Vault Hunting 101: A Guide to RNG and LOOT!

D’Oh! I knew I was tired when I wrote that; it’s 40% or 2 in 5. Fixed, thanks!

Thanks for the link; I hadn’t realised that was still available on the oldforums archive, and the other version I have is missing most of the images. Unfortunately, the link appears to be missing the table of intervals? It is for me, anyway. I need some time to go through a number of things before filling in more content, but it’s New Year’s Eve… I’m not trying to present a “stopping rule” as such, so I’ve made some edits to try and clarify that. It’s more the point at which I start to feel comfortable sharing my data with others so it can be pooled.

Anyway, time to go!

Figuring the pooch would get screwed in the archiving of the old fora, I preemptively made complete page backups of those posts.

I don’t know how portable they are (internally the links are mapped to a local folder on disk), but I could poop them to PDF (you see everything, but no working animations) and a separate folder of the images (so animations work and you can use images directly. N.B.:images are usually transparent background with colors mapped to work with the old fora dark background, e.g., some formula images will just look like a white sheet naively viewed - they need to be on a dark BG (new fora here should be fine) or put over an appropriate BG in Photoshop/etc. ).

If you think you’d find that useful, let me know, I’ll scrape it together and put it somewhere you can get at it.

1 Like

Hmmm… How about zipping the folder? As long as the links are relative within the folder, everything should be good. Besides, I can work with broken links as long as the content is physically there.

Snagged it - thanks!

I think I have one more part to add to the OP; anything after that will be an addendum posted in the thread and linked from the top post.

Well, let’s give that a shot.

Here’s a zip with the 4 main guides I did on the subjects & mathematics:

  • Mathemorphous the Invincible
  • Random is Random (mostly)
  • A guide to sampling for drop rate inference from experiments
  • A Guide to RNG and its use in games

A quick local test seems to indicate this works - images/animations are intact, only web links are missing (e.g., cross-link to some other posts I did), but you can assume such links simply point to one of the above posts.

Let me know if it works/does not work, and reminder that most images are transparent and color mapped for dark background, so if you open a raw image and see what appears to be blank canvas, try putting it over a dark layer.

Feel free to use as you see fit, no attribution needed - I did these as a public service…

Here’s a graph of grinder streak probabilities - perhaps useful to you.
image

1 Like

This is an incredible post.
In regards to the Can you game the game? section I believe a special mention needs to be made for the multi-drop bosses, especially the infamous BNK-3R and Saturn. While a now dead thread of mine was intended to explain the Legendary Hotfix, now Update, it was also about explaining that 1:10 in more detail and we know that every vanilla boss as a single 1:10 chance (with the exception of Terramorphous who has two and Vermivorous who’s guaranteed to drop something) for anything in their pool which still leads to disappointing expected returns (“The Hive just won’t drop”/“Why can’t I get a Sham?”).
So I just would like to suggest adding a sweet graph for either Saturn or BNK-3R with the Observed drop rate and keeping Skins/Heads in mind. Terramorphous and Hyperius (no Seraph drops though) are also good candidates for that, the former especially because of two chances per kill.
(and maybe also the Creepers, they are suspiciously generous. Yes, I am aware it might very well a case of me being very lucky)

I also still believe UVHM has a better general drop rate but proofing that without Word of God is much more difficult thanks to low chances (and locked rarity drops). The only thing I have are 500 normal vs 500 UVHM2 Hyperius kills with 1 vs 3 random Legendaries respectively but considering the nature of random we’d need more than that.

(Please don’t say ask why I am posting again. This was just interesting enough).

SWEET!

I noticed there was code in the comments to perform the calculations required, but I hadn’t had a chance to try using it anywhere yet. Thanks![quote=“HeyCarNut, post:16, topic:1153869”]
Well, let’s give that a shot.

Here’s a zip with the 4 main guides I did on the subjects & mathematics:
[/quote]

Lookin’ good, Houston.

I for one am not complaining!

I have sketched some stuff on the vanilla game raid bosses, because Paul added detail in his other dev blog post about them, and particularly terramorphous. It might be a while though - things are about to get crazy again IRL, so time will be a bit more limited.

