Crawmerax statistics - a study of >120,000 drops

Hi everyone,
I took on this study to try to figure out a few things about what Craw drops, how often, and how we might maximize our chances of getting what we want.

I began this on another thread over here, if you’re interested in the backstory and some preliminary conclusions and discussions.
Craw drop stats?

Methods

Playing on PC and using a few simple mods, I decided to kill Craw countless times, collect everything he drops, and use the exporting abilities of WillowTree along with some basic Excel calculations to tally up every single part of every single weapon, shield, and COM.

Here are the modifications used:

  • modded backpack size to 9999, in order to fit everything for many many runs at once
  • enabled the console, and used the “notarget” command to freeze Craw so I could kill him with minimal fuss
  • made myself a cheater shotgun w/ 200k+ damage…every little bit of speed helps
  • 4-player slider, set to 4 players to make him drop more stuff each kill. I did a smaller set of tests to establish that only the quantity, not quality, of the drops changes with the slider enabled.

I tried using the only user-created Craw farming map I could find, but it was useless for these purposes (altered drop tables). So, all my testing was done on the official Craw.

I ran one set of ~20,000 drops on each character, and another two with Mordecai and Lilith with their Find Rare Items COMs equipped. So, six sets overall, with each one representing 160-170 kills (4-player mode).

Each kill yields about 120 drops (4 player). A single-player kill is about 65 drops, for reference.


Results

Each table below contains the summed and percentage totals for each weapon part/type. The far right column represents the chance of getting that particular part. These can be multiplied together to get a % chance for a theoretical weapon.
e.g. I want a perfect Torgue Cobra, as do we all. Looking at the percentages:

  • body5 = 82.8%
  • barrel4 = 4.3%
  • stock5 = 13.6%
  • mag5 = 34.1%
  • mat3 = 48.4%
    For the sake of sanity, let’s say I don’t care about the sight.

0.828 * 0.043 * 0.136 * 0.341 * 0.484 = 0.08% chance of getting all those parts. Damn.


Weapons

Repeater Pistols
image

Revolver Pistols

Machine Pistols

SMGs

Rocket Launchers

Bolt-Action Snipers

Semi-Auto Snipers

Combat Shotguns

Assault Shotguns

Assault Rifles (Burst Fire)

Machine Guns

Eridian Weapons

Items

Class Mods

Shields

Grenade Mods

12 Likes

OK, let’s talk about the shiny stuff!
The lists below show how many of each oranges and pearlescents each character found.

image

Not surprising to anybody, but the Equalizer is far and away the most common legendary. It accounts for 20% of all the orange beams. I will discuss this further below; before getting too deeply into the relative abundances, remember that these bulk totals are slightly skewed by having 2 sets each of Lilly and Mordy.

Overall, the total Legendaries = 5,707. Total weapons drops are 62,859.
5707/62859 = 9.079%. So, almost 1/10 weapons he drops is orange. Not bad; too bad 1/5 of these are Equalizers!

image

Pearls are freaking rare! Notice that in >1,000 Craw kills, I did not get a single Bessie or Nemesis. They are so rare that the stats above are of limited value.

Overall, total pearl weapons = 12, shields = 11. Total item drops are 56,908.
12/62859 = 0.0191%. Approximately 1/5250 weapons is a pearl.
11/56908 = 0.0193%. About the same. 1/5175 items is a pearl shield.

Call it 1/5000 - that is the takeaway here…at ~65 drops per 1-player kill and 120 per 4-player kill, that is approximately 1 pearl every 77 single player kills (42 4-player kills). The old hearsay was 66:1 legendary:pearl, but unfortunately it’s more like 250:1.

Better get to work.

12 Likes

Overall Stats


Weapon Types
The table below gives the distribution of weapon and item types across each character.

A few things to note right away:

  • The weapons:items ratio hovers just above 1.1, with the exception of FRI Mordy. We’ll talk about this later.
  • Average weapon and item rarity is virtually identical for all sets
  • Only Lilith gets Eridians. Hooray?

