# The proofs in the spreadsheet

For the record, I am slightly lower than most of the ones who comment the most, but I just looked at one character, my cleric, and all of his rare pieces are between level 7 and level 9, and all of his common pieces are level 12 or 13. It does seem fairly consistent just from that admittedly very unscientific look.

Each character is a replicate, which internally controls for when the player began acquiring gear for that character.

Each character has common and rare weapons at half the rate of other items of the same rarity.

There are 6 slots and 7 common items. All else being equal, if item drop rates were equal, then the probability of the common weapon being the lowest is 2/7. The probability of both weapons being the lowest is 1/21.

With 10 characters, and all of them having the weapons of all the characters lowest is (1/21)^10. Then you square that when you consider the rares too.

Okay. So that’s if all things are equal, which I acknowledge is not necessarily the case.

Yet, how strong must selection bias be to beat out this highly unlikely ((1/21)^20) pattern from emerging over and over again across different user accounts?

Your reasoning A very flawed - a small selection bias over a huge lot of choices in one direction could easily lead to all or most going one way. partic if the epic and legend are considerably better than their counterparts for weapons but not other gear

My wife is demanding a neck massage

Will get back to u

Ok maybe I am maybe I a missing something simple that you are seeing. Admittedly it’s been a long time since I looked at spreadsheets regularly and looking at someone else’s screenshot while on my phone isn’t the best way. However, I see very few instances where there are more commons than rares. Just counting the blank boxes which he said was maxed gear, I see 14 maxed in the rares and 27 maxed in the commons.

Also it wouldn’t be that hard for selection bias to win out if he is a grinder, which guessing from the overall amount of gear he has I would think is true. If he gets even a decent amount of his gear from grinding challenges as opposed to opening chests and cards, he would easily be able to have more of one thing than another. If the card packs drop rates were equal and then he grinded for specific gear and added that on top it wouldn’t take much at all to skew the data.

If every piece evened out at 100 from chests and card packs and you grinded another hundred Of a piece from challenges that piece would be double. So if there were seven slots at 100 a piece, and then grinding another 100. It would only take 1/8 of his total gear coming from grinding a Specific piece of gear to make that double the average

I would think most people just instinctually pick rares over commons almost from the beginning.

? Rets is saying that within rares the weapons drop way less than the other items

Within commons same thing

Ok got it… I still think it is easily explained by choice in challenges. I would have to look to be sure but it seems to me that most of my favorites spells come from something other than weapons so I would pick those items first unless the amount of cards I got for a weapon was much higher.

Also like I said from just looking at cleric pieces all of my rare pieces and all of my common pieces are within a few levels of each other and I am definitely not a grinder.

Agreed re challenges - I think the diff is coming from that selection bias (and possibly that drop rates is for weapon vs other items - so each weapon has half the chance of being dropped)

Don’t agree to within a few levels point - there are very small number diffs required to get to next level a low levels but heaps Of bits required at higher levels. Hence even if drop rates were substantially different would expect everything to be within a few levels

This is the top three dice for each of the top four dungeons I have access to. If you were to able choose 20 ten times in one dungeon and 18 ten times in another, which would be the same odds, you could have a gap as big as 20 off one piece to 140 of another, just by making a choice of one over the other. It wouldn’t take long for the gap to be huge

Dat true

So a very quick look at my characters just rare : archer, warlock, rogue no where near conforming to other data.

Got more of at least one of the rare weapons than most other categories

Haven’t looked at all characters yet

Interestingly I have level 9 rare axes for archer yet I never choose them. I have been collecting the rare bow since day dot and that’s only level 8. Both r on a par or better than other rares some of which I am collecting and some not.

Shows how much effect the random drops of huge amounts have

All of this being said, I don’t mean this to say that I think there isn’t the potential that the odds stated are incorrect. I honestly haven’t thought that much one way or another about it. I am just saying, again, that there are way to many variables in how gear is given/chosen to use a simple count to prove one way or another

Yup my point all along it’s too confounded

I highly doubt it’s malicious coding. What probably is happening is the code checks the slot of the item and then, if there are multiple types (I.e. common and rare weapons), it chooses one or the other. That would naturally lead to lower weapon drops over all. They probably didn’t envision the high demand on rares and commons, so probably didn’t give it much thought.

I have suggested tha may be the case many times in this thread

Excluding Jarlaxle I have maxed out 170 of the 180 non-legendary items. The 10 I am still working on are all weapons. I assumed it was because only weapons have the 2 extra pieces - 1 extra rare and 1 extra common.

2 Likes

Thanks @gpinsky1313. I think that between Tyr, mine and your own collections, we can safely conclude that the drop rates for common and rare weapons are lower (likely half) of the drop rates for the other items of the same rarity.

I suspect that @Turbotarry and @Mkb617 are likely to be suffering from small sample size from their own collections, and so natural variances are masking the decrease in drop rate. I think that only when the collections are large enough (i.e. have sufficient cards opened from chests) that the actual pattern emerges.

At this point it’s a matter of determining whether the model of equal item drop rates or decreased (rare and common) weapon drop rates is more likely to explain the apparent patterns that have manifested it more mature accounts. I think the answer is obvious.

I have already laid out the math above. I invite those who remain contrary to offer an alternative theory on why players might consistently prefer useless non-weapon gear over useless weapon gear.

Note: if you agree with the OP’s conclusions that rare and common weapons drop less frequently than other gear of the same rarity, the main application of this knowledge is to assign a premium to rare and common weapons during challenge rewards or gear purchases, and prefer picking them if given a meaningful choice.

Rets it was me that suggested from the start that one of the likely possibilities was that the drops were weapon, helm, trinket etc not weapon 1, weapon 2, helm, trinket

Ie half probability for weapons

I also said the data that was initially presented was completely confounded - which it was

The data from Pinsky is way more valuable

So I’m not sure why u r acting all high and mighty

This data is as confounded as my own and @Tyr’s; on what basis are you differentiating the quality of the data sets?

If Pinksky is trying to collect the whole set then he won’t be selling

So I should say less confounded - he still collects from the guild and he still get the big random hits of gear etc

I think that from the spreadsheet presented by @Tyr, it is also apparent that he was also trying to collect the whole set.

Not ness a celery

Pinsky started he was, thingy never made that claim nor gave any details about what he was doing at all (selling, buying, collecting, etc etc).

He has huge numbers of things because he has been playing a long time - wouldnt need to set out to get all gear maxed

As someone with a lot of experience in stat design and analysis I don’t like being lumped with data with no details of the confounding variables

Try getting data of that type and hasty conclusions like yours into a good journal - I don’t think so