How to do some stuff with scraping backpack.tf or using API

[Smoker]

Hey all,

 

I'm doing a project where I try to analyze the probability of unboxing an unusual item from a crate.

 

I already sort of built a simple unbox simulator based on my hypothesis on what the underlying unusual probability algorithm is.

 

My next step is to evaluate how accurate my unbox simulator is by comparing the number of unusual items unboxed based on what my simulator says vs. what the actual numbers were.

 

For example, say for the month of November, 2020, there were total of 100,000 unique crates unboxed, and total of 650 new unusual hats have been unboxed. Then, I input 100,000 in my simulator and simulator shows 700 unusuals were unboxed, so for the month of November, 2020, the error was 50 (700-650). Something like this, and I was hoping to get this data for at least 12 months of data.

 

I looked everywhere on backpack.tf API's documentations, but no luck.

 

So if there is no API, my only option is to write scraping script for each unique crate and all possible unusual variants from each crate and see how many were in circulations, basically some complicated nightmare I don't want to deal with.

 

Any suggestion?

[Mememememem]

cиm

[Smoker]

hmm...

[Worfy]

I'm not exactly sure if it's possible to get the kind of statistics you were looking for but if this helps at all, in CS:GO the odds to get a knife/gloves is about 1/400. Most of the time when I see unboxings in TF2 people get unusuals a lot more often than people get knives in CS, so unusual odds are very likely higher. A lot of people have suggested 1% but that feels quite high to me.

 

Also during the crate depression according to the TF2 team roughly the same amount of unusuals were made tradeable as there are new unusuals unboxed in a month so if you can figure out how many unus that were unboxed during the crate depression were made tradeable that could help you a bit.

[Smoker]

@Worfy - thanks for a response. Yeah I think most people think the chance of an unusual is 1%. I've done some research on this topic, and there are in fact multiple tf2 crate simulators out there + some discussion threads on this topic from several places, and there people seem to think it's around 0.65%, and I've used their simulators and ran them many times, and as the number of unboxes grow, the probability always seem to converge around 0.65%, so I'm pretty sure they just implemented a random binomial distribution with p=0.0065.

 

And yeah I also saw a bunch of vids about the crate depression, and one of them got to the math you were talking about, let me try to search that revisit that logic. What's even more perplexing is that during the crate depression, only certain types of older crates were affected (guaranteed to return unusuals at 100% chance). This sort of suggests valve's item server maintains separate algorithms for different types of crates, and I'm gonna try to dig into this more.