Curse

FoxBat · Apr 18, 2008, 03:20 AM // 03:20

I've heard rumors about this here and there, so I decided to do some controlled tests. Results are far from conclusive, but the initial findings are disturbing.

This thread will mostly describe farming attempts, but the conclusions drawn from it affect all PvE, so I placed the post here. If mods feel it's better off in farming then w/e.

I took the same mesmer character and did some Minotaur runs in Elona Reach normal mode. I use two very different builds, a Sandstorm one that can quickly level a large group of foes at once with earth magic, and a Signet of Illusions build that slowly whittles enemies down one at a time with degen and empathy. These builds have demonstrable differences when solo in terms of the number of drops, I've tested and explained this elsewhere, it is the core of lootscaling that making kills close in time to each other significantly reduces the amount of drops received. You can try this for yourself if not convinced, after just a few runs the results are very noticable.

Anyway. What I was curious about though, was what would happen if I had a party of henchies with me in radar range. In theory, loot scaling is only supposed to effect full parties. Therefore, one's killing speed with henchies in range of drops should not affect the drops you recieve. However, data from 20 runs with each build seems to suggest otherwise.

I took the first 5 henchies from the party window, and enter Elona Reach NM. I flagged the henchies right inside the starting gate, aggroed the two Minos on the hill, and pulled them down to the area beyond the hill where tons of minos congregate. Henchies were within radar/not grey on party screen each time. (Drops are less than you'd expect for true solo anyway).

After killing the minos, I screenshotted the drops, counted the number of lootscaled items and gold coin drops. (One Dye dropped for me during an SoI run, presumably dyes are not affected by LS so I therefore discounted it from this list.)

I interspersed the run of each build with each other, to control for any supposed "repeated entry anti-farm code" that some claim is still around. Each time I exited the mission by /resign. The drop instances are listed in temporal order within build, and it doesn't seem to show any repeated entry effects anyway.

Edit: added more data, 10 runs each. On day 2 I stuck the henchies up on the hill, in case them not being "close enough" was somehow messing things up.

For both builds I had +3 fast casting, +1 illusion, +1 inspiration, and a 19% enchanting staff.

[build=Signet of Illusions;OQZDAWwzOogtGOwAIU3qfVNA]

This build basically tanks all day using Signet of Illusions to power the earth skills and empathy, and slowly kills one target minotaur at a time.

Code:

Day 1

Drops		Gold

0		7
2		2
1		3
2		4
2		4
0		3
1		1
1		5
1		3
2		5

Avg Drop	Avg Gold

1.2		3.7


Day 2

Drops		Gold

0		5
1		3
2		3
1		6
0		5
0		3
1		3
1		4
1		5
1		2

Avg Drop	Avg Gold

0.8		3.9

[build=Sandstorm;OQZDA5kJPLBuqqwVYM2qfVDA]

With this build I would put up GoEP, then stoneflesh followed by stonestriker, and wait until I could do that again as the minotaurs slowly gathered around. Then I would arcane echo, sandstorm, sandstorm, wait slightly, earthquake, aftershock. Often this would leave minos running away at a sliver, then I'd wait for them to return and finish them all at once with aftershock.

Code:

Drops		Gold

1		1
0		2
0		2
0		1
1		0
0		1
0		2
0		2
0		1
0		2

Avg Drop	Avg Gold

0.2		1.4

Day2

Drops		Gold

1		0
1		1
0		2
0		1
1		1
0		2
0		3
2		0
2		2
2		1

Avg Drop	Avg Gold

0.9		1.3

Total Averages from both days:

Sandstorm: 0.55 drop, 1.35 gold piles per run

Signet of Illusions: 1 drop, 3.8 gold piles per run

I'm going to admit the drop data is a little questionable with only 0, 1, or 2 as possibilites. But the difference in gold drops is sizeable and very consistent, with almost three times as many gold piles in a slow kill. I'll have to find a different spot to test the item drops more thoroughly.

What disturbs me at this point though: that there is any significant difference based on killing speed for full parties. I thought the "elite farmer filtering code" was only intended to affect solo farmers, but these results strongly suggest that this simply isn't true. I hope this is just a long-seated bug that can be brought to light and fix rather than an intentional change; I don't think making AoE elemenalists scorned for their drop-robbing vs. more staggered killers like yelloway/ursan is "good for the game".

What should be further tested is whether these results still hold true with a full party of human players; I don't really have the resources to test that, so it will fall to someone else. It's possible this only applies to hero/hench for whatever reason, although that is still bad IMO and should be fixed. It could also be that this only works correctly in "8-man" zones, but I think the drops wouldn't differ so widely if that was the culprit.

I hope some others will try to corroborate this data with their own tests, although likely this thread will just be filled with speculation and anectdotes as usual.

Coloneh · Apr 18, 2008, 04:06 AM // 04:06

Quote:

Originally Posted by FoxBat

data from 10 runs with each build seems to suggest otherwise.

useless data. come back once youve done 100 runs each.

FoxBat · Apr 18, 2008, 04:40 AM // 04:40

Quote:

Originally Posted by Coloneh

useless data. come back once youve done 100 runs each.

100 runs would be good for approaching accurate % estimate; it isn't needed to show that there is any difference when averages vary as widely as 300%, and the results are normally distributed assuming your average random number generator. If only I could remember how to do those P-tests... the mino population is also 15-20 per run, it's not as if I'm turning up a random number 10 times but more like 150-200.

I'll do another set of 20-20 runs or so, (p=30 was good in psychology with much, much smaller differences), but I'm not a serious farmer, (or a physicist) and more want to get other people investigating this in other ways.

cellardweller · Apr 18, 2008, 06:09 AM // 06:09

You're after a two sample pooled t-test, not a P-Test. I'm too lazy and my brain is too cobwebbed to do it for you, but my eyeball says you've got enough data to get a 95% CI.

FoxBat · Apr 22, 2008, 04:20 AM // 04:20

In my search to find easier drop spots, I went looking at some low-level monsters, such as the termites outside the Astralarium. The drop quality wouldn't be amazing, but I'm just interested in quantity for collecting data. Also in 4-man areas I can mess around more with hero or human configurations to further determine how loot scaling is interacting with party size.

What I found surprisingly was that no matter how quickly or slowly I killed the little level 2 buggers, I seemed to be getting roughly the same kind of drops. So I suspect that level is somehow a part of the equation, that low-leveled monsters are either exempt or subject to much less scaling. It may be relative to the players level (which would be especially bad for lvl 28 monsters), it may smoothly scale or there could be a cut-off point, I'm not entirely sure.

As a preliminary investigation (and to establish some baseline info for later tests with party members), I took a 105 monk to the gargoyles outside nolani academy, which are around level 7-8 in normal and 22-23 in hard mode. I walked past the group on the first stairs you run into, and up the stairs all the way into the clearing of empty buildings, where about 4 groups of gargoyles await. Usually the group at the bottom of the stairs would lose aggro as I gathered these other four groups.

[Mystic 105;Owok4sPIUuijBpesSurqX2LAReA]

Although the kill here is not as clean as you can do with SS, enemies still tend to die fairly close in time to each other to SoJ. The knockdown throttles their attack rate so the ele and mesmer gargoyles die about the same time, wheras trails with VwK left the eles alive long after the mesmers had killed themselves. In hard mode these guys sport lightning surge, necessitating the anti-knockdown enchantment. Since the Mystic Regen nerf this build ends up with only 1 net pip of regen after hexes, so it basically relies on keeping SH or SoA on constantly for 0 damage.

I went and did 8 runs in normal and hard mode, alternating, completely solo. Without henchies to steal drops, the data is (hopefully) more consistent.

Code:

 Normal

Skulls	Gold	Other Drops
		
3	6	3
4	1	6
5	1	7
2	2	7
4	5	5
3	4	4
3	1	7
1	4	4

3.125	3	5.375	Average

Hard

Skulls	Gold	Other Drops

0	2	3
1	2	3
1	2	1
1	1	1
1	1	2
0	1	2
2	1	3
2	1	1

1	1.375	2	Average

Sample Normal Run:

Sample Hard Run:

I seperated out the singed gargoyle skulls because I had a thought that it may be item value which is scaled rather than monster level, IE loot scaling would only apply to more expensive dropped items (20 gold+). But the results here suggest that these are just as subject to scaling as the near-max whites and larger crafting material piles in Hard Mode.

Again we are seeing roughly a 1:3 ratio in terms of loot scaling as with the minotaurs. If correct, that would imply that loot scaling is hardly enforced at all on the lvl 7-8 monsters, but completely enforced on level 22-23.

To completely eliminate alternative hypthesis (such as HM drops themselves are just rarer) I will need to find a build that lets me solo-kill the gargoyles in relative safety (this build is pretty difficult for me in HM, I've had many botched attempts, miss one enchantment in that shifting hex status stack and it's over quick.) Also I'd need to find a high-level place in Normal Mode and compare to Hard Mode drops. But regardless this does at least confirm the conventional wisdom that drops for easy-area collectible farming (for festivals) are better in normal than hard mode.

Tarkin · Apr 22, 2008, 07:52 AM // 07:52

...
A long time ago i and lots of others ppls kown, if you kill tons of monsters to fast the trash drops and money dont drop... solo or in a full party... and i dont make any research just look the drops.
...

FoxBat · Apr 22, 2008, 05:41 PM // 17:41

Well good for you all, but alot of people don't seem to know this and need convincing (I was one of them), so here it is. You have seen how long some the loot scaling threads are... it's evident many aren't aware that this hits everyone and not just solo farmers, or to what degree they are affected.

Then there's people who think this is some kind of "bug" with drops getting stuck in AoE blasts, although I think it's an intentional and formulaic change, and only some more testing is going to prove anything there. (IE if loot scaling barely applies in low-level areas despite AoE... it's unlikely to be a simple bug.)

