Item drop rate changes

[zakaluka]

I flung this at the o-forum as well, but they're in an uproar at the moment and that crowd isn't receptive to this kind of stuff

From the patch preview article, after 1.0.3 monsters in all acts become able to drop item level 63. In act 2, the breakdown is:

Item level 61 - 19%
Item level 62 - 12%
Item level 63 - 4%
Grand total: 35% of your drops will be required level: 60.

I have collected data in excel on over 20 full bags of items stashed/sold/trashed, magic+ after filling my bags doing all different kinds of runs in act 2. Excluding rings and amulets, because you can't tell what item level they dropped at. This is a representative cross-section of all drops. Let me be the first to say that there is a hidden nerf in these numbers. Currently (1.0.2b) it looks like this within a few percentage points:

Item level 61 - 30%
Item level 62 - 20%

someone will probably say, but zaka! you need thousands of data points to make such a claim. This is actually not true. With a sample size of over 400 items, I have a pretty narrow confidence interval (someone with a real background in statistics feel like chiming in?)

See the trade-off? right now, 50% of your drops in act 2 are required level: 60. Post-patch, 35%. 15% more of your items will become vendor trash. Hey, at least there will be more gold in the economy for item farmers, yay! In return you will see a top-tier item once in a blue moon.

Good or bad? I personally don't think it matters much, just think that the numbers shown in the article don't show the whole story. Since people don't know what the story was pre-patch, the common interpretation people will make is that they're going to get more good items. It's not really true. The bar on quality went up, but you'll be getting more trash drops.

(on a different note, the NV changes are.... interesting. Value of MF on gear just went through the floor.)

[Coriolis]

The problem is that the highest ranges on most of the good affixes only spawn on the highest ilvl items (i.e. 70-80 res all only on ilvl 63). So in the long run (which we're obviously very far from, 3 weeks after release), only the ilvl 63 gear is going to be in the "non-crap" bin, by default. So this change means that you'll have less chance of good-but-not-perfect gear, but some chance of "perfect" gear. In terms of the long-run item inflation, I think the much bigger question is how that 8% ilvl 63 items in a3/a4 compares to the current value, but I can't farm a3 well enough to know.

I think it's a good change. It means that you aren't "forced" to do a3/a4 to have a chance at top tier gear, while still giving the best chance in the latter acts.

[shufti]

Nobody cares about Act 2 or Act 1. Unless you are able to kill champion packs four times faster than in Act 3 (which I highly doubt), it will be a non-factor in farming. You also forgot ilvl 60 items in that equation.

The only interesting thing to know would be how exactly the drop rates are now for Act 3/4 and whether they are truly equal for both acts.

[zakaluka]

You also forgot ilvl 60 items in that equation

Ilvl 60 items are level required:59, so they're completely unimportant. Usually vendor trash unless you get an optimal affix pool and all good rolls.

The act 2 context is merely what I collected data for, and have a frame of reference for comparison. It won't surprise me when a similar thing happens in act 3. Also, yes, I currently CAN kill 2 times faster in act 2 than in act 3 (2x is the key number, not 4x).

This is derailing the original point a little, though - overall, drops will be more slanted to low ilvl than they are now. This is the hidden cost of unlocking ilvl 63 across the board.

[moranor]

Nobody cares about Act 2 or Act 1. Unless you are able to kill champion packs four times faster than in Act 3 (which I highly doubt), it will be a non-factor in farming. You also forgot ilvl 60 items in that equation.

The only interesting thing to know would be how exactly the drop rates are now for Act 3/4 and whether they are truly equal for both acts.

It matters for HC.

[korrekt]

Item level 61 - 30%
Item level 62 - 20%

someone will probably say, but zaka! you need thousands of data points to make such a claim. This is actually not true. With a sample size of over 400 items, I have a pretty narrow confidence interval (someone with a real background in statistics feel like chiming in?)

Your CI (alpha = .05) for lvl 61 items is between 25.5% and 34.5%, CI for lvl 62 items is between 16 and 24%.
CI = p-k to p+k
where k = c⁢⋅ squareroot(p(1−p)/n)
c = z-value of your alpha
p = percentage of lvl 61/62 items
n = sample size

[Hitori M]

Just a thought/question:
When you say currently your drop rates are 30% and 20% (61/62 respectively) do you -
1. Know the item level? Or guessing it? I'm assuming you used a database of sorts to determine the ilvl?
2. Those combined 50% are in contrast of 50% ilvl 60 and below, or do they include non-magical items? I fear you may have reached false conclusions regarding Blizzard's official numbers. I'm fairly certain that Blizzard's combined 35% are in relation to non-magical item as well.

Also, assuming my other 2 points are of no concearn, I'd argue that 4% ilvl 63 is worth (gold and upgrade wise) at LEAST 20% of any other ilvl items you may find (say an average of ilvl 61-62).

[zakaluka]

1) knowing the names of items in each tier. I only have to count 3 times per full inventory.
2) Magic+ is a representative cross section of all drops. A random subset of all drops is upgraded to magic/rare/legendary, so this restriction has no impact on the results as long as the dataset is large enough (which is in doubt).

My point is that I believe they tuned the change to keep progression a constant. Item max quality goes up significantly, average item level of drop goes down significantly. Overall result is that progression doesn't change much, even if better drops are available to all.

This experiment ended some time ago, but my 95% CI is wider than I like if korrekt has given me correct info (ha ha!) I think the experiment would have to be repeated again with a larger dataset (50 full bags or more), but the arbitrary restriction of magic+ has no impact on the outcome. Biggest problem with my experiment is the confidence interval. I don't think I care enough to do it again.

[Hitori M]

I'm not sure I follow, being new here. By "Magic+" you mean you've only counted the magical items and not the white drops?
If that's the case, are we certain Blizzard's numbers are using the same arbitrary method? It would seem to me they may be using that method, but I've no idea.

I guess even if they did nerf it, it's worthwhile for everyone. At some point the current loot tables get really boring for anyone, even if you are one of the capable Act 3/4 farmers, being forced to play only half of the game's content (if that) is a bummer. If you're able to clear acts 3/4 within reasonable time, act 1 should be a fun 15 mins for you if not less, to get a nice chance at some decent loot (no matter what's your gear level). I believe it is ideal, even if a nerf to "lesser" items is included.

[zakaluka]

again, magic+ is a representative cross section. I followed you the first time

It doesn't matter whether or not blizzard's info follows the same restrictions.

It's a representative cross section. I know we're running into a language barrier here, but I don't know different words to describe the concept.

The only problem with my experiment is that my dataset was too small. I asked for someone to evaluate that for me, and it turned out I was wrong. Dataset was too small.