User talk:Oap

From Team Fortress Wiki
Jump to: navigation, search

Case unbox stats

Following up on our conversation on IRC, I have added graded item totals to my inventory history script to get a sample of 6578 case unboxes (the item discrepancy between graded items and cases opened is from the unlocked war paint cases from contracts which I didn't count since they had civ grade items which would throw off the count). Additionally I will be removing 200 items from the tough break case since it doesn't have elites

43/6378 = 0.674% 204/6378 = 3.198% 1036/6378 = 16.243% 5095/6378 = 79.884%

With the largest proportion being estimated at 0.8, a 95% confidence interval for being within 1% requires 6147 unboxes which I am comfortably above, thus I am 95% confident that these are the correct values within +/-1%. After I fully finish and release the script and get other people's unboxes we should be able to push this higher.

As far as I am aware this is the most unbiased data we have for cases. As I mentioned on IRC, backpack is not fully unbiased due to trade ups, private inventories, and the current API problems that prevent most public inventories from staying up to date. For anyone reading this that did not see our IRC conversation, using backpack stats for cases is inherently flawed for the mentioned reasons and can be proven by looking at population data of old cases such as the quarantined case (note these values are the exported values from this aggregations link). Something I failed to realize in the original IRC chat was I didn't filter to found in crate items only. Here is the image of those stats; in either situation the shooting star, a commando grade item, has a larger population than two mercenaries. We can see the same thing in the confidential case too, with the commando grade capper outnumbering the merc grade head full of hot air. Jh34ghu43gu (talk) 08:44, 4 September 2023 (UTC)


Thanks for the detailed data. I agree there's no reason to believe the unbox rates for the same grade would be different. Therefore, we can probably round then divide by the number of items in the grade so it adds up to 99%, with 1% being allocated for unusuals. I don't believe we should have all items in the case add up to 100% because unusual unboxes are fundamentally different than case items. As for the unusual rate, that's for a different discussion. Also - backpack.tf data is going to be somewhat more accurate the higher in tier you go, because of fewer trade ups.
It looks like with your data, there's a 5x multiplier based on tiers. We can probably use a general 5x rule. I have a script that compares the backpack.tf item amounts as a rate of the total drop percent. I'll cross reference backpack.tf stats with the 5x multiplier, particularly at higher grades, and see if it generally lines up. There will be outliers, and backpack.tf isn't totally accurate, but there also needs to be a way to verify your data with the general stats of the item. If there's a statistically significant 5x rate increase across item tiers, we can use a 5x rule. Let me know what you think. Oap (talk) 15:49, 4 September 2023 (UTC)


The Unusual page's estimate is 0.66%, I would use that.
BLU Wiki Cap.png | s | GrampaSwood (talk) 16:01, 4 September 2023 (UTC)


I forgot to look into the rolling 20% odds I mentioned, it did come from csgo where they were forced to release the odds to china; as stated, each higher tier appears in a ratio of 1:5 of the lower tier so 5x like you said or rolling 20%. I don't think we can use the covert odds = unusual odds because its been somewhat known that csgo rare item odds were lower than tf2's. As swood said, 0.66% appears to be the unusual rate from both my data, backpack's data when it was good (pre-2013 when all inventories were public), and that one finance article from the early 2010's.
The fact unusual items are still tiered means that we don't even need to care about removing that 0.66% from the other values since it probably keeps the same odds that your unusual will be X grade. An exception to that is going to be the invasion cases since not every tier can have an unusual. Probably should put a footnote about that or something. This is a different story for crates since unusuals are a separate item (except for winter 2014 nice crates). Jh34ghu43gu (talk) 18:51, 4 September 2023 (UTC)

Modules on this wiki

Hey Oap, While I'm generally in fond of Lua in mediawiki (and it certainly makes my life easier), this wiki has a drought of users who actually understand templates, much less Lua. We've got a handful of mods (other than me) who have an "intermediate" understanding of templates, but next-to-nobody who can handle Lua. If there's something truly complex you need for a template, feel free to reach into Lua, but otherwise, I'd greatly appreciate you sticking with pure template syntax here.

Thanks, Darkid « TalkContribs » 23:22, 4 September 2023 (UTC)