HUGE Magic Find Analysis - how magic find affects the loot

[Classicclean]

Hello! I just came over this post on diablofans.com and thought I should share it!

Greetings fellow gamers.

As a first I would like to point out that english is not my native language. Spelling errors are likely and my grammar is probably horrible, but I will do my best.

Recently I have noticed a lot of threads appearing regarding the item affix Magic Find and how it actually works. People usually get the same responses; they get told that MF works the same as in Diablo 2 and people refer a lot to blue posts. And even in spite of all of these fine answers, there are many questions that keep rising: Does MF affect item lvl? Do I get more items with more MF? Does MF increase the rolls on item stats?

I thought about it for a bit and I decided to jump into collecting a lot of data that could back up the statements that people use out there. Some of you may already know that I have performed statistical work about treasure creatures before.

In the following I will therefore present my findings. Please note that these findings are not final; the results will most likely be continually updated as I increase the data pool! So please, if you think the sample sizes are not large enough - give the data time to grow. As long as I can collect data this thread will be updated and potentially the data pool may become large enough to be valid for direct calculations.

- - -

Contents:

1. Introduction
1.1 Reason for this work
1.2 Previously known statements about MF
1.3 [Important] magic find and nephalem valor
1.4 Description of the procedure for data collection

2. Data Analysis and Results
2.1 Data and sample size
2.2 Gear quality as a function of magic find
2.3 Item level as a function of magic find
2.4 Number of items found as a function of magic find
2.5 Bonus: treasure creatures spawn type rates
2.6 Treasure bandits and blacksmithing plans

3. Conclusion

4. Outlook

- - -

Abbreviations:
ATR: attributes
GC: gear class (being common/magic/rare/set/legendary)
iLvl: item level
L: legendary
M: magic
MF: magic find
R: rare
S: set
TB: treasure bandit
TG: treasure goblin
TS: treasure seeker
TP: treasure pygmy
W: white

- - -

1. Introduction

1.1 Reason for this work
I simply wanted to assign as much information about magic find I could in a single thread so it may serve as a single viewpoint for newcomers and older members alike, should they feel in doubt about how magic find works. I hope that the data I collected will help display how magic find works and I hope people will realize how elaborate this kind of data-collection is - especially with the requirement of manual writings due to the use of 3rd part program being in contradicting with the Terms of Use for D3. In the long run this collection of data may serve as a baseline for computations on MF and potentially interesting parameters from these computations can be extracted e.g. what level of MF is required to wipe white items off the drop board.

I would like to point out, should anyone feel to start such discussions, that we are all human. If I have made a mistake in one of the following sections; please correct me on it and bring some proof that I am mistaken. This project has turned out to be larger than I first presumed and hence errors are likely to occur, but changes may also be made in the future which can make these findings misleading.

1.2 Previously known statements about MF
A lot of threads have been going up since launch in which MF is discusses or questioned. The best proof there is as of how MF really works is based on the Diablo III Developer "AMAA" transcript in which the question related to this topic can be found here.
There has also been some blue posts (some people state this and I think I have seen some of them myself) about how MF affects rolls. I will need to find these posts before the following can be correctly argued, however, MF should not affect the rolls on iLvl nor the ATR/affixes rolls on those items. I will come back to edit this once I find the related posts, however, I have in my research looked into the effect of MF on iLvl and luckily my data backs up the above statement.

A lot of people tend to give the following example about the effect of MF:
>>If you have 0.5% chance to get a rare item with 0 MF, boosting your MF to 100 would increase the chance to 1.0% of getting that rare item.<<
While this is a very fitting example for how magic find works as a mechanic I am not entirely sure it explains it in its' full aspect. There is a negative feedback system that should be taken into account as well. For instance, if you have 30% chance of getting a whilte quality item, 50% chance of a magic quality item and 20% for a rare quality item, it is in theory possible to boost magic find up to a point of which you will have ~100% chance of getting a rare quality item and close to 0% for a magic quality item. And here you might think - but that does not contradict the above statement: in fact it does. The above statement can be used for a magic item instead of a rare and if that is taken into consideration, then it says that it is possible (with enough MF) to reach 100% dropchance for a magic item without considerating the other qualities and in we should therefore only find magic items with high MF which is clearly not the case.