Calibration!
What I would like to do is use the above data to quantify the widely accepted view that the different characters have meaningfully different drop rates for the various weapon types. “Roland finds more machine guns” and “Mordy finds more pistols” are universally accepted, and some digging around in the editor by Scottes seemed to confirm this.

In order to do that, I need to standardize this data set to create a theoretical “base rate”. This base rate would be independent of the weighting applied by which character you choose.

To calibrate the data, I tossed out the FRI Lilith and FRI Mordy data, and used their non-FRI counterparts. Lilith has the smallest number of weapons found (10,208), so I need to truncate the other 3 characters’ data sets in order to compare apples to apples. Doing that gives me this:

image

So, what is this showing us? The “% of bulk” column for each character is the important one here. It shows the amount of each weapon type that each character found, relative to this base rate. Numbers greater than 100% mean that character found more of that weapon type, relative to other characters.

It’s kind of difficult to explain. Let’s look at Brick, and his 144% in Rocket Launchers. He found almost 1.5x the amount of rocket launchers than he would have, if this drop rate was not weighted. Same with Lilith and her 120% in SMGs. This way of presenting the data is not perfect, and I welcome any suggestions on how to improve it. Case in point - Roland and his 98% in RLs. This would lead us to believe he doesn’t find all that many launchers, but in fact he found quite a few more than did Lilith or Mordecai, just not as many as Brick.

This data is nowhere near as conclusive as I had hoped. I certainly believe that the larger deviations from 100% are real. The smaller ones I am not so sure about. For all 3 pistol types, I am unconvinced that there is any difference. MPs have the largest variation in %, but the smallest sample size.

What it does clearly tell us is that you can find any weapon type with any character. You may have a slightly better chance of finding a good Bastard with Roland, or a Bitch with Lilith, but playing with the “right” character is certainly not a requirement. The story is different with Class Mods…


Class Mods
The Class Mods table in the first post looks like a huge mess of numbers at first glance, but there are some striking differences between characters there. The COMs that Craw drops are HEAVILY weighted for the character that you are playing. Between 60-65% of the COMs dropped are for you. The remaining 35-40% are split about evenly between the other three characters.

So, if you want a Professional, play as Lilith. If you want a Patriot or Bad Ass or Gunfighter, you are damn near wasting your time playing as Lilith. If you really want an Atlas Champion for Roland but can only kill Craw with Lilith, better start practicing - using Roland has 5x the likelihood of getting you one!

The rest of the COM parts seem to be exactly evenly weighted, which I find odd. None of the other weapons or items behave this way.


Materials
The distribution of material grades is as follows: Mat1:Mat2:Mat3

  • Weapons: 29,406 : 3,910 : 29,524 or approximately 7:1:7. You have about a 1 in 15 chance of getting that precious Mat2.
  • COMs: 19,118 : 4,224 : 2452 or approximately 24:5:3. Lots of 4-liners out there.
  • Shields: 7,535 : 6,670 : 2,603 or approximately 15:13:5.

Other Notable Stuff
First and most important - in the first post above we have all the part abundances for each weapon type. But the question we have is, are the part distributions the same for legendaries and pearls? Can I estimate the chance of getting specific parts on an Anaconda or Crux?

To answer this, I looked at the part stats for just Equalizers. We have the most of these, so if they mimic the overall revolver stats pretty well, we should be happy. Do they? Sloppy table says:

image

Yeah, pretty much. Left bold column is overall revolver %s, then the Equalizer subset. There’s some erratic behavior going on with the accessories, but other than that it is spot on. I’m happy with that. I think the answer is yes - we can use the part stats in post 1 to estimate our (lousy) chances of getting perfect legendaries.

