Guild Wars 2

Please note that my conclusions are a result of my data, which may not be representative of the underlying trend. However, based on the data I have I conclude that:

I have NO significant evidence that Magic Find has any influence on the drop rate (the probability of a monster dropping loot of any description) or drop quality (the quality of the item being dropped) for environmental monsters in The Cursed Shore or Frostgorge Sound.
I also have NO significant evidence that magic find increases the probability of receiving gems from mining or woodcutting. (Note that I only have about 100 observations here so take this with a grain of salt).

But how could this be?
Perhaps magic find applies to boss drops, or chest drops instead. Or perhaps I missed some key element of the diminishing returns system that messed with my data. Perhaps 1,109 observations aren’t enough. If anyone knows something I missed I would love to know!

METHOD
Over the course of several weeks I collected 1,109 observations while ranging from 0% to 222% magic find. These were collected mostly in the Cursed Shore (level 80 zone in Orr), but also some in Malchor’s Leap and Frostgorge Sound. I did not farm monsters for more than an hour at a time, and if I did, I would change to a different zone before doing so. While I was in a zone I would run around more or less randomly; killing whatever mobs I came across. I used the same character (a level 80 Thief) to collect data. I would not gather data from events where it was difficult to keep track of what I had killed and what had dropped, etc.

As part of my analysis, I also assigned a “drop rank” to each item, to value them in worth. (You may disagree with the order but in the grand scheme of things it would not make much difference).
But the ranking I ended up using for the graphs is (1 – lowest, 5 – highest):
0. No drop
1. Grey items
2. White items (included armour, weapons, salvage items, T1 crafting materials*)
3. Blue items (Black Lion Chests, armour, weapons, T5 crafting materials*)
4. Unidentified Dyes and T6 crafting materials*
5. Green items
I did not receive any rare (yellow) or exotic (orange) items.

The Graphs:
First off, zooming in might help. Sorry I had do bunch everything up into one picture; I can only attach one image to the forum post. :P

1. This is a pie chart of all the types of drops I received. As you can see, “Nothing” has by far the largest proportion.

2. Scatter plot of MF along the x-axis and Drop Rank (see above) on the y-axis. If magic find did increase the quality of drops, one would expect to see an increasing trend through the data. But the trend can probably be represented by a straight line: this means that there is no relationship between MF and Drop Quality.

3. A pie chart of the percentages of rare crafting materials dropping. My data suggests that there is an 18% chance of getting 1 gem and a 6% chance of getting 2 gems.

4. This is a box plot of the rare crafting material drops against magic find. The first box corresponds to getting no gems, the second box to one gem and the third box to 2 gems. Again if MF did increase the chance of getting gems, the second and third box plots would be positioned higher than the first one. But they are all at about the same height which suggests that there is no relationship.

5. This graph shows the distribution of drop ranks for groups of magic find. Along the left hand side you can see 50, 100, 150, 200 and 250. These correspond to the magic find groups (“50” includes all observations less than or equal to 50, “100” includes all observations from 51 to 100, etc). Along the row you can see the distribution of drop ranks that magic find bracket produces, where “0” is no drop, “1” is grey drops, etc. (see above). These graphs all look pretty much the same, which further establishes the idea that MF does not influence these drops.

6. This is the same as Graph 5 except it has removed the “0” category (no drops) so that you can better see the variations in quality of drop from monsters that did actually drop loot. But again, they all look pretty equal to each other.

So yeah, I hope you found it interesting!

P.S. I am happy to share my data with any other fellows out there who are doing/going to do other investigations where it would be useful!

Haha that’s a shame! Don’t worry – you’ll have more than your share of blues and greens at level 80!

How dare you use actual numbers to back up your claim! Forums are for baseless subjective arguments thown around as fact!

I dare say that it MF did, for a short while, make a massive difference. I would literally hit DR with one event and three or more rares to dust. Now? After the fix and all the whine about how “bad” DR is… I might as well farm naked. (thowing in a total exageration for effect but I’ll refrain from the whole refund, game rework, amazing idea tangent)

I’ll wear my MF set for the same reason I wait 15 minutes without looting to dust my rares… and avoid breaking mirrors, walking under ladders and crossing paths with black cats.

