Compendium of MF Mechanics and FAQ

General Discussion
1 2 3 10 Next
Highly Rated
Information current to 1.0.5

1.0.4 Ended the era of the "Unidentifed Item Code" exploit, wherein it was possible to determine the exact number of affixes on an item. This bug was significant, because it allowed us to compare how Blizzard said the magic find system worked with actual results. I would have loved to get a larger set of data to work with, but I'd like to give the best summary I can muster based on the limited data that was made available.

I will be pulling HEAVILY from this thread:

Many people contributed to this project, but I would like to extend special thanks to Cyeron, who organized and managed the project, and to Nubtro and Murskautuminen who contributed some key ideas and data in the final days. I'm not going to give credit for every thing in this post point by point because it will make it clumsy to read, but I hope you'll read the original thread and see where this information came from.

Introduction and Basics
The focus of this guide will be on elite kills. Many of the concepts and math can map onto goblins, bosses, destructibles, and trash, though varations exist.

As the game guide tells us, magic find affects the step of the loot determination process when the game decides how many affixes an item will have. The game first "rolls" to see if you get a legendary. Failing that, it rolls to see if you get a 6-Affix piece of loot, then 5-Affix, etc.

There are two numbers of interest in such a system. There's the nominal chance of a roll hitting and there's the resultant chance of a piece of gear being that. For example, if there is a 10% chance of rolling a success on the 5-affix roll and there's a 95% chance that the 6-affix roll failed previously, 10% is the nominal chance of 5-affix, but only 9.8% of our loot will actually be 5-affix because producing that loot depends both on the nominal chance and on the chance of previous rolls failing. The nominal chances do not have to sum to 100% but the resultant chances will.

Slot Theory (special props to Nubtro and Murskautuminen here)
Monsters have drop tables. It's list of all the things they can drop. It contains all possible gear drops, tomes, gems, etc. If you record the game and watch a dying monster in slow motion, the order in which loot pops out is actually reflective of this table. It's a list and the game generates the drops in order. For an elite, there are 3 gear drops that are guaranteed to always drop. They are gear drops 1, 2, and 4. Gear drop 3 has a 50% chance not to drop a piece of gear. Gear drop 5 is activated 100% of the time upon achieving 5NV. This produces 3.5 drops per elite on average before 5NV and 4.5 after. Here is the drop order for an elite (taken directly from the source listed above):

Gear drop #1
Gear drop #2
Tome drop
Gold drop #1
Gold drop #2
Gold drop #3
Gold drop #4
Gear drop #3
Gold drop #5
Gold drop #6
Gold drop #7
Gold drop #8
Gear drop #4
Gear drop #5 - NV drop
Globe of Health drop

The gear slots are not created equal. Although the game guide does not detail this, the nominal chances of 6-affix, 5-affix, etc vary per slot. I would have liked to have more data on this, but what data we do have indicates to me that gear drop 1 and 4 have identical tables, that drop 2 and 3 have identical tables, and that drop 5 has the same chances of legendary, 6 Affix, and 5 Affix as slots 1 and 4 but its nominal chance of 4 affix is 100%.

The following numbers will not be exactly right, but they seem to match our data very closely and I'm not sure we'll ever be able to refine them additionally. Also, the chance of legendaries is so low that our data does adequately capture their drop chances, so I'm excluding them here. Their chance to fail is so close to 100% that it would have no noticeable effect on the other numbers anyway.

It's difficult to capture the rates of magic items in chart form here, because the existence of items that can only roll as magic or better (jewelry, follower special, etc) skews the magic numbers. For the sake of demonstration, though, I'll post the rare numbers so you can see the process:

Base 0MF Values

Result - Nominal

Gear Slot 1 and 4 at 0MF:
6 Affix: 1.6% - 1.6%
5 Affix: 3.9% - 3.9%
4 Affix: 11.1% - 11.7%
2 Affix: All Others

Gear slot 5 (theoretical guaranteed rare from 5NV @ 0MF)
6 Affix: 1.6% - 1.6%
5 Affix: 3.9% - 3.9%
4 Affix: 94.5% - 100%

Gear Slot 2 and 3 at 0MF:
6 Affix: 0.1% - 0.1%
5 Affix: 0.2% - 0.2%
4 Affix: 0.4% - 0.4%
Magic and White: All Others

Magic Find acts on the nominal chances for each drop. As the chances for higher affix-count items goes up, it can reduce the occurrence rate of things lower on the table. This effect is called "pushing it off the table".