Did some normal mode runs to compare loot scaling of level 8 gargoyles with a ranger. One AoE build and one single-target build.

[Trapper;OgATcT8mVS2hLM5Q5o4I1gVy3AA]

With this build I pulled the four groups together into a doorway before unleashing traps and killing nearlly all of them in a shot.

Code:

Skulls	Gold	Other Drops

3	2	6
3	2	8
5	4	6
3	4	4
4	4	4
2	2	7
5	3	6
2	3	8

3.375	3	5.375		Average

[Screaming Shot;OgATcT8m1Z4Mbu1PmiTqhV6bAA]

With this build I had to pinpoint and take out the Ressurect gargoyles quickly or be there all day. Then I took out the boss, as it was sometimes a draw conditions monk. Finally I took each gargoyle down one at a time, with only apply and a single screaming shot. This throttled my kill speed down to 8-10s per kill.

Code:

Skulls	Gold	Other Drops

2	8	8
4	5	9
9	6	1
6	7	7
7	4	11
2	7	9
3	7	9
2	6	8

4	6.25	7.75		Average

So there does appear to be some loot scaling going on here... even if less extreme than what is expected with minotaurs or hard mode. It's basically 12 expected drops for the AoE build (same as normal mode SoJ) vs 18 for taking 10 seconds per monster.

Edit: and now some Normal Mode runs on lvl 2 Termites outside Astralarium....

[AoE Mesmer;OQZDAWwzSrAIglAGACVbVfVdeA]

Aggro is gathered by tabbing and wanding each termite. Then one sigged Ether Nightmare followed by a well-timed Wandering Eye kills everything but the Iron Drake, which is finished off by Clumsiness. "Trophy" includes both Termite Legs and Drake Flesh.

Code:

Trophy	Gold	Other

0	1	2
1	3	3
0	3	3
2	1	2
4	2	1
1	1	2
1	2	1
2	3	0
3	2	4
4	1	2

1.8	1.9	3.9	Average

[Empathy mesmer;OQZDAWwzOoAAQJwAIU3qfVNA]

This build uses a Sigged Empathy to kill termites one at a time. I skipped using a staff to throttle Empathy kills to roughly 10-11 seconds, similar to the ranger throttling.

Code:

Trophy	Gold	Other

4	2	3
1	3	5
2	3	0
6	2	0
2	3	2
2	2	4
2	3	2
4	2	1
3	2	2
4	2	2

3	2.4	2.1	Average

The total drops average out to 7.6 vs 7.5. There may be a difference in there that could be established with more data, but it's rather small. Further support for Level affecting loot scaling... and evidence against the AoE drops being some kind of game logic bug.

Edit Again: HM Gargoyle Results

[Empathy SoA;OwUUI0h9YoNaAlvMIBUQurAE5BA]

A very slow, but safe and controlled build. Usually takes 2-3 empathy casts to kill a foe. Power lock is to nail resurrect when there are two rez gargoyles, and wastrel's just plain punishes power locked opponents.

Code:

Skull	Gold	Other Drops

6	4	7
4	6	8
3	6	6
6	4	10
7	6	9
6	2	10
5	8	3
7	6	3

5.5	5.25	7	Average

Average total drops is 17.75 for a single-target HM build.... compared to 18 for single-target NM (screaming shot). It looks like virtually no loot scaling is taking place in those instances. The differences observed in hard mode mass-kill are again likely because of loot scaling affecting higher level monsters more than lower level ones.

Now all the baseline work is done, and I'll being comparing these runs with different player/hero/hench configurations next.

Skye Marin · Jun 28, 2008, 06:14 AM // 06:14

Pretty interesting stuff.

So you get more drops if you kill a number of monsters over time, and the number of drops are throttled through loot scaling only if they have higher values (say, merch value of 50g or more).

Now, combine that with what we know about monster-drop assignments in this thread:

http://www.guildwarsguru.com/forum/s...php?t=10225077

That basically says that what drops from what monsters is defined when you enter a zone. This means that I do the raptor farm over and over, and the lever is already pulled when I enter the instance on if I'm going to get a gold or not, even if I'm solo. Also, gold weapons are declared to be exempt from loot scaling, which I would assume is actually all forms of loot scaling, being both party and kill-rate. I don't have evidence to support that, but I'd assume that in your runs, you didn't encounter many gold weapon drops (and to be able to plot a trend regarding gold drops would need a very large sample size, and a larger mob pull would help too).

My theory is that it is still always best to run a fast, simple run and kill as many monsters as you can, as fast as you can, because the gold drops and rare materials will always be there to collect. Anet has designed the system so that the best, fastest way to make money is to sell to other players, directly (7 unids for 5k) or indirectly (material traders).

Moral55 · Jun 29, 2008, 01:47 AM // 01:47

Could you do a drop per second analysis? It's all well and dandy that we know you get more drops if you go slower, but if you can kill stuff 4x+ as fast and get 1/3 loot it might still be better to go quickly.

I.E 4 * 1/3 = 1 1/3