Good read. Do you know of any data collected on Hyperius drops? More specifically on the 360? I have been grinding for a Norfleet for a very long time, in excess of 250 runs. I’ve finally gotten fruatrated enough to document my runs in attemt to prove that either Gearbox accidentally nerfed it on 360, or my game is bugged.

So far I am 37 runs in.

I don’t have time to look right now, but I’ll search through the drop rates survey archive data for you later today.

But do we have data on them post-buff, aside from mine? My data, after some talks with a certain someone, suggests that the hotfix/patch actually generalized loot pools, possibly an unintentional side effect, as I got far more Norfleets in 700 Normal mode kills (14 to be precise) than the old rate would suggest, even when tripled.

I would gather some Xbox 360 data myself but I am having trouble redownloading the Season Pass content >.<

Ouch, sorry to hear that.

I am gathering data for 360 in UVHM OP8 right now.

Before gathering data I uninstalled and reinstalled the game, to hopefully eliminate any glitchiness associated with my loaded game. I am performing my tests with a mule present, without any Vault Hunter relics.

So far I have 80 runs (documented, I am discounting all the previous runs as they are undocumented), no Norfleet yet.

No, but we should be able to compare rates of Norfleet versus anything else. The drop rate should not (assuming the buffs worked the way I think they were intended) be worse than it was.

@kgk4569’s overall numbers, while still a very small sample, are consistent with a legendary rate of 1 in 10, but that one obviously can then be one of a large number of items. Did you track total legendaries over those 700 runs, or just Norfleets?

Yeah, a lot of folks seem to have been bitten by that bug. Did you try re-downloading through your purchase history?

I have 110 runs documented now. Norfleet dropped on 105. Sometime this week I’ll put my runs in spreadsheet form.

Sledge’s shotgun was the overwhelming winner of drop %.

Looking through the original survey data (pre-buff), this is what I see:

  • Hyperius:
    • 22 legendaries in 830 runs = 2.65% observed drop rate (nature of legendary not specified)

There doesn’t appear to be an entry for Vermivorous, unless I’m missing something?

I tracked absolutely everything (important) as detailed as I could. Overall I got 73 Legendaries, 75 if we include the Pyrophobia. I make this differentiation because it seems to drop FAR less than everything else, suggesting it’s not in that pool.
Overall we have 10 Kisses of Death, 14 Norfleets as mentioned, 17 Hornets, 13 Sledge’s, 11 WTFs and 8 Shredifiers. Thus, 73 Legendaries (since the KoD clearly counts. Unfortunately) or 10.43%, pretty close to the expected drop rate.

Like I mentioned earlier, the Pyrophobia simply doesn’t seem to work into this as is rather clear by these numbers. Therefore, my assumption is that Hyperius has 6 drops in that pool, leading to a realistic drop rate of 1.6667% of any of these drops, most specifically the Norfleet.
I tracked the occasional Seraph drops as well but they are too rare in Normal and aren’t what we are aiming for anyway.

As a matter of fact, I haven’t. Although it seems like that may no longer be necessary.

1 Like

The original survey data did NOT track uniques though, so to do a direct comparison you’d need to take those out. Which gives us… 9%, or roughtly 3x the old data rate. Close enough in my book for >1K trials, anyway.

Yeah, they could just be world drops from one of the however many world loot pool rolls. It does make me wonder about the Norfleet though - does it just take a larger portion of the possible outcomes for any legendary, or is there one roll that’s specifically for the Norfleet? My hunch would be the former, since that sits better with “increased chance” than the latter, but the latter approach would be more consistent with the way the mini-bosses (Lee, Boll, etc.) work.

Not anymore. I got (without looking at my numbers) 3 or 4 Pyrophobias, that is more than WTFs, Hornets, and the Norfleet combined.

My data so far has Pyrophobia and Norfleet inverted from what you saw.

True, things were different back then.

My guess is that the Norfleet in this case is “just another” drop from his pool. No bias or anything.

Well, I am confused now. This seems to extreme of a change for simple cases of bad/good luck.