They also nerfed magic find too =(

“Haha that’s a shame! Don’t worry – you’ll have more than your share of blues and greens at level 80!”

I got some T6 mats from salvaging blues, greens I just sell to vendors.

So is it a fact Magic Find does nothing? In short, what’s the purpose of the stat when you stack it to 200%?

Kain Francois.4328:

So is it a fact Magic Find does nothing? In short, what’s the purpose of the stat when you stack it to 200%?

Yup, I found that MF did absolutely nothing! even when I got to 222%. That said, it might affect other things (like boss drops/chess drops) differently that environmental mob drops.

There are plenty of much more in depth tests (3,000+ kills both with and without MF) posted on Reddit that prove it does indeed have an effect.

I’d check the dates on those studies… they’ve changed MF due to DR whine at least twice in the past 2-3 weeks.

Maybe Magic Find slows down DR of loot?

ExcretusMaximus.9120:

There are plenty of much more in depth tests (3,000+ kills both with and without MF) posted on Reddit that prove it does indeed have an effect.

Thanks, I thought there would be bound to be other studies. Nice to know that there is some hope for MF! I had all but given up :P

Kain Francois.4328:

Maybe Magic Find slows down DR of loot?

That’s possible, but seems overly complicated…

i think people has misconception about magic find. when they see +200% magic find they thought the item surely will drop, it is 200% ! no it doesn’t work that way…

+200% magic find will make your magic find 300%, which means tripple the chance for an item to drop… if it has 2% chance to drop, now it becomes 6%….. if it has 15% chance to drop, now it becomes 45%….. that’s all

azizul.8469:

i think people has misconception about magic find. when they see +200% magic find they thought the item surely will drop, it is 200% ! no it doesn’t work that way…

+200% magic find will make your magic find 300%, which means tripple the chance for an item to drop… if it has 2% chance to drop, now it becomes 6%….. if it has 15% chance to drop, now it becomes 45%….. that’s all

Yeah that’s what I thought as well, so when I had 200% MF I expected drops that were say, 6% chance with no MF to increase to 12% chance. But I did not see this at all in my study. Drops that ere 6% chance with 0% MF were about 6% chance with 200% MF.

From the GW2 Wiki

“Magic find is an attribute that increases a player’s chance to receive rare loot from drops.”

So Magic Find does not improve drop rate or gathering , it only improves your chance to get a yellow item when a body can be looted . If ,lets say, the chance to get a yellow item is 1% without any magic find , and you have 222% magic find your chance goes to 2,22% (multiply 1% x 222%) which is still very low .

I love your data!

This is an amazing study! However, Karina, if what people are saying is true and it has such a low effect of 1% to 2%, then 1,000 might not be enough to actually have low enough margin of error to be able to tell the difference between the two. I mean, using a t-test, in order to have a low enough margin of error to tell between a single % requires a ridiculous amount of data points.

I don’t know. What test did you use to determine the significance of the data? My statistics are a wee bit rusty mind you, but I don’t think it’s a problem with your study persay, you just might need more data to be able to figure it out.

Of course, your data -does- show that the amount of difference that MF makes, if any, is either non-existent or relatively small though. Thanks for the interesting info.

I don’t have any data, but anecdotally, I seems to see an improvement in rare drops from an even modest MF boost from foods. However, MF from armor or items has appeared to either have zero benefit, or has actually decreased my drops, even if combined with food.

Could food MF boosts be working, but item/gear MF be broken; either non-functional or actually bugged to the point where they are offsetting any positive benefit from food?

In any event, other than food, I avoid the MF stat completely and I would never spend a dime on the cash shop MF boosters, with the value of MF appearing to be extremely questionable.

BTW, it did seem to me that the MF food had a much bigger impact during the first week or two, so something may have changed; either they “fixed a bug” that made MF “too valuable”, or they broke it intentionally or unintentionally.

