Gradual Reduction in MF Effectiveness

[LiquidDamage]

Just a quick note : a person testing MF reduction does not need to note down failed uniques/set, since for any one creature the proportion of failed uniques/sets should be constant and therefore able to be discounted.

To clarify - a pattern of unique/set drops over time should not be statistically different to the pattern of unique/set/failedunique/failedset drops.

I disagree with this, and here is why:

There are 298 weapon types and 203 armor types, plus rings, amulets, charms, and jewels make for a total of 506 item types in the game. There are approximately 350 item types (if we assume ladder play) that can be unique in the game based upon looking in the mpq files and leaving out different uniques of the same type. There are only 114 item types that can be set.

For the purpose of this argument, let us assume that "good" drops are set or unique. this means that if you run a level 95 monster with an equal chance of dropping each item (no monster like this exists, but lets say it did for simplicity's sake), 31% of all unique drops will be failed, and a whopping 87.5% of all set drops will fail.

Yes, in a really massive run count, ignoring failed items will still allow you to see a drop in "good" drops, but you will have to do a much larger number of runs when ignoring them to get the end results to be as statistically accurate as possible.

Most of those willing to test here have said that they want a testing method that takes them as little time as possible. Therefore, ignoring failed drops would not be something they would want to do.

It should be noted that if runners are going to take failed uniques and sets into account, they should NOT take failed OR successful sets and uniques that have no durability, such as bows, throwing weapons, and phase blades. You cannot count a Blood Raven's Charge as a "good" drop and a magic Matriarchal bow as a bad one when the magic bow could have been generated as either set or magic, for all you know. Doing otherwise skews your results in an unknown manner (especially if you find 10 rare Matriarchal bows off of Andariel, and count them as "bad" when they actually could all be failed Blood Raven's Charges, since she can drop the bow but not the unique).

Ok, for people who want to test this, I suggest the following:

1) do Andariel runs, since she always does her quest drop, or exploit the quest drop bug on one of the other act bosses. They will never drop anything less than rare, meaning that you know any magic item she drops is a failed set. The appropriate thing to do would be to count the number of rares vs the number of sets vs the number of uniques per run, over a long stretch of runs.

2) Do not count items without durability as "good" or "bad" no matter WHAT type they are, as you cannot tell if rares are normal rares or failed uniques. Counting the uniques, sets, and failed sets without knowing if the rares are failed or not will give you inaccurate results.

3) Use a character with teleport to get to her as quickly as possible, and kill nothing else on the way. Grab the items, dump them in town, and have a friend id them and tally the "good" and "bad" drops while you go on to the next game, to put as little time between Andy kills as possible.

4a) If the test is to find out if MF is gradually decreased when a specific monster is killed repeatedly in a small timeframe, do as many runs as you possibly can in a row. If 50 is the most reasonable you can do in one sitting, then do several (10+) sets of 50. Add up the results for each numbered run. This means if you do 10 sets of 50 runs, add up the results for run 1 over the 10 sets, run 2 over the 10 sets, etc, and check to see if the numbers decrease as the run number increases.

4b) If the test is to find out if MF is gradually decreased over a character's lifespan, then get to your goal run level as fast as possible, and then do the runs as quickly as you can for as long as you can possibly tolerate for the life of that character. Rinse and repeat for several (10+) characters, add up the items for each run number across the different characters, and check to see if the numbers decrease as in 4a.

This should give you something reasonable to go on. If say, 5 people performed the same tests and it came up with similar results, then I'd say you had something.