1.3 [Important] Magic Find and Nephalem Valor
As Ulto has pointed on page 4 of this thread nephalem valor changes these presented results with a possible negative feedback: Farming elites with nephalem valor buff stacked (or bosses for that matter) results in a guaranteed drop of 1 (or 2 for five stacks for bosses) rare. This will not change the effect of magic find in itself, but it will change the effective numbers (and most likely decreasing them) as a result, as the presented analysis (for now - remember to see 4. Outlook) is based on no guaranteed drops and thereby flat baseline droprates. When I have performed some more calculations on this direct effect, this section will be updated, but it is at this point assumed that the feedback is negative.

It is therefore important to state that the presented results cannot be directly applied to elite farming before more elaborate research is applied - and I will get into that in the following week.

1.4 Description of the procedure for data collection
I will give a short description of the procedure I used to collect the data that can be found in the analysis. I have simply farmed treasure creatures with varying values of MF on my gear. Yes it is again these poor gobos who met my datahammer! Muahahaha. I did not involve any NV buffs nor did I use gear swapping for the data collection, just to make sure the data are completely "clean".

Just for making it clear, it is these guys who helped me collect the large amounts of data:


Treasure Goblin Treasure Bandit Treasure Seeker Treasure Pygmy

After each kill on a treasure creature (either goblin (TG), bandit (TB), seeker (TS) or pygmy (TP)) I noted the following information from the items dropped:
Total number of drops (except potions).
Number of gear of white/magic/rare/set/legendary quality.
Number of Tomes of Secret, gems and plans/designs.
The iLvl of the items dropped within their respective gear quality range.
I wrote all of these findings into one excel file (this can be very time consuming).






2. Data Analysis and Results

2.1 Data and sample size
My datasheet can be downloaded here, should you be interested.
I found a total of 858 treasure creatures of which I managed to kill 832 of them (the rest simply managed to port away due to low dps in high MF gear or due to silly monsters backstabbing me).
From those I killed I have collected six data points of varying magic find (0%, 76%, 150%, 234%, 252% and 270%). These data points provide average values over 100, 100, 100, 200, 200 and 100 creatures killed, respectively. It requires many creatures to assemble just one data point, so my charts are still weak predictionwise, but they should be quite precise.

2.2 Gear class as a function of magic find
I will start out with the most important results: testing the class of gear (white/magic/rare/set/legendary) as a function of MF.

The data results are (I will put these values into a table or similar later):
MF=0 : W=32.49%, M=57.67%, R=9.84%, S=0.00%, L=0.00% (Sample size: 437 items)
MF=76 : W=31.52%, M=54.57%, R=13.91%, S=0.00%, L=0.00% (Sample size: 460 items)
MF=150 : W=30.30%, M=51.30%, R=18.40%, S=0.00%, L=0.00% (Sample size: 462 items)
MF=234 : W=26.28%, M=44.54%, R=29.06%, S=0.00%, L=0.11% (Sample size: 898 items)
MF=252 : W=24.10%, M=42.57%, R=33.00%, S=0.00%, L=0.34% (Sample size: 888 items)
MF=270 : W=23.17%, M=40.83%, R=36.01%, S=0.00%, L=0.00% (Sample size: 436 items)

The values are rounded to the second decimal.

!!!Please note that these values are NOT droprates - they are the items I found on my farming and therefore only reflect the droprates from goblins, they do not define them nor can they be directly generalized to elite/boss farming (especially due to the guaranteed drops from the Nephalem Valor buff which has a negative feedback on MF) until more research is made.

The amounts of items found being set and legendary items are extremely few compared to the amount of white/magic/rare items found. It is therefore only those three quality levels I will base my charts on.