I’d love to see a test using only MF boost foods, to see if there really is a disparity between food and item based MF.

I’d also love to see something official on MF, what it’s supposed to do and if it’s actually working as intended.

Magic Find affects this:
http://wiki.guildwars2.com/wiki/Crafting_material#Rare_crafting_materials
you get more of these.
(Of course since this is forums, this is an entirely baseless statement without any sort of backing with numbers. Someone is going to test it while farming destroyers and will back it with numbers for us in the near future)

“I don’t know. What test did you use to determine the significance of the data? My statistics are a wee bit rusty mind you, but I don’t think it’s a problem with your study persay, you just might need more data to be able to figure it out.”

Thanks Endless! And I actually used logistic regression to test for significance; since I had a continuous explanatory variable (magic find) and a discrete response variable (rank drop with 5 levels) I used logistic regression to find the odds of getting different drop ranks with changing levels of magic find (if that makes sense). The p-value was about 0.25 from memory.

I also tried other ways, like simple linear regression using the drop rank as a continuous variable (1-5) and testing against the hypothesis of no slope. P-value there was like 0.93 haha.

And your point about needing more data… the thing about p-values is that they only measure if there is a difference from 0 and the more data you have (as p-values are a function of how much data you have) once you get into the thousands you’re almost guaranteed to get significant p-values! This is a trick with statistics! A difference of 0.1 is different from zero. 0.00001 is different from zero also. So getting “significance” by getting more data isn’t always the answer… if that makes sense! :P

Fiontar.4695:

I’d love to see a test using only MF boost foods, to see if there really is a disparity between food and item based MF.

Haha I can do this study for you if you like! Feel free to help me collect data

Karina.9871:

“I don’t know. What test did you use to determine the significance of the data? My statistics are a wee bit rusty mind you, but I don’t think it’s a problem with your study persay, you just might need more data to be able to figure it out.”

Thanks Endless! And I actually used logistic regression to test for significance; since I had a continuous explanatory variable (magic find) and a discrete response variable (rank drop with 5 levels) I used logistic regression to find the odds of getting different drop ranks with changing levels of magic find (if that makes sense). The p-value was about 0.25 from memory.

I also tried other ways, like simple linear regression using the drop rank as a continuous variable (1-5) and testing against the hypothesis of no slope. P-value there was like 0.93 haha.

And your point about needing more data… the thing about p-values is that they only measure if there is a difference from 0 and the more data you have (as p-values are a function of how much data you have) once you get into the thousands you’re almost guaranteed to get significant p-values! This is a trick with statistics! A difference of 0.1 is different from zero. 0.00001 is different from zero also. So getting “significance” by getting more data isn’t always the answer… if that makes sense! :P

Actually it does. Like I said, I’m a little more than a wee bit rusty. :P I’m mostly familiar with the basics of t-tests and such that you use in Biology.

But don’t p-values of .25 or .9 something seem a bit…. I dunno, too different? Though I mean, depending on the type of test you can get different results. I guess the thing would be finding the perfect type of test for the data.

Actually, wait, better question: What was your null hypothesis? Was it that Magic Find did make a difference or that it didn’t?

Sorry, attempting to try and remember what I can about stats, but it’s been a few years. XD

EndlessDreamer.6780:

Karina.9871:

“I don’t know. What test did you use to determine the significance of the data? My statistics are a wee bit rusty mind you, but I don’t think it’s a problem with your study persay, you just might need more data to be able to figure it out.”

Thanks Endless! And I actually used logistic regression to test for significance; since I had a continuous explanatory variable (magic find) and a discrete response variable (rank drop with 5 levels) I used logistic regression to find the odds of getting different drop ranks with changing levels of magic find (if that makes sense). The p-value was about 0.25 from memory.

I also tried other ways, like simple linear regression using the drop rank as a continuous variable (1-5) and testing against the hypothesis of no slope. P-value there was like 0.93 haha.