9 Likes

Find Rare Items and Scavenge Extra Items

Man, these have been debated half to death on the dusty old forums. Conventional wisdom was that FRI doesn’t work, except maybe to increase the number of Eridians. Lots of people still believed in it. SEI was believed to work, but nobody seemed to care all that much. So, I tested both of these - Lilith’s Catalyst and Mordecai’s Scavenger COMs have FRI, and the Scavenger also has SEI.


Find Rare Items
In order to test this, I ran full 20k item sets with Lilith and Mordy, both FRI and non-FRI. I would call the results damning - feel free to check the tables in post 1.

Here’s what it does not appear to do:

  • increase weapon rarity (average rarity is ~identical)
  • influence the relative abundance of the individual parts of weapons.
  • change the relative abundance of weapon types
  • change the relative abundance of manufacturers
  • increase the shield rarity, or influence shield part probabilities
  • increase COM rarity, or influence COM part probabilities
  • make certain COM types more or less common than “normal”
  • increase grenade rarity, or change abundance of elementals or types
  • do anything at all

If anyone can think of any other possibility that I have not accounted for here, please say so! I would love to see this work, but I think if it had any effect we would be able to see it in a data set this large.


Scavenge Extra Items
This one does work, and on Craw himself! Here’s what it does:

  • causes Craw to drop money and ammo in addition to weapons and items. This is a pain, because it gets in the way of picking up other stuff.
  • increases the amount of ITEMS ONLY by about 10%. In the same number of kills, I got 10,232 items with SEI active and 9,343 items without it.
  • again, ITEMS ONLY! This means elemental artifacts, shields, COMs, and grenade mods.

It does not do anything at all to weapons, and does not change the quality of items, only the quantities. So, if you can spare the time, it’s worth swapping to a Scavenger COM just before killing Craw. I speculate that this increases your chances of finding pearlescent shields, but I can’t prove it.

12 Likes

Slow clap that leads into a standing ovation

Seriously nice work good sir. I am still figuring out how I am going to work my test thread once I get the time lol.

4 Likes

Thanks Stouty. Keeping your data organized from the very beginning is super important - try to make sure you never have to ask “what did I do here” or “where did this go, I know I had it…”

You did a pretty damn CRAZY job here !!! :dukeaffirmative:
You’ve won all my respect dude !!! :relaxed:

Now, I can actually say that my “best parts” Bessie won’t be found tomorrow… :sweat: :scream:

2 Likes

Kinda funny to read the chances of finding a Bessie. I found not more than 8 pearls, but 2 of them were Bessies. 3 were Undertakers, just for the record. And 2 Ironclads. RNG at its best!

Great job MeltintheSun for collecting this massive amounts of data. Really appreciateted

1 Like

Hey, it might be! Too lazy to do the calculation now, but on the old thread I estimated about a 1/2000 chance for a DVL590 Barrel4 Bessie w/ Heavy accessory. Some people have certainly found them!

2 Likes

So what your saying is the chances of finding an orange eridian is greater than 1:120000. This doesnt surprise me in the least bit. Did you at least find a fireball(blue one)

Can’t really say what the drop rate is. I only got 20k weapons with lilith, which is the only character that counts in this respect.

No blues, only greens. I would have loved to see something else, cause I have never gotten a blue from Craw or an orange at all. In my normal playing I used almost exclusively Mordy, so had no chance of him dropping one.

I’ve never had craw drop anything other than green eridian weapons, and that’s with a godawful amount of kills with Lilith. Had two legendary eridians, one world drop and one in the armory. I don’t recall having but one blue iirc that is…

1 Like

Interesting, I haven’t done craw runs in years (I had played BL1 on xbox 360) But the only Perals I ever found were

lv 48 Undertaker (Lance Chest) With Brick if memory serves right
lv 48 Tsunami(Secret Armory Run) sigh why couldn’t it have dropped when I was a higher level :frowning:

And from Craw(alll drops only happened in Co-Op Lilith(Me),Lilith, Mordicai, Roland) with the agreemnet being duping was okay as long as it was someone who helped fight craw and no one else.
lv 66 Bessie
Lv 67 Bessie
Lv 68 Bessie
Lv 67 Jackal
Lv 68 Serpens most beautiful thing I ever got in BL1 replaced the Defiler as my armored Craw killer

No idea how the different players affected that but the pearl you never saw was the only one I ever really got to see…
Never saw a legendary Eridian from Craw (only ever greens)or any of the non shock elemental ones ever.
Only Legendary Eridian ever I believe came from a secret armory run…

The point is this puts into perspective how biased our sample was.
My sole hope was for a level 68 Tsunami which never happened ever…

Yeah, even as large as this study is, it needs to be many many times larger in order to draw any meaningful conclusions about whether certain pearls are more common than othere. They are just too rare. Same goes for the blue and better eridians.

1 Like

OMG, OP, that is just outstanding work! Thank you for this info, it’s amazing. I can’t believe you have cleared the mystery of the Find Rare Items COMs.

And yeah, I’m in a similar RNG boat, I don’t think I have ever seen a Bessie drop from Craw himself, and in all these years I have killed him hundreds of time.

Oh well, at least I’ll keep farming to see if I get a good AX Unforgiven masher from him.

2 Likes

First: Incredible work! So much information and the dedication of yours, splendid!

Second: Looking at this I really wonder how rare a Nemesis Invader would be. I don’t think that the Nemesis has the same chances of having the Invader sight as the other stuff but if it does then every 400th Nemesis would be hybrid. Approximately 1/5 of Craw’s drops are Repeater Pistols, regardless of character. At face value that means we would need 385 solo kills on average to even get a single Nemesis, and that is certainly far from a guarantee because random. Then, if it has the same chances, we would need 153,846 kills on average for the Nemesis Invader.
Per Repeater, the chance would theoretically be 1 in 2 Million or .00005%

I really don’t think that’s right and I hope I am wrong about this this but sweet heavens. The only hope I have here is that the Nemesis simply has a higher chance for that sight.

Derp

The only issue I see with your calcs is that repeaters are about 1/5 of the weapon drops, not the total drops. I just ran some quick calcs and it hovers around 11% of total craw drops. So, as bad as you thought your drop rate looks, I think it’s about half that again (1 in 4 million-ish).

This relies on two (seemingly reasonable) assumptions:

  • 11% of drops are repeaters = 11% of pearls are nemeses
  • invader sight occurs on the nemesis at the same rate as other repeaters.

The second I believe is true, based on the stats I ran on the Equalizers (~identical part distribution to revolvers as a whole). While I can’t guarantee anything, I see no reason it would be different. The first assumption above seems logical, but I have no data whatsoever (0 Nemeses found) to back it up.

Happy hunting :confounded:

Oh, and when you do find it, there’s a ~55% chance it will have barrel0, 1, or 2 :frowning:

Oh, derp, I looked at the data in the wrong way. Thanks for the incredibly depressing correction (completely serious).

It sure is funny because the Nemesis was my first Pearl ever back on the Xbox 360. Since then I found quite a few others on PC and PS3 but unfortunately mostly shields, with the Serpens being an exception.

SO can a Pearl ever drop as a Hybrid? I had always heard that they had parts restrictions thus assumed Gearbox would have prevented that seeing as they should have known about the Unique Hybrids,
That said most of the pearls would be unable to be affected as the pearl exclusive part conflicts with the legendary determining part… The Nemesis might be the only one that could overlap as most guns legendary/pearl parts are barrels or elemental capacitors…(I think those are separate in BL1 right? been too long)

The nemesis was my first pearl as well - I remember it fondly…BLR, barrel3, action2, no scope, small mag. Even with slightly crappy parts it was a beast.

Yes, but only the Nemesis. Others seem like they should be possible (matching manufacturers for the defining Legendary+Pearl parts; e.g. Bessie + Skullmasher, Jackal + Bulldog) but are not, for whatever reason.

2 Likes