Soul

But [Power of the High Chief] is just a [Briarwood Reed] less the RNG aspect, and that trinket was introduced in its current form in 1.4 (after the spell damage item budget revaluation, but before Dire Maul). I suspect that in the olden classic days, trinket procs and activations were simply grossly overvalued and the item designers simply weren't fond of using trinket slots as mere stat pools. I know for a fact that in the old classic days, different slots of equipment had different weightings for different stats - for example, spell damage on wands took up less budget than spell damage on other slots or other stats (such as INT) on wands.

 

ildon

I think that was only on green wands, and it was probably an oversight/bug. If I remember correctly, blue quality wands with spell damage on them had expected spell damage for their budget, which was also much lower than the level 58 green wands.

 

Thorgred

As the author of a lot of the more recent updates on the wowwiki discussion on this topic i'm quite interested in the itemisation formulas. Mostly, i'm interested in working out which are the over/underbudgeted items, seeing as there appear to be so many of them in WotLK.

I based my assumption that the stats were 'removed' for gems just on my own observations, as giving a fixed stat cost to slots didn't fit at all. It was just a theory .

The reason that socket-able items tend to be 'better itemized' is that they will remove stats that you don't particularly want. Ie, an item with 2 blue sockets was created by removing 48 Stamina. However, you want to socket them both with Spellpower: the same item, stacking spellpower, will be a virtually higher iLevel because of the way the system penalizes stacking one stat high.

AS far as i'm aware, a Spellpower weapon has 61.5% of the DPS as an equivalent iLevel 1H weapon does, with the remaining 48.5% converted to spellpower at a rate of ~7.5 SP per 1 DPS. That is only preliminary observations but hopefully it might spark some discussion .

 

mek

Those calculations are correct, and the same as they were in TBC. Socketed items (with blue gems, a big caveat) are no better than unsocketed items with the same color stats, if you mindlessly pick your gems according to the formula. But you don't. You might stack one stat in every socket and ignore bonuses, or take some stat that is optimal for you but not on the item at all. (eg. if the item has two blue sockets, 75 str and 50 sta, and you put two strength gems in, that item is now "over-budget".)

And then when you get your hands on gems a tier higher, those are effectively free stats.

 

Soul

What blue quality wands were there with spell damage back then?

Note that this is all derived from hindsight, since the very concept of item budgeting as we know it wasn't extrapolated until about WoW 1.9, IIRC. Back then, we had no clue that the green wands were over budget at all because epic spell damage wands all had stats and we didn't know how the stat budget worked. We just knew that as casters, stats didn't really do much for our performance as compared to spell damage, so we were sticking with our greens.

Note also that when the budget formula approximation was extrapolated by Hyzenthlei, he found that resistance on rings also had a lower budget than other stats on rings or resistance on other slots.

 

ildon

[Wand of Biting Cold]
The example has kind of been ruined by all the spellpower/itemization changes since AV was released, but this wand originally had 16 frost damage, and was less frost damage than a [Dragon Finger] of Frozen Wrath (21 frost damage). After the green wands were fixed, they gave 11 frost damage.

Edit: If you compare, for example, a Dragon Finger of the Owl or of Spirit (which were never nerfed/changed), to a blue dungeon drop wand with static stats, even without knowing the itemization formula, you can see the blue item is strictly better. It was obviously a bug/oversight that only affected "of X Wrath" items. Just because you don't know it's a bug because you don't know the item budget formula doesn't make it not a bug. :P

 

Seratha

I touched on this in the first post, but sort of waved away the explanation to it, but in current weapons there's no longer a fixed ratio of spell power to DPS lost. Prior to WotLK, you could trade one DPS for four Spell Power, but the way it works now is that the ~43% of the weapons DPS is no longer present and converted to spell power based on the item slot value. It's no longer a fixed trade and you can see that by taking the amount of DPS sacrificed and comparing it to the amount of spell power gained. The rate at which it's lost will go from about 6.7:1 up to about 7:1 as the item levels increase.

In the 200 item level range: [Knife of Incision] vs [Blade of Dormant Memories]
DPS difference: 60.7
Spell Power: 408
Ratio: 6.72

213 Range: [Murder] vs [Ice Spire Scepter]
DPS difference: 66.2
Spell Power: 461
Ratio: 6.95

226 Range: [Sinister Revenge] vs [Torch of Holy Fire]
DPS difference: 72.2
Spell Power: 520
Ratio: 7.2

The ratio of spell power to weapon DPS is just going to get larger as we get higher tiers of gear. Based on three data points, I could come up with a rough equation for item slot value to spell power for epic items, but that will just get revised when Ulduar comes out and we get another point or two.