Below is a chart showing my six data points of which the % of total gear found for GC of common(white), magic and rare. This chart should help illustrate the effect of magic find on the class of the obtained gear.





It should now be easily concluded that magic find increases the class of the gear found.

2.3 Item level as a function of magic find
I also sampled the iLvl of all the items dropped and collected them into a chart. If magic find increases the iLvl of the items found, then it would should be clearly visble. The chart below shows the found results. It is created from a total of 3137 items.





it is immidiately obvious that MF does not seem to affect the iLvl of the gear dropped. The sample size is not the largest but it is still large enough to give a representative picture.

As a little bonus is has previously been posted on a blog that the iLvl found in act II are 19% for iLvl 61, 12% for iLvl 62 and 4% for iLvl 63. As it was found above the iLvl of the dropped items are not affected by MF, so just for shits and giggles I calculated my found distribution of items from iLvl 51-63:

iLvl Found %
51 0.7
52 2.8
53 0.8
54 0.7
55 8.7
56 4.4
57 4.0
58 11.3
59 4.8
60 22.3
61 20.7
62 14.8
63 4.1

Remember this only reflect the loot in act II !

!!! It has just come to my attention that these drop rates will soon be changed as a result of a hotfix (source). While this will change the numbers, the results still stands their ground that iLvl is not affected by MF. The results will be updated when a new sample size is collected.


2.4 Number of items found as a function of magic find
Again, these findings are most easily represented in a chart.





Again the sample size is not the largest, but I honestly think it's large enough to illustrate the finding that magic find does not increase the number of items found.


2.5 Bonus: treasure creatures spawn type rates
As a little bonus I can also present the relative spawnrate of treasure creatures. I found a total of 858 of which 223 were TG (26.0%), 215 were TB (25.1%), 208 were TS (24.2%) and 212 were TP (24.7%). I think it's fairly safe to presume the spawnrate is 25% for each type.


2.6 Treasure bandits and blacksmithing plans
While I do not have enough data to conclude anything about this (yet), it did seem through my data collection that treasure bandits have become very pleased with dropping blacksmithing plans!
I do not think MF has an effect on them, but with a larger sample it may become clear. Until now I will just give you lot a list with the plans I have collected.

Blacksmithing plans only dropped for me from bandits (except from a single pygmy who dropped plans for Exalted Fine Strike Wand), but as these can generally drop from any creature don't take it as a surprise to find some dropped by the three other types. Off the 215 bandits i managed to kill, the following plans dropped:
Exalted Flesh Ripper
Exalted Phantom Bow
Exalted Piercer
Exalted Slag Hammer
Exalted Fine Pallium (x2)
Exalted Fine Slag Hammer (x2)
Exalted Fine Sovereign Greaves

3. Conclusion

So, as a result of these findings and to re-imburse the statements from blue posts and elsewhere.
Magic find increases the class ("item-color") of the gear found.
Magic find does not increase the number of items found.
Magic find does not increase the iLvl of the gear found.
Whether or not magic find affects affix rolls is not investigated here, but there is no reason as of to suspect that magic find would increase the stat rolls.
Last and most important: Future data collection may help to provide data for computational research on magic find. It will also be investigated if this research is directly applicable to elite farming.
Magic find is not useless, as some people may have pointed out it may feel like. It increases the class of the gear found and as the % of iLvl is static, the more MF you have the more items of higher quality will you receive in the long run and it may therefore seem (as you get more useful gear) like you get higher iLvl gear.


4. Outlook
People are more than welcome to contribute to this research. All I am asking for - if they wish to contribute to the research of section 2.2 - is to keep track of white/magic/rare items dropped with the MF they are running with (and if NV is included note the stacks as well) and note what they are killing (elites/bosses/gobos/ponies/whatever). The research I am doing at this moment is focused on goblins, but I won't be killing them forever and it would be nice to see if elite farming has similar results (even though they are influenced by guaranteed rare drops). One thing that would be very interesting is to make a large sample with 0 MF as this gives a hint about the baseline %'s.