And your point about needing more data… the thing about p-values is that they only measure if there is a difference from 0 and the more data you have (as p-values are a function of how much data you have) once you get into the thousands you’re almost guaranteed to get significant p-values! This is a trick with statistics! A difference of 0.1 is different from zero. 0.00001 is different from zero also. So getting “significance” by getting more data isn’t always the answer… if that makes sense! :P

Actually it does. Like I said, I’m a little more than a wee bit rusty. :P I’m mostly familiar with the basics of t-tests and such that you use in Biology.

But don’t p-values of .25 or .9 something seem a bit…. I dunno, too different? Though I mean, depending on the type of test you can get different results. I guess the thing would be finding the perfect type of test for the data.

Actually, wait, better question: What was your null hypothesis? Was it that Magic Find did make a difference or that it didn’t?

Sorry, attempting to try and remember what I can about stats, but it’s been a few years. XD

Haha no worries! Well for the two tests I mentioned (linear regression and logistic regression) they would have different null hypothesis. So the linear regression null hypothesis is that beta nought (the gradient of the line)=0.
The null hypothesis for logistic regression is that the odds of each the drops occurring (with different groups of MF) are all equal. So yah the p-values are quite different but they’re testing quite different things

Turtle Dragon.9241:

Magic Find affects this:
http://wiki.guildwars2.com/wiki/Crafting_material#Rare_crafting_materials
you get more of these.
(Of course since this is forums, this is an entirely baseless statement without any sort of backing with numbers. Someone is going to test it while farming destroyers and will back it with numbers for us in the near future)

Thing is I did measure those (it’s under T5 and T6 crafting materials) and the proportions of them I received while using 0% MF was not really any different from when I had 200% MF!

I have been using MF gear and food to clear the maps for my 100% completion, I never seen a difference in drop rates or gathering, in the 5 days it took me to clear the last of the 7 maps and towns, using MF I had 2 rares drop, strangely when doing a level 55 – 60 area in my normal gear as I was levelling up I had 3 rares drop one after the other.

Another example on 2 occasions fighting the Dragon in Frostgorge, on opening the chest whilst fully buffed to my ears with MF I received all greens and blues, oddly on another 3 occasions wearing my more serious battle gear with no MF I have received rares.

I’m aware that its not solid data, but I don’t think any number crunching and data collecting en-masse is really required to proof that MF has no significant effect on drops, if any at all. But if you believe that cracking a mirror brings you 7 yrs bad luck then carry on believing.

I respect what you’re trying to do here, although the graphs themselves aren’t particularly convincing, since high-ranked magic items constitute such a low proportion of finds. I like the design in general, I’m not sure how you applied the tests… Are you regressing numbers for single categories or multiple categories?

If I’m reading your scatter plot correctly, the total number of 4 and 5 rank items is quite low such that it really doesn’t indicate anything except that you need more samples. Right now you don’t have a good estimation of that rare/magic drop %. Are you randomly deciding which rate of MF to go with at the start of each session?

If you knew or could estimate with some confidence the normal rate of rarer or magic drops, you would be able to generate a null hypothesis for your results based on what magic find is advertised to be. If you did that, I think you would find that your current results are also in line with that null.

Hi SoggyFrog, about how I applied the tests: I left out most of the heavy stats side of things because I figure most people wouldn’t be interested, but…
I did a whole bunch of tests, I really tried to get my data to tell me that MF was working, haha. I’m not sure how much Stats you know, but:

My test on the scatter plot was just straight up linear regression.
My test on graphs 5 and 6 was an analysis for the differences in row distributions (e.g. does the 50% group have a statistically different distribution from the the other groups).
I also tried a 5-way logistic regression analysis (does the odds of getting higher ranks of loot increase when having higher magic find?)
I tried an ANOVA on graph 4 as well.

I agree that having more data is always a good thing… but something I said earlier was that, really, if you have to get 10,000 observations to get significance in your data… it’s probably too far. I know that I have relatively few observations in the higher tiers, but even if I got another 10,000 observations I expect that the distributions WILL be more or less the same and I would get the same result.