On sockets, I'm still of the opinion that they subtract stats to create sockets, but I'm open to the idea of a fixed, independent cost for them. I wouldn't be surprised if the metagem socket is a fixed cost since there's really no stat that you can use to quantify it unless you assume the static value of a metagem is the cost and the bonus multiplier is just free.

 

Mman

That is not exactly right though. If you go through the epic pieces without any gem slots you will find that their actual ilevel is almost always below their expected, even if it is just a small portion of an iLevel. For instance, [Undiminished Battleplate] has an actual iLevel closer to 212.7. While this would seem like an insignificant amount, it would actually allow you 8 crit or agility to be added the item. That is half a gem.

If you go through an look at socketed items though, the opposite is almost always true. Their actual ilevel are above the rounded number. The more sockets you add to an item the further it gets away from the reported ilevel. [Gown of the Spell-Weaver] has an actual ilevel of 213.4 (socketed for spellpower). [Chestguard of the Exhausted] would actually have an ilevel of 214.4 if you put strength gems in. The resist items with 3 sockets, like [Glacial Robe] wind up with an ilevel of 216.7 socketed with stamina.

You wind up with socketed items being a good deal better than nonsocketed ones as long as you pick your gems according to the formula. The ability to socket whatever stat you want into a slot actually makes socketed items worth less, in terms of pure ilevel. Adding 16 of a new stat is way cheaper than adding 16 of the highest stat. Either 1 of 2 things are going on with socketed items then. Either Blizzard is just making a mistake and giving socketed items too much value, or they are calculating the potential loss in ilevel (by socketing in a new stat) into the formula somehow.

 

◊ vorpalblade

(emphasis is mine)

Sort of. You're absolutely correct in that the ability to socket whatever you want will often result in a "lower" ilvl, simply because the stats you're adding can be underrepresented on the pre-socketed piece, and as such don't consume the same itemization points that socketing for an already represented stat would consume.

That said, I think for the purposes of this discussion, it's very misleading to be using the terms "better" and "worse". You can pump the ilvl of a socketed item by dropping in gems that pile on more of an already well-represented stat, but higher calculated itemlevel doesn't make something "better" or even over-budget. Hence, it's misleading to think of itemlevel as something you want to try to maximize, so it's probably better to avoid qualitative comparisons like "better" and "worse". The itemlevel is used to determine a way to allocate stats to an item, and thus far the formula specified earlier regarding sockets has been very close, when you assume that sockets are consuming itemization points equivalent to one of the stats already present on an item, as opposed to a previously unrepresented stat. Those items aren't necessarily "over-budget" even though you can give them a higher apparent ilvl, and that flexibility is a main part of their appeal.

Sure, there are still some unknowns, such as the Red Yellow and Blue sockets that you noted which are present on pieces containing only Stamina and Resistance, raising the question of which stats are they consuming itemization points from on these pieces to budget for those sockets? These need to be accounted for in the formula, but their existence doesn't invalidate the application of the formula for more traditionally allocated items.

 

Mman

But it isn't only the 3 socket item where the formula doesn't work. That one just works the worst. Using the socket theory in the first post resulted in socketed items always having a higher ilevel than nonsocketed ones. I think that formula is in the right direction, but clearly there is something else going on. The problem I keep running into is that nothing is consistent. Using the theory for various 2 socket 213 ilevel items sometimes got me between 213 and 214 and sometimes got me an ilevel as high as 215. I suspect, because 2 socket items result in a higher ilevel, that the second gem is somehow less valued.

Also, from the brief testing I did, it doesn't seem like 1 socket is taking the highest stat and the second is taking the second highest. I need to do more testing on this though.

 

Thorgred

In the spreadsheet i created to investigate BC iLevels, I found that the best way to find the correct iLevel was to give each stat +/- 0.5. This could result in a difference of up to 5 iLevels either side of the initial expected value.

To demonstrate what i mean by this, i'll use Green chestpieces of intellect
An iLevel 114 chest has 53 Intellect.
An iLevel 120 chest has 56 intellect.
An iLevel 117 chest has 54.5 intellect, to be exact. However this is rounded up to 55 intellect. If you did a reverse calculation, then this item would appear to be slightly over-budgeted.

So therefore in my calculations, i displayed 3 values of what the iLevel range could possibly be: One with the exact values as you see on the item; one calculated with the values - 0.5; and a third with all the values +0.5.

 

Scrufola

It was not spell damage but only a school of damