Curse

Yawgmoth · Aug 21, 2009, 09:15 AM // 09:15

B. it's all luck and your fuxedup perception.

If you farm specificaly for the tusks you see them in a whole different light than if you were farming for something else and getting them as typical trash drops.
To see it in work just go and farm some of the harder Nicholas items from previous weeks and try to get 2 sets of them. For example 10 Sapphire Djinn Essences. Nobody farms them now, you'll see their droprate is the same as back then when they were *heavily* overfarmed.

Targren · Aug 21, 2009, 02:37 PM // 14:37

Quote:

Originally Posted by Martin Alvito

. Conduct a controlled experiment with sufficient data to demonstrate that the hypothesis isn't true, and I'll believe you.

An interesting idea. What sort of control group would be appropriate for such an experiment, do you think? (Not a challenge, a question since you mentioned doing this stuff for a living and I've wanted to work this out for awhile).

Martin Alvito · Aug 22, 2009, 02:47 PM // 14:47

Ton of ways to do it.

The most obvious two methods have already been mentioned. Yawgmoth suggests using the Traveler as the control. Farm the stuff on Monday/Tuesday, and then go back after the Traveler moves and compare.

Alternately, you can use hours when traffic varies as the control. Farm UW around 8-9 PM Eastern. Then farm it around 3-4 AM Eastern. Be sure to do this during the week so that most of the Euros are sleeping.

Personally, I think that the temporal control is more reliable. You can't be certain that you're drawing a time when many people are farming for Traveler items, but logon patterns are more or less consistent.

Either way, iterate until you have enough data to say something substantive. Classic test is to assume that the data generation pattern is the same and just run a difference of means on the data. If the generation pattern is not the same, you'll feel it given sufficient iterations.

You won't be able to just farm up two sets of Traveler items, though. You'll need a lot of iterations to make an inference about chance, because the drop odds are pretty bad. Something like the tens of thousands of smites I popped with 55 Prot Bond would do. (Had three free weeks of time between leaving work and starting on the Ph.D.)

Investing all that time is costly, which is probably why it has not been done. Since I had a preposterous right-tail observation (if the generation pattern was random), I've never been able to be bothered. There have been other clear instances, also. Eg: Urgoz HM farm before the rest of you showed up. Gold drops went to about 2/3 to 1/2 per unit of time invested. But since I wasn't systematically tracking the ex ante drop rate (too busy farming, Wild Blow testing the unid'ds and selling), I didn't have proof I could show anyone else.

It would be difficult today to replicate that magnitude of variance in activity, though.