And I did not randomly decide on the MF rate, I tried to get more or less and even spread of data from 0-222%, with at least 40 or so in each 10% bracket.

But your comment about using the drop rate of rare/magic drops as the null hypothesis is a good one… Perhaps I could get lots and lots of data from mobs at 0% MF (the base line) and then from this determine the overall drop rate. THEN compare it to the drop rate I have with higher magic find and determine whether it is statistically different.
Is that what you mean? Because that sounds like a great plan to me!

If the effect is not clear over one standard play session, I think I would prefer adding other stats to my gear so I can kill more mobs in the same time (either attack to kill quicker or defense to make bigger pulls).

in my experience MF on armor does nothing
the greatest benefit comes from the guild perk and possibly food as well.
im fairly sure armor and weps do nothing noticeable to contribute

I don’t have a snazzy graph but i’m now fairly convinced MF actaully works.
I’ve been farming the Shelter/ Penitant Events for the last week with my epic maxed-out MF gear and a food boost. On avarage I get about 1 rare per event.
Today started off the same (1 rare per event). I then used a 1 hour MF boost from the Black lion trading… 3 rares per event on average.
After the hour was over I switched to another epic set with no MF, and got a ton of loot but only 1 rare in 1.5 hours.
When farming for rares and gold I found MF is important, however doing enough dps to achieve a “Hit” (% damage need to warrant drop and exp from a mob) is more important. If your attacks don’t register a hit in, you get no loot what-so-ever. Thats why grouping up is important, your party helps you do damage to mobs.

Interesting graphs. I have thought about salvaging my Magic Find set more times than once, but I think it helps somewhat.

However, another thing to consider is trying to do these tests in a party. I’m somehow convinced that being in a party improves your chances of obtaining decent loot. I know being in a party has influenced the acquisition of Badges of Honor in WvW (either that or it was just a big fluke to jump from 5~10 badges a night to more than 20 in an hour); but I have yet to do more tests in PvE content.

Karina.9871:

“I don’t know. What test did you use to determine the significance of the data? My statistics are a wee bit rusty mind you, but I don’t think it’s a problem with your study persay, you just might need more data to be able to figure it out.”

Thanks Endless! And I actually used logistic regression to test for significance; since I had a continuous explanatory variable (magic find) and a discrete response variable (rank drop with 5 levels) I used logistic regression to find the odds of getting different drop ranks with changing levels of magic find (if that makes sense). The p-value was about 0.25 from memory.

I also tried other ways, like simple linear regression using the drop rank as a continuous variable (1-5) and testing against the hypothesis of no slope. P-value there was like 0.93 haha.

And your point about needing more data… the thing about p-values is that they only measure if there is a difference from 0 and the more data you have (as p-values are a function of how much data you have) once you get into the thousands you’re almost guaranteed to get significant p-values! This is a trick with statistics! A difference of 0.1 is different from zero. 0.00001 is different from zero also. So getting “significance” by getting more data isn’t always the answer… if that makes sense! :P

Was the p-value from the effects test (whole model) or just individual categories of drops? Also, would you mind making your raw data available?

Great job by the way!

Tangleknot.8605:

I don’t have a snazzy graph but i’m now fairly convinced MF actaully works.
I’ve been farming the Shelter/ Penitant Events for the last week with my epic maxed-out MF gear and a food boost. On avarage I get about 1 rare per event.
Today started off the same (1 rare per event). I then used a 1 hour MF boost from the Black lion trading… 3 rares per event on average.
After the hour was over I switched to another epic set with no MF, and got a ton of loot but only 1 rare in 1.5 hours.
When farming for rares and gold I found MF is important, however doing enough dps to achieve a “Hit” (% damage need to warrant drop and exp from a mob) is more important. If your attacks don’t register a hit in, you get no loot what-so-ever. Thats why grouping up is important, your party helps you do damage to mobs.

