Why Witch loot drops suffer so badly - the stats:

Cronk, your argument for the lack of witch drops is based on stats that are obtained from your test sample. Your argument is predicated on the stats, so people who are addressing the sample size are addressing the topic, just not in the way you intended. Their argument isn't that you are wrong, as they haven't proved that. Rather, they are saying that you haven't sufficiently proven your own position.

There are a number of posts that go into some detail as to the reasoning behind their claim that the sample size is insufficient, but you haven't said anything beyond 'it's fine'. In particular, DestroTheGod made a good post referring to the effects of item levels that would drop within your designated test zones and the limited testing in regards to player/area levels.

I've never been very good at statistics, so I'm not going to say whether or not your sample is big enough or not, as I simply don't know. But I did a lot of study on logic, and your dismissal of the arguments presented weakens your own. I recognise the amount of work you put into acquiring the sample and analysing it to the extent you have, so I would be keen to see you defend it adequately against arguments that do question it's validity.

"

sabledrakon wrote:

Exactly, except Path's rolls don't suffer from roll method bias or physical defect bias. Which means more samples start to show a more complete model of the function.

To be clear, this response is perfectly contributive. But for anyone else: Enough with the "Do more!", *contributes zero*.

Based on the data you provided, you expect 10% of the drops to be Int based. You received 8.5%. That seems perfectly within reasonable bounds given your sample size. All you have shown is that the drops in your sample seem to be fairly representative of the parent population.

And as far as that parent population go, that's likely just down to the particular level band you chose.

"

Cronk wrote:

So tell me, do you disagree with my analysis of the weighted currency drops from a sample of 433?

Not entirely. I can't disagree or agree with raw data I haven't seen or collected first hand. But I will say that it's a good start to begin mapping the function. While incomplete, generic shape is starting to come to light. It's just a matter of feeding more datapoints into the computation.

Since Deceptionist threw my post out the window and again dismissed people as trolls, I would like to state that what Malladin said about wanting to see OP defend his foundings in a matter that would want other people to discuss the subject, is pretty much my point.

If he feels that he have a good enough sample, dont tell people that they are trolls for arguing against it..

Well, I gave you a hard time for not having any data, and this time you have come back with data. This is good, and the data is interesting to look at.

But already you start with a strange premise, that "RNG is not actually RNG", and further continue on this line with your description of currency drops:

"

Cronk wrote:

The currency system is NOT pure RNG. The currency system is 'weighted' RNG whereby the lower the value of the currency, the more probability/chance there is of it dropping. It would NOT be possible for all 50 of those drops to be 50 Exalted Orbs. This is because the RNG is functioning under a set of rules which produces the following 'pattern' that I got from the four screens:

The idea that weighted drop rates is not pure RNG reflects a misunderstanding of what an RNG is and how it works. It is simply an algorithm by which you draw a random number. It does not mean that the results cannot be weighted.

Consider the following analogy. Suppose you are hosting an event for enthusiasts of model trains, and you wish to hold a drawing to determine who will win a very valuable, one-of-a-kind model train. You allow people to buy tickets to support the event, with each ticket costing $2. A person who wishes to win (or simply wishes to support the event more than others) may buy multiple tickets.

So the average eventgoer might buy one ticket, whereas another man may buy 200 tickets, if he is willing to spend $400 on it.

After the ticket purchasing is complete, you put all of the tickets in a large box, shake it very thoroughly, and then reach your hand in and pull one out.

Now, in this example, a few things are obvious:
1) The probability of each person winning is weighted, since a person may purchase multiple tickets.
2) The process of drawing a ticket from the box is random.
3) A person who only purchases one ticket is far less likely to win than someone who purchases many. However, everyone who purchases a ticket has a chance to win.

Weighting RNG results is perfectly normal and common in games, and is typically accomplished in a similar way to the example above. You have an algorithm that produces a random number. You draw a number from it, and then take the modulo of that number by another (i.e. the remainder when you divide the first number by the second), and you can weight the results as you see fit. In practice it's a bit more complicated than that, but that's the basic idea.

This is still "pure RNG". The weighting comes into place after the RNG has already completed its work. And this is the way it should be, unless you really believe exalted orbs should be as common as scrolls of wisdom.

As far as the actual data, I think results get messed up somewhat if you consider weapons, since weapons are so diverse and not actually corresponding with a particular type of character (such as int, or dex, or dex/int) as armor is.

Probably the best bit of data in your post is this:

"

97 items found (25.5% of all Equipment items)

Str - 15 (4%)
Str/Dex - 17 (4.5%)
Str/Int - 15 (4%)
Dex - 23 (6%)
Dex/Int - 17 (4.5%)
Int - 10 (2.5%)

This was the distribution of gear collected excluding chestpieces, weapons, and neutral items. This is the fairest comparison, and you can see the sample size is indeed rather small.

Nevertheless, the results produce an average of 16.167 drops per item type. We see most of the item types falling very close to that, one falling quite a bit below (int), and one falling quite a bit above (dex).

This is exactly what you would expect of a non-weighted RNG based system. If your complaint is that drops are weighted against witches, I don't believe the data shows that.

If your complaint is that the data should be weighted in favor of witches while you are playing a witch, I strongly disagree.

Drops are already sufficient to get you through most or all of the game without trading, they are weighted to have a higher chance for higher level items to drop (which is good, so that you don't see level 80 players complaining of constantly getting level 1-20 gear and very little of anything else), and beyond that, the class distribution is roughly equal.

This actually works out better in the long run, because it means:

1) You have a decent chance of finding a drop that might be very useful to an alternate character of yours.
2) You have a decent chance of finding a drop that you can sell to other players for a good amount of currency.

If you weighted the drops so that witches find "witch gear" much more often, then the endgame would actually become far more difficult, because the odds of finding an item that you can trade to others would become much smaller, and those who find it will likely need to be using it themselves, and will thus be unwilling to sell it. You will find more gear that is applicable to your class, but most of what you find will still not have the stats that you want. Unless you are running a niche build, it probably won't have stats that will make it usable to other witches, either, making it effectively garbage. And since you would be finding so many more witch items than items for other classes, you are either helped or severely hurt by the how many players there are playing your particular class.

In addition, if you are playing in a style that does not use gear that is aligned with your class attributes, drops will always be strongly biased against you. This is a Bad Thing.

Weighting drops by class would make the early game easier, but the late game much, much more difficult.

I am glad that you have collected the data, and respect you for the effort, but still strongly disagree with your conclusions.

"

Cronk wrote:

Pure manure. Any PoE player can see and agree with those currency stats. It is you who is failing in the 'factual' side of the issue, and the issue you're having is 'the need to win an argument at all costs'.

Can't agree or disagree with what everyone knows to be right? I'm sorry, but LOL. Nice try though bud.

Sorry, man.

That's people for you. They will always make others do the dirty work, and upon that dirty work, it will always involve a lot more than what is necessary -- and it doesn't matter if you know how to break it down better they ever could, they are going to continue making you do their stupid work and that's a cold-hearted FACT.

Have you notice that none of them are even speaking for themselves or thinking for themselves? All of them are basing this off what they interpreted out of probability class. Apparently they didn't run the numbers between the factors of ignorance and intelligence. :}!