Here is the table with 200MF applied:

Gear Slot 1 and 4 at 200MF:
6 Affix: 4.8% - 4.8%
5 Affix: 11.3% - 11.8%
4 Affix: 29.5% - 35.1%
2 Affix: All Others

Gear slot 5 (guaranteed rare from 5NV) at 200MF
6 Affix: 4.8% - 4.8%
5 Affix: 11.3% - 11.8%
4 Affix: 84.0% - 100%

Gear Slot 2 and 3 at 200MF:
6 Affix: 0.2% - 0.2%
5 Affix: 0.6% - 0.6%
4 Affix: 1.2% - 1.2%
Magic and White: All Others

Here is a chart that compares all the data that was collected in general with my extrapolations based on this data. It compares general results (R%/M%/W% at various MF levels) vs the results I got by taking the limited data we had that broke out the slots individually at 0MF and extrapolating that forward.

Here is a summary of the slots at various levels of MF. These are the results values (the actual % you will see in game). They can be calculated using the MF and the nominal values shown previously:

Loot tables at growing MF values

Affix Count - Slot1&4 - Slot5(NV) - Slot2&3 - Avg w/o 5NV - Avg w/ 5NV

1&4 5(NV) 2&3 Avg w/o 5NV Avg w/ 5NV
6-Affix: 1.59% 1.59% 0.05% 0.9% 1.1%
5-Affix: 3.88% 3.88% 0.19% 2.3% 2.7%
4-Affix: 11.05% 94.53% 0.41% 6.6% 26.3%
2-Affix: 83.47% 0% 4.41% 50.2% 38.9%
1-Affix: 0% 0% 8.48% 3.6% 2.8%
White: 0% 0% 86.47% 36.4% 28.2%
Total Rare: 16.53% 100% 0.64% 9.8% 30.1%
Total Magic: 83.47% 0% 12.88% 53.8% 41.7%

1&4 5(NV) 2&3 Avg w/o 5NV Avg w/ 5NV
6-Affix: 2.78% 2.78% 0.09% 1.6% 1.9%
5-Affix: 6.71% 6.71% 0.32% 4% 4.6%
4-Affix: 18.52% 90.5% 0.71% 11% 28.9%
2-Affix: 71.98% 0% 6.03% 44.2% 34.3%
1-Affix: 0% 0% 11.4% 4.8% 3.7%
White: 0% 0% 81.45% 34.3% 26.6%
Total Rare: 28.02% 100% 1.12% 16.7% 35.4%
Total Magic: 71.98% 0% 17.43% 49% 38%

1&4 5(NV) 2&3 Avg w/o 5NV Avg w/ 5NV
6-Affix: 3.98% 3.98% 0.12% 2.4% 2.7%
5-Affix: 9.47% 9.47% 0.46% 5.7% 6.5%
4-Affix: 25.3% 86.55% 1.02% 15.1% 31.1%
2-Affix: 61.25% 0% 7.64% 38.7% 30%
1-Affix: 0% 0% 14.18% 6% 4.6%
White: 0% 0% 76.58% 32.2% 25%
Total Rare: 38.75% 100% 1.6% 23.1% 40.4%
Total Magic: 61.25% 0% 21.82% 44.7% 34.6%

1&4 5(NV) 2&3 Avg w/o 5NV Avg w/ 5NV
6-Affix: 4.77% 4.77% 0.15% 2.8% 3.3%
5-Affix: 11.27% 11.27% 0.56% 6.8% 7.8%
4-Affix: 29.45% 83.96% 1.22% 17.6% 32.5%
2-Affix: 54.51% 0% 8.7% 35.2% 27.3%
1-Affix: 0% 0% 15.96% 6.7% 5.2%
White: 0% 0% 73.42% 30.9% 24%
Total Rare: 45.49% 100% 1.92% 27.2% 43.5%
Total Magic: 54.51% 0% 24.66% 41.9% 32.5%