That’s what I call lucky

MF probably helps, but if you got 200%+ (which isn’t real 200% since you can’t add up % like that) that’s still 1.x multiplier to the actualy drop % of a certain item.

Let’s say corrupted lodestone got 0.5%, so now you got 0.8/9%.

The rates are there, but at a price of dps.
And I – for one, cannot stand slow killing.

I have my doubts that a regression would yield any kind of result, especially if you’re considering the drop rate of common and rare things in the same test; the former is going to be about the same, and that will just obscure any sort of signal that you expect to find.

I do have to say it’s not a very sound statistical practice if you find yourself doing a bunch of different tests. It’s always better if you are able to decide beforehand what kind of tests you wanted to do, and then design your experiment so that it’s best able to answer that question. Then you run whatever tests you have planned and make your conclusion. This experiment is still useful as a trial though, since it gives you an idea of what kind of numbers you’ll obtain. That will help a lot for future designs, if you’re still planning more experiments.

My suggestion would be to start with a simpler test: compare 0% to 222%. I think it’s fair to use a single kill as a single sample, but it is hard to control what kind of mobs you hit/the proportion of damage you deal especially given that you’re not blinded to what kind of gear you’re running, so keep that in mind. As well, perhaps you could focus on an easier metric of comparison; rather than considering all the ranks of items separately, lump the high tiers together and the low tiers together and just compare the proportions.

Regarding my question about randomness, I did assume that you had the same or a similar number of samples for each MF ; what I’m asking is, were the trial runs conducted in a random order (so that you don’t have all 0 MF and then 222% in blocks)?

Gosh, so many technical questions guys! I’ll try to answer as best I can though xD

@Tangleknot and Ari Kagura: yes there are so many other things to explore here! My data was just about environmental monsters standing around in fields, but I do have a hunch that drop rates might be different for parties/event mobs/dungeon monsters. But like they say, it’s only a good stats test if you’re left with more questions than answers…

@Lunesta: I think you’re referring to the first p-value? Well, I think it’s a bit of both. This p-value measures the test for no relationship between the five groups of magic find (which test that the Odds Ratio (which is the likelihood of that event happening to not happening) between groups =1). Therefore since my p-value>0.05 there is no evidence that the having higher magic find is related to me getting different loot in the 5 categories.
And I can certainly make my raw data available! Though I’m not sure what the easiest way to do that is…

@SoggyFrog: you sound like my Statistics lecturer: “GET YOUR QUESTIONS FIRST” :-P which is completely true. It’s not the best practise… my original questions were “what is the optimal MF rate to get the best loot?” and when I saw that it was so terribly useless I scrounged around trying to discover something significant. Which I know is not the best statistical practise – my apologies!
And yes your suggestion is actually EXACTLY what I’m taking on board right now, I’m in the middle of collecting 0% MF data.
And to answer your question about randomness: I tended to start with a level of MF at random (say 80%) and then during the course of my killing I would add or remove MF gear to get a spread say from 60-90%. Then another day I might start at 150% and take a spread around there, and so on. I did take most of my 170%+ data all in one block though I have to admit – it took the most resources and I tried to get it over and done with relatively quickly.

That is how it is with experiments, isn’t it? You start out thinking you’re going to do one thing, and you often end up changing the plan quite a bit. But there’s always something you can salvage.

Apologies if I do sound too much like your lecturer (I guess they’re doing things correctly then!), you seemed earnest enough to take a sincere criticism well.

What I was saying about the order of collection is that you should either alternate or randomly “plan” when you’ll use MF and when you won’t. That’ll help even out some of those factors outside your control, especially since you’re kind of your own test subject here.

i got rares from drops even without magic find. it was a level 50s item though.

Magic find improves the drop quality, not rate. With around 150% in myself, i find when farming orr my loots changing from lots of whites to have a minimum of blue. Also, before MF i almost never found a yellow by killing mobs. With the MF each event in orr generally gives me one.

That’s my experience though.

SoggyFrog.4170:

That is how it is with experiments, isn’t it? You start out thinking you’re going to do one thing, and you often end up changing the plan quite a bit. But there’s always something you can salvage.

