Guild Wars 2

Earlier today I decided to try my hand at answering some of the more common questions people have about farming with your friend and mine: mathematics. I decided to approach the problem as a casual gamer adventuring in high level zones on the weekend, but who would like to be properly equipped, and hunting the proper enemies, so they will make as much money as they can given their time constraints. I constructed a sample by killing a variety of level 80 enemies in the cursed shore, every type of enemy several times. To smooth out my magic find into a more continuous variable, I made use of a superior sigil of luck. Unfortunately, no such methods were available for gold find. Basically, I set my base magic find to some value using food and gear, run from waypoint to waypoint killing things until I hit 25 stacks, then relog to reset the counter and start over on a different path. I collected 485 observations of drops, collecting enemy name, enemy type, drop, drop quality (crafting rare, magic item), vendor value, and coin drop amount in copper while taking care to note my magic and gold find at the killing blow. Once I did this, I performed some statistics magic, mostly linear and logistic regressions, my results follow.

Gold Farming

The basic formula for calculating the average vendor and coin value of a kill, if you vendor everything that drops, is:

Copper/kill = 13.4*** + 0.043 x MagicFind + 0.37 x GoldFind**

With both find values in whole numbers. Gold find is the clear winner here, as the 40% gold find you get from eating an Omnomberry Bar will roughly double your expected yield. It is worth mentioning that the magic find value here is not very significant, neither statistically nor otherwise. At my average magic find of 33%, about the same as gained by consuming an Omnomberry Bar, I would only expect to see, on average, an extra copper per kill (if any at all). Another interesting result is that construct type enemies, like Orrian Spectral Guards, render on average an additional 19 copper per kill. All other enemy types are roughly comparable to one another, with the exception of animals for which you can expect around 5 copper less.

If magic find works intuitively, that is to say that it is some base rate multiplied by 1+0.MF, I can work backwards from my average drop rate of 0.031, divide it by 1.33, and arrive at the average magic item drop rate of trash mobs in Cursed Shore of 2.3%, or approximately 1 drop per 43 kills. Even with the substantial magic find boost of relatively expensive food, it would only increase to about 1 drop per 32 kills. Simply put, for the casual gamer who isn’t taking down difficult bosses for somewhat better loot odds, magic find simply isn’t worth the slot it takes up.
I should probably also mention that I can find nothing to suggest that magic find actually increases the drop rate of magical items at all, and 485 is not an insubstantial sample size.

Rare Crafting Mats

The best guidance I can give here is which monster types offered the best and worst drop rates in the bunch. Oozes drop rare crafting materials about 30% * of the time, which is the highest drop rate, whereas animals drop rare crafting mats about 15%* * * of the time, which is the worst. Humanoid enemies come in somewhere in the middle, with drops rates of around 22% * * .

Summary

Be nice to wildlife, eat your Omnomberries, and good hunting.

• * denotes significance at 90%, * * denotes significance at 95%, * * * denotes significance at 99%

Correction:

A fellow named Nhalx in another forum, while talking about magic find, suggested that higher magic find leads to an increase in any kind of item being dropped. In order to test this hypothesis, I generated a variable representing if a drop occurred, then ran a simple logit model with that variable on the left, and magic find on the right.

Turns out, it does increase the odds of getting an item drop of any kind. Here are some sample values.
+0 MF: 34.9% drop chance
+10 MF: 37.2% drop chance
+20 MF: 39.6% drop chance
+30 MF: 42% drop chance
+40 MF: 44.5% drop chance
+50 MF: 47% drop chance
+60 MF: 49.5% drop chance

I stand corrected. Oddly enough, magic find still does not appear to increase the odds of getting crafting rares or magical items, leading me to suspect that the thing it really increases are drops of greys and whites over no drop, with perhaps lesser gains to more substantial drops.

Holy crap man, that’s awesome. Seems like you put a lot of time into this, and as someone who is definitely a casual gamer let me be the first to say: thanks! This is really good to know!

Math is awesome. Thanks. I’ve been trying to suss out the effect of MF too but hadn’t summoned the patience to start documenting in this detail. Do you plan to try again an a larger sample?

Unkemptwolf:
No problem. Time wasn’t really an issue, it only took about 10 hours from start to finish. The difficult part was if I pulled 5 mobs simultaneously I would have to make sure that I kill only one at a time, and remember the exact sequence in which they died. Also, typing in excel during combat should be its own video game.

TabMorte:
I hate to say it, but the question of drawing a larger sample, at this point, is really more of a philosophical question than a statistical question. If the standard deviation for magic find in my logistic regression is reasonably accurate, I would need 1680 more observations to accurately determine how magic find affects the drop rate of magical items.
It may be an unscientific view of the matter, but I think if I need to go out and kill almost 1,700 more enemies to see if magic find is doing anything for me, I think that question answers itself.
I would, however, be interested in exploring other problems.

Thanks for actually testing this out with good sample sizes. This is the kind of stuff I was wanting to see from people testing this, not a “looks like this, but never really paid much attention” type of answer. If your wanting to try and make better sample sizes for this to get better data, send me a PM, would be happy to lend you a hand.

Excellent post. Gets my vote for forum post of the year.

Thanks for actually testing this.

Even if this sample size “is not big enough,” every test I have seen has said that MF does not make that big of a difference. And every test has people saying sample size not large enough. If that is the case…shouldn’t some of the tests say that it does make a difference? Instead of them all saying it does not?

Ok tc can you explain this in “DUH” terms now please

hahawinner:
Be careful what you wish for. I have another idea to determine if magic find increases item quality, for which I will need a much larger sample.

Yukishiro:
Thank you.

Alle:
You’re welcome.

Dinkus:
Yes, we would expect that. I don’t know how many people have actually been recording observations and analyzing their results, but my guess is that most of them are just going off their impressions. If the effect is sufficiently small, it would be invisible to the casual observer, so the general impression would be one of no significance. Here, we’ve got that it increases item drop rate overall, but not the average value of drops. That’s something, kinda.

karma:
Gold find is good, magic find is meh, kill oozes for rare crafting mats.