1&4 5(NV) 2&3 Avg w/o 5NV Avg w/ 5NV
6-Affix: 6.37% 6.37% 0.2% 3.8% 4.4%
5-Affix: 14.77% 14.77% 0.74% 8.9% 10.2%
4-Affix: 36.89% 78.86% 1.62% 22% 34.8%
2-Affix: 41.98% 0% 10.8% 28.9% 22.4%
1-Affix: 0% 0% 19.34% 8.1% 6.3%
White: 0% 0% 67.3% 28.3% 22%
Total Rare: 58.02% 100% 2.56% 34.7% 49.3%
Total Magic: 41.98% 0% 30.14% 37% 28.7%

1&4 5(NV) 2&3 Avg w/o 5NV Avg w/ 5NV
6-Affix: 7.96% 7.96% 0.25% 4.7% 5.4%
5-Affix: 18.15% 18.15% 0.93% 10.9% 12.5%
4-Affix: 43.2% 73.89% 2.02% 25.9% 36.6%
2-Affix: 30.69% 0% 12.88% 23.2% 18%
1-Affix: 0% 0% 22.48% 9.5% 7.3%
White: 0% 0% 61.45% 25.9% 20.1%
Total Rare: 69.31% 100% 3.19% 41.5% 54.6%
Total Magic: 30.69% 0% 35.36% 32.7% 25.3%

So this is the basic mechanics. I'd like to follow this up with a FAQ section, so feel free to ask questions and I will do my best to answer them.

For what I have learned about affix selection and rolling, I created this thread:

There is also this ongoing supplemental research:

Thread regarding top-rank affix rolls being more common on 2-affix gear than rares:


This section is for some additional information.

ilvl drop rates per Act according to this:

With Monster Power Disabled:

Hell III and IV Inf. I Inf. II Inf III&IV
ilvl61: 13.9% 23.9% 23.3% 27.1%
ilvl62: 3.45% 12.6% 18.6% 21.7%
ilvl63: 0% 4.8% 9.3% 16.3%

With Monster Power Enabled (MP>0), all Inferno Acts have the same drop rates as Inf III&IV

Revision History:
9/20/12 - Major Revision regarding MF effectiveness and effect of magic+ items. Details:

Technical Functions of Magic Find
Q: What does Magic Find affect besides the number of affixes?
A: At this time, there is no evidence that it has any effect on ilvl, slot, durability, which affixes are chosen, which ranks of affixes are chosen, and how those affixes roll within those ranks

Q: When does the game look at my Magic Find?
A: Magic Find is surveyed when the monster is killed and the loot drops. For this reason, "gear swapping" is possible, wherein players take the majority of a monsters health wearing normal gear and then swap to high Magic Find gear just before finishing the kill. This still works as of 1.0.4, although the paragon system is designed to make this less attractive over time. Rare drops, though unidentified, are already 100% determined, and nothing can be done after they drop to improve their quality in any way.

Q: Does the percentage of 6-Affix rares go up with MF?
A: Yes, though by how much depends on whether you mean the percentage of rares that are 6-Affix or the percentage of loot that is 6-Affix. The percentage of loot that is 6-Affix goes up like you would expect: 200MF means that 3x more of your loot will be 6-Affix. However, if you look only at the rares, you will see an additional smaller trend that is the result of pushing 5-Affix and 4-Affix slightly off the table: the proportion of rares that are 6-Affix slowly creeps up, but the effect is slight and does not really have a macro effect.

Q: Does Magic Find give you better loot or just more loot?
A: The average piece of gear that drops will be better with more magic find because it will have more affixes, allowing additional mods to be applied. It also does not increase the total number of drops, so the result is better loot and not more loot.

Q: What is the difference between a rare (yellow) and a magic item (blue)?
A: Only the number of affixes. Rares have 4, 5, or 6 affixes. Magic items (blues) have 1, 2, or 3, but note that 3-Affix blues do not drop. They can only be crafted.

Q: Does Magic Find work on Legendaries?
A: It is hard to answer this question with 100% confidence because of their low occurrence rate, but I believe they do. Of the 16.5k dropped data that could have been legendary or set, 17 of drops were legendary or set items. All 17 of these occurred at 200MF or higher, even though ~37% of the data was gathered at 0MF or 75MF. The number of them is so low that there is no apparent trend in the data (we have less than 10 drops at each MF level), but I think the fact that none were found at lower MF levels indicates that MF is working on them. Remember, too, that the game guide's description states explicitly that MF affects Legendary/Set rolls.

Q: Is there a cap to Magic Find?
A: A 300MF cap has been instituted. This includes gear, fortune shrines, paragon levels (3% per plvl) and follower bonuses.

Q: How does Nephalem Valor, Paragon Level, and Monster Power work on Magic Find?
Each stack of Nephalem Valor gives 15% MF up to 5 stacks (75MF). This can exceed the 300MF cap. 25% bonus MF is granted per Monster Power level up to a 250% bonus at MP10. This can exceed the 300MF cap, as well. 3% permanent MF is granted per Paragon Level up to a 300%MF. This is subject to the 300MF cap.

Q: If the maximum number of affixes is 6, why do I see some rares with more than 6 properties?
Affixes apply mods. What you see on an item is the mods applied. Most affixes apply just one mod (thorns, crit chance, etc). However, a few affixes apply two mods such as +min dmg/+max dmg or +vit/+str. Only one of the dual-primary affixes can be applied. I suspect you can have as many as 9 mods on a rare, since you could have +min dmg/+max damage, a dual-primary affix, and an intrinsic guaranteed value like IAS on quivers.

Q: Are you sure there are no 3-affix blue items dropped? I've seen plenty of 3-affix blues.
What you're seeing is 3-mod blues with only two affixes. Magic items can have up to 1 prefix and up to 1 suffix (or both), but that is the max they can have. Note that rare affixes are not constrained by "prefix" or "suffix". See previous question for additional clarification.

Q: Does Magic Find have diminishing returns?
A: It depends on what you are hunting. If you only want 6-Affix, then it does not. 0MF gives 1x, 200MF gives 3x, 400 MF gives 5x. However, if you also want 5-Affix and 4-Affix, then there is a slight diminishing of those over time. Base chance for a rare is around 10% when we exclude the guaranteed 5NV rare. 400MF, though, does not give 50% rares. It only gives about 42% because of the 4-Affix rares being pushed a bit off the table.

Q: How does the 5NV guaranteed rare affect overall drops
A: This is an independent drop that is in addition to what you would otherwise get. It has no impact on any other drops. The 5NV drop does have a >0% chance to be legendary/set, and its 6-Affix and 5-Affix chances are the same as gear slots 1 and 4, which are the "high quality" drops.

Q: How many rares do I get per elite pack?
A: You can use the tables above to calculate this. If you have 5NV, you'll have 4.5 drops per pack. Without 5NV, you get 3.5. If you have 300MF with 5NV, you'll get 4.5*49.3% = 2.2rares per elite group. If you only care about 6-Affix rares, replace 49.3% with 4.4%. If you don't have 5NV, replace 4.5 with 3.5.

Q: How do I know whether it's better to run Act 3 with 75MF (5NV only) or Act 1 with 400MF (including 5NV)?
A: Start by calculating the number of desirable drops per elite pack. This is based on the calculation in the previous questions times the percentage of drops that are the ilvl you want (whether 63 only or 63 and 62, etc). iLvl drop rates are posted at end of Original Post for reference. For example, if I want ilvl 62 and 63 drops that are 6-Affix and I am running Act 1 with 400MF, this is how many I will get per pack: (4.5 * 5.4% * (4.8%+12.6%))=0.04. I then have to measure my elites packs per hour and multiply in. If that number is 30 elites per hour, then my total haul is 0.04 [drops/elite] * 30 [elites/hr] = 1.2 drops/hr. Compare this across the different acts and your different MF levels and see what provides the most benefit. Keep in mind that there are some hidden benefits of later Acts such higher chances of dropping high quality patterns. See an example of this sort of calculation here:

Does Magic Find work on every loot drop?
A: No, in 1.0.3, Blizzard removed their effect on destructibles, corpses, etc.

Q: What are the sources of Magic Find?
A: You can find gear with Magic Find, Nephalem Valor gives 15% per stack up to 5 stacks, fortune shrines give 25% magic find temporarily, and 20% of whatever magic find gear you put on your follower is given to you. Note that you get the 20% from your follower even if her or she happens to be dead when you kill the monster. 1.0.4 adds the paragon system, which adds 3% MF (and GF) per paragon level up to 300% max. Note the FAQ about Magic Find cap.

Q: Does it matter if another player gets the last hit when applying Magic Find?
A: No, as long as you are in range of the kill and gear drops, you will be given gear that is reflective of your personal Magic Find.

Does Magic Find work is I let my follower, pets, or hydras get the last hit?
A: There is a bug report that currently says that Wizards are not getting their MF when hydras get last hit. This was tested by myself, another player, and the dev team, and the results show that the MF is working as expected.

Q: Why do I sometimes not get even a single rare from an elite pack even with 5NV?
A: Unless there is a bug of which I am not aware, one of the following is probably true: either you had four stacks before the kill and five after and mistakenly believed that you killed them with five stacks or the elite was a scripted (purple) elite, for which no bonus NV rare is given.

Q: According to the guide, I should be getting # of * type of drop, but I only got # in # runs. Could your numbers be wrong?
A: Yes, the numbers could absolutely be wrong. However, we've gathered enough samples now that the numbers are probably pretty close to correct. We all know that it takes a "sufficiently large sample" to prove that the numbers are not right, so here's how you can compare your results to the data and see if you've logged enough data to indicate that your deviant findings are probably not due to chance. Call "p" the chances that I've listed for something to happen. For example, p for "ilvl63 rare in Act 1 at 0MF" would be 4.8%*9.8%=0.47% or .0047. Call "n" the number of samples you get. For example, killing 200 elite groups would give n~700. Finally, "x" is the number of success you actually measured. If you ran warrior's rest at 0MF 200 times and got 2 ilvl63 rares, x=2. The chance for that to occur randomly if my numbers were right is (e^-pn)*(np^x)/x!. e is ~2.718. In the case shown here, the chance of 2 rares would be 20%. Chance to get 1 rare is 12.2% and chance to get 0 rares is 3.7%. If you log something like this and find that the chance to get your results is very low (<2%?), please let me know so we can take a close look at our numbers! Note that this is an approximation of the true binomial function, and really only works right with large sample sizes and small chances (which is what we are usually dealing wit here).

Q: I'd like to get the exact number for 250MF. Can I just interpolate them?
A: Yes and no. The affect on 6-affix numbers by MF is linear because nothing is pushing it off the table, so they can be exactly calculated. If you wanted average ratio of 6-affix rares at 250MF with 5NV, you could do (250-300)/(200-300)*(3.3%-4.4%)+4.4%=3.85%, which is exactly correct. You can also get approximate numbers for other values this way, but they may be a few percent off.

Q: Regarding Hellfire Rings, does magic find affect key and organ drop rates?
A: No, not at all.

Q: I want to buy Unidentified rares. Should I care about the MF of the seller?
A: Somewhat. 400MF doesn't make his rares 4x more valuable, but the chances that a rare from a high MF player will be 6-affix is 25%-50% higher. Details:

Questions about data gathering and analysis
Q: Did the data come from Act 1 or multiple Acts and could Act affect these numbers?
A: The sequence data (breaking down the drops by slot and finding the % drop of each affix type) was all from Act 1. A great deal of data was gathered in Act 2 as well, but it is only R%/M%/W%. It is obvious from the non-sequence data that the chances of rares, blues, and whites are the same regardless of Act. We can conclude from this that the distribution of rares must be the same in later Acts. If it were not, then even if the base chance at 0MF were the same, the drop rates would drift apart as MF rises. Since the data extrapolated from the Act 1 sequence data matches all general drop rates across multiple Acts, we can safely conclude that these numbers should apply elites/champions in all Acts.