Apologies if I do sound too much like your lecturer (I guess they’re doing things correctly then!), you seemed earnest enough to take a sincere criticism well.

What I was saying about the order of collection is that you should either alternate or randomly “plan” when you’ll use MF and when you won’t. That’ll help even out some of those factors outside your control, especially since you’re kind of your own test subject here.

Yes that’s exactly how these things tend to go! It’s never a dull moment with Statistics And no worries about the lecturer thing! I meant it in good spirits. Constructive criticism is most welcome!
And I will take that on board, some times I will use 0% MF and sometimes I will use 222%, but not all in one big block.

By the way if you’d like to help me collect data you’re welcome to! :P

Sirevanac.3178:

Magic find improves the drop quality, not rate. With around 150% in myself, i find when farming orr my loots changing from lots of whites to have a minimum of blue. Also, before MF i almost never found a yellow by killing mobs. With the MF each event in orr generally gives me one.

That’s my experience though.

Thing is, my data seemed to show that MF affected neither! (you can see by graph 6 I think it is). But if you want to challenge it, feel free to collect some data and compare

You might just be unlucky or Magic Find could be kittenched.

It works for me, it works wonders to be honest with you. Just yesterday I consumed a Lemon Bar which gives you + 16% magic find I believe. Found 2 dyes, tons more gems, a couple of great rare weapons and so on and so forth. For me, It is something that works perfectly, I recieve more quality weapons and armor with more magic find and find more dyes and gems which is just awesome.

ErkiB.8375:

You might just be unlucky or Magic Find could be kittenched.

It works for me, it works wonders to be honest with you. Just yesterday I consumed a Lemon Bar which gives you + 16% magic find I believe. Found 2 dyes, tons more gems, a couple of great rare weapons and so on and so forth. For me, It is something that works perfectly, I recieve more quality weapons and armor with more magic find and find more dyes and gems which is just awesome.

Well perhaps it works much better for some people than others xD
Or I wonder if it’s a mental thing. Like, “OOH I got a dye/rare! It must be from the magic find bar I ate 20 minutes ago!” Although when you don’t have MF you don’t take as much notice, so relatively speaking having MF makes you think you’re getting better loot?
Just a theory I have :P

This is really interesting, and thank you for taking the time and effort to empirically test Magic Find. This confirms everything I’ve ever experienced from Magic Find – people who kit their gear out in MF and then receive a drop see the MF as the reason for that drop; those who receive it out of the blue see it as simple luck.

Magic Find is the Emperor’s New Clothes.

First of all, thanks a lot for your tremendous effort. The problem with really good drops is probably that they are rare events in a statistical sense and you would probably need a very high number of samples to reach a satisfying conclusion.

Hey Karina I really like the thread.

Since your experiment has changed to finding the optimal mf, to wanting to see if magic find works at all or not, would it not be much simpler and much faster with greater accuracy to pick 2 stat points and see if there is any correlation at all (ie 0% and 220%) (sorry if I am wrong, I only have a moderate understanding of stats from my engineering courses).

In fact, we could even simultaneously test whether food does work and gear doesn’t, or vise verse, by picking a stat point that is easily obtained by either method, say 30% mf (which is 5 accessories or the omniberry bar). You could easily have the community help with collection by posting images of excel files with their findings in them (or even make a public google doc that people can add their findings to, though you run the risk of sabotage by randoms on the internet… but at least google docs let you see individual changes by user and undo them. You could also make a group and put up files as read only, still dose not help against false data but meh).

Im sure you probably could get enough community help to put together a decent survey and do follow up experiments depending on the results.

EDIT: with enough data points you would also be able to see correlation with absurdly low drop rates as sometimes even a couple thousand data points is not good enough to measure these (and rares seem to be in this area, not to mention exotics not even sure if they drop)

What is needed is a way to find your hidden luck number which is generated at character creation and is used in the formula that determines what item from which loot type drops for each character from kills, chests, bosses, etc.