I hope that the original post will eventually be so well constructed that it will serve as a single thread where people can get all of their questions about MF answered. And in order to get there I most likely need a bit of help.

People are also more than welcome to find relevant blue posts / similar official statements or previously performed research on MF (such as how the rolls are made) and contribute with this, as the original post will be continually updated with such. Just make sure it is reasonable.

- - -

I hope you lot can use this. And as I mentioned in the introduction: as long as I have stuff to farm I will increase the size of the data pool and keep this thread updated. If you have something to add, feel free to post below or pm me. I have most likely forgotten something somewhere in the mess!

.

TL;DR: MF increases only the class (magic/rare/set/legendary) of the gear dropped, not the amounts or the iLvl. See the graph in 2.2 for an illustration of gear class vs. magic find. Potentially this research may lead to computations on MF.

[naniqs]

Really nice post!, like it thanks for sharing!

Cant rep atm, but i will rep tomorrow have given out 2 much rep today :S

+ rep!

[TanookiLover]

to sum up, blizzard blues were telling the truth lol

interesting to see that magic find after 150 almost doubles the rate of rares

good work, hopefully this will curb the amount of people spouting nonsense as fact

[CuT]

[Ongoing Project] Magic Find and its efficiency: A statistical insight - Theorycrafting and Analysis - Diablo Discussion and Community Forums - DiabloFans.com

Posted 28 June 2012 - 09:14 AM

Can't tell who was first :P Thanks if it was you

[d3trader]

Great post nonetheless, wouldn't have found this info otherwise. Thank you

[Storm301]

That's basically how most of us have always assumed MF works since D2, but there is definitely some interesting data there. I'm surprised that the overall amount of blues decreases from 76 mf on. I would have expected blues to increase at lower MF's and then decrease later on because of the larger amount of rares. I wonder if it would be possible to write a program that reads the memory of D3 and tracks MF, the creature killed and the drop quality, without doing anything that would get a player banned. I would like to see a compiled list over tens of thousands of kills across all of the different creatures to figure out if certain areas are better to farm.

[Classicclean]

[Ongoing Project] Magic Find and its efficiency: A statistical insight - Theorycrafting and Analysis - Diablo Discussion and Community Forums - DiabloFans.com

Posted 28 June 2012 - 09:14 AM

Can't tell who was first :P Thanks if it was you

I just wrote into the first line that I've copied the post from diablofans.com! Just sharing it. I just edited the post to have a clear overwiev!

[Classicclean]

Do I have to make it fat? Just read the first line of my post ...

[coolme]

Thank you for the wonderful post, very informative +rep

[Kylah]

Nice find, even if you didn't do the original research. +rep, er once I "spread it around" I apparently gave you rep recently. You're just all kinds of helpful

[Classicclean]

Nice find, even if you didn't do the original research. +rep, er once I "spread it around" I apparently gave you rep recently. You're just all kinds of helpful

thank you really appreciated

[joshuazzz]

WTF is wrong with people like you?
Do you have a hobby called "making sure everyone gives credit or I'll make him look bad"?

He stated right up-front that it was obtained from somewhere else. I KNEW this the moment I read the post.
And it was a freaking nice post to read too... something I would have missed otherwise.

Thanks Classicclean for the read.

Give credit? people are thinking you're the one who made the post.

[Classicclean]

WTF is wrong with people like you?
Do you have a hobby called "making sure everyone gives credit or I'll make him look bad"?

He stated right up-front that it was obtained from somewhere else. I KNEW this the moment I read the post.
And it was a freaking nice post to read too... something I would have missed otherwise.

Thanks Classicclean for the read.

Thanks, atleast one person reads the first line

[TanookiLover]

they should have posted what little legendary data they had so we can at least guess how it's affected by mf
if is treated the same, that would mean there's a max mf point where it actually lowers the number of rares and raises the chance of legendary? or they both eat into the magical item rate? or mf doesn't affect lengendary?

[xPocky24]

Good read!