Q: I am a casual player, but would like to help contribute. What can I do?
A: This answer is a bit long, so please refer to this post later in the thread:

This is an epic post, I can only imagine this took a ton of work, time and effort.

Basically, nominal chances for 6, 5, and 4 affix drops are multiplied by (MF + 1). Nominal chances for 2 and 1 affix drops are multiplied by (MF/2 + 1).

I believe the bolds should read as
(MF/100 + 1)
(MF/200 + 1)

Edit: Can't believe I forgot to say Thank You. D'oh.
Thank you, this work you've done is an incredible service to the community.
Fair enough. I was thinking of MF as a %, but I suppose most people think of it as a 0-400 (not 0-4.0) value. I'll update accordingly.
Added some additional resources at the bottom that may be of interest.

you and your partners on this project are truly selfless to spend so much of your time on a post that may get buried under the avalanche of flame threads here :(

This is the type of info that Blizzard really ought to release to the public. I'm not sure what the point of forcing players to extrapolate this data for themselves is.

This isn't the type of information that hurts the public when released, this helps, no clue why Blizzard suppresses it as much as they do. Why no transparency?

Thank you again
I don't know why this sort of information isn't included in the game guide. I would love to see full disclosure of this sort of detail. Sad that we're left trying to deduce it from the front end.
No questions about MF? I may start looking up other threads and answering questions here, I suppose.
I've got one.

Since NV5 stack guarentees a rare drop off certain mobs and bosses, does this mean that a NV5stack reduces the chance of getting a legendary?
Holy crap OP this post is fantastic and well sticky-worthy.

Well done!
sticky requested
Bump this... it's good knowledge and well thought out.
I've got one.

Since NV5 stack guarentees a rare drop off certain mobs and bosses, does this mean that a NV5stack reduces the chance of getting a legendary?

No - because you average 3.5 drops without a 5 stack, and 4.5 drops with a 5 stack, ie the 5 stack bonus is clearly an extra item that doesn't affect other rolls.

Also there is screenshot proof that the extra item can roll as legendary, its just cut off at 4 prop if it fails its legendary/6 prop, 5 prop rolls.
Gear Slot 1 and 4 at 0MF:
6 Affix: 1.6% - 1.6%
5 Affix: 3.9% - 3.9%
4 Affix: 11.1% - 11.7%
2 Affix: 83.5% - 100%

Very nice work.

Question: was this data gathered in a single act, or across multiple acts? If the latter, are you assuming these probabilities don't vary with act / monster level?
I've got one.

Since NV5 stack guarentees a rare drop off certain mobs and bosses, does this mean that a NV5stack reduces the chance of getting a legendary?

No - because you average 3.5 drops without a 5 stack, and 4.5 drops with a 5 stack, ie the 5 stack bonus is clearly an extra item that doesn't affect other rolls.

Also there is screenshot proof that the extra item can roll as legendary, its just cut off at 4 prop if it fails its legendary/6 prop, 5 prop rolls.

This is a correct answer :) I'll add it to the FAQ anyway.
Gear Slot 1 and 4 at 0MF:
6 Affix: 1.6% - 1.6%
5 Affix: 3.9% - 3.9%
4 Affix: 11.1% - 11.7%
2 Affix: 83.5% - 100%

Very nice work.

Question: was this data gathered in a single act, or across multiple acts? If the latter, are you assuming these probabilities don't vary with act / monster level?

Good question. Added to FAQ, and I created a new section regarding how the data was gathered and how the analysis was handled.
Finally, some real info on MF. Tired of seeing all the threads arguing that MF doesnt effect the quality of item. With higher MF you get a higher percentage of getting more affixes thus a better chance at a good combination of stats.

Its funny watching people complain about drops when they have no MF at all. I remember trying to farm without any MF and it doesnt even compare the quantity and quality of items I get now with running MF.
Thanks! I've expanded the FAQ a bit more. Will continue to grow this.
Added tables for various MF levels and also expanded FAQ.
It would be nice if some of this as well as others faq etc would get sticked. I mean have you seen the class forum stickies they are a joke. The current wizard information sticky and the wts ads in there while looking for anything useful make me lol.

Join the Conversation

Return to Forum