Karina.9871:

ErkiB.8375:

You might just be unlucky or Magic Find could be kittenched.

It works for me, it works wonders to be honest with you. Just yesterday I consumed a Lemon Bar which gives you + 16% magic find I believe. Found 2 dyes, tons more gems, a couple of great rare weapons and so on and so forth. For me, It is something that works perfectly, I recieve more quality weapons and armor with more magic find and find more dyes and gems which is just awesome.

Well perhaps it works much better for some people than others xD
Or I wonder if it’s a mental thing. Like, “OOH I got a dye/rare! It must be from the magic find bar I ate 20 minutes ago!” Although when you don’t have MF you don’t take as much notice, so relatively speaking having MF makes you think you’re getting better loot?
Just a theory I have :P

Funny how my one post in this thread was deleted…

Anyway, what I wanted to say is that I think your theory is right on If magic find indeed has an effect it’s so small that it’s not worth kittening yourself over. I do however eat Omnomberry bars for the extra % coin.

I run the event chains in Cursed Shore without any magic find stats and i get plenty of rare drops, at least 3-4 rares each session, and i figure that at the price ectos are going for, most people aren´t having much trouble with drops as well.

If magic find doesn’t do anything then that’s a very good thing! The best thing to do would be to remove it, entirely, for which see here:

https://forum-en.gw2archive.eu/forum/game/gw2/Why-Magic-Find-is-a-Poor-Mechanism/first

jweltsch.1832:

Hey Karina I really like the thread.

Since your experiment has changed to finding the optimal mf, to wanting to see if magic find works at all or not, would it not be much simpler and much faster with greater accuracy to pick 2 stat points and see if there is any correlation at all (ie 0% and 220%) (sorry if I am wrong, I only have a moderate understanding of stats from my engineering courses).

In fact, we could even simultaneously test whether food does work and gear doesn’t, or vise verse, by picking a stat point that is easily obtained by either method, say 30% mf (which is 5 accessories or the omniberry bar). You could easily have the community help with collection by posting images of excel files with their findings in them (or even make a public google doc that people can add their findings to, though you run the risk of sabotage by randoms on the internet… but at least google docs let you see individual changes by user and undo them. You could also make a group and put up files as read only, still dose not help against false data but meh).

Im sure you probably could get enough community help to put together a decent survey and do follow up experiments depending on the results.

EDIT: with enough data points you would also be able to see correlation with absurdly low drop rates as sometimes even a couple thousand data points is not good enough to measure these (and rares seem to be in this area, not to mention exotics not even sure if they drop)

Thanks Jweltch!
And yes that is exactly what I’m working on right now: collecting data just from 0% and comparing it to 222%. I’m hoping to get ALOT of data so your idea about community involvement is an appealing one – I have to do less work, haha. I have contacted other people who have done surveys like mine and got their data, but I thought people out in the world would not be overly interested in collecting for me to be honest… But I guess it’s worth a shot!

I created a google doc:
https://docs.google.com/folder/d/0B-q8UK6yZeLRNEtGMkxUUDFrWlU/edit
It has my data in it for those who are interested (feel free to use it!) and if you have some data/want to help collect data, post it up there! (I hope it works).

Data at 0% MF or at >200% MF is particularly useful. It does not have to be in the same format as mine with the same categories and everything. Put it up with as much detail as you want to, and I can make sense of all of that later.
Thank you!

Just a little semi-off-topic question:

Is there any way to read your total amount of effective magic find in game, besides making the addition yourself?

MF has always been a long term investment in just about every game that has it. A sample size of 1000 is hardly enough to see a noticeable change in revenue. If you were to run as maxed as possible, gaining 1 exotic for 4 hours worth of play time as opposed to half (say 1 every 8hrs.) that would seem to be a more likely way to gauge MF and that’s basically what i’m seeing for the most part. We are talking very small percentages here, doubling or even tripling those numbers only adds up over time. Playing diversely in this game is where it’s at, not sticking to an area running the same things over and over.