to go in to a bit more details about my 2 bans I have received. This is not an appeal for these bans, I have tickets open with the support for that, I just cant explain my suspicion of the bug without going in to the bans, they are a crucial part of why the tools are wrong.
To make this a bit easier for the readers, I will begin with a short introduction with some crucial information about this problem, I will try to keep it short and concise but also need to include as much information as possible to prove my assumptions. Sorry in advance for forcing my life story on you. For the people who don’t care about how this happens in detail, just want to know what’s going on, please read the summary at the bottom.
I have informed the support about this, but due to previous conversation I’m not sure how much they will listen. Its hard to describe it all in a 3k comment to a ticket. Here I can use a bit more text formatting, including images and what not to easily show the issue, It will just be easier to direct the support to this post instead of trying to describe it within the ticket.
So short introduction
There are some dates and timestamps here that are quite important, they are in my local time zone but the time it self is not too important, its the timing difference that is.
- I uploaded a big project, which takes days to complete. 2023-10-13 21:55:00
- It got really popular, 3k downloads within two weeks.
- I got banned by the automatic system due to thinking I had bots download the files. 2023-10-26 08:09:00
- It was manually verified, that I did not use any bots, that no bots had anything to do with my downloads, it was a false alarm from the automatic system. 2023-10-27 12:53:40
- For 2 days my model became even more popular
- I got a second ban because, to quote the support “Upon thorough investigation, we have observed multiple accounts starting to print your models within one minute multiple times”
Due to users receiving rewards when people download their models, it is quite easy for me to prove from the front end analysis that I have in fact not received multiple prints on my model within minutes of publishing. Here is a table of all my models related to the project, the time they were uploaded, the first reward they received, the time I got banned and unbanned and finally when they received their first reward after the ban.
I’m not allowed to post tables so picture from my excel sheet, sorry.
As you can see, I have at no point received “multiple prints” within one minute of publishing my models.
- The fastes model(FOSBOS - Drawer 12u) took 15 hours to receive its first 10 downloads before my ban. In other words Its impossible to have received more than 9 prints within the first minutes. It feels highly unlikely I received 9 prints the first minute, then it took 15 hours to receive the 10th print.
- When republishing my models(FOSBOS - Hotend with Nozzle Bin) after the ban, the fastes model took 3 hours to receive its first reward, the reward of going from 20 to 30 prints. Since this was published before I don’t know exactly how many prints it received when republished, but its a maximum of 10.
So now we have established either 2 things, A) either the analysis tools the support has access to is wrong, B) they are lying. Like I said I do not believe they are lying. So to explain what I think has happened and how the bug works
Explanation of the bug
Due to my project being so large, taking days to finish. A lot of people had time to download the models and start printing before I got banned by the automatic system believing my models were downloaded by the bot. The models got set to private. Which means that the models were being printed while I was banned while the models are technically inaccessible. They were downloaded before, the users started to print, I got banned, the users finished their prints.
This can be proven by checking the table above, I did receive rewards while I were banned and the models where private.
I then appealed the ban with the support, got manually verified, there was no bots downloading my models. 2 days later I was banned again, this was due to the expected bug I’m trying to explain. reason was “Upon thorough investigation, we have observed multiple accounts starting to print your models within one minute multiple times”, but like explained above, this is impossible
So what I assume actually happens
When I was banned, people were still printing my models and I were receiving printing points(Explained above). When I was then unbanned, I republished all my models. The Bambu Lab internal analysis tools have no idea how to keep track of this. To this tool it looks like I received 100’s of prints the exact same minute the models were published, but in fact I received these 100’s of prints during the 28 hours I was previously banned and the models were inaccessible.
Note that this is possible due to the way the reward rules are set up:
The regular model reward applies to each model without any time restrictions. You will earn points at each milestone.
1 successful print will count as 2 downloads.
The regular print profile reward has no time constraints.In order to recieve points, each print profile needs to keep the average rating ≥4 starts, then you will earn points at each milestone.
1 successful print will count as 2 downloads.
The model does not actually have to be accessible for you to receive points on it.
Example
- You create a model
- 10 people download the model the first day, you will receive 10 points on day 1.
- You set your model to private
- The next day the 10 people print their model, you will recieve another 10 points for their successful prints on day 2
- The next day you republish your model, day 3.
On your front end side this looks correct, during day 1 you receive 10 points, during day 2 you receive 10 points, no points on day 3.
However on the backend side, I assume its messed up. On the first day it looks like you received 10 points, but they do not see your successful print points on day 2, instead they will see you receiving 10 points the same minute you republish your model on day 3.
Summary
The reward rules allow a user to receive downloads while the model is inaccessible, hidden from the public, private. When the user then republicize the model, in the internal analysis tools, all these downloads will be counted the same minute the model went public, depending on how many downloads the model received, it will to the support look like the user got 100’s of downloads the first minute of publishing a model. But that is not the case, the user got all these downloads during the days/hours the model was private.
Please have your second line look at this issue before more people get banned. If this assumption is correct innocent users will get hurt.
Thank you for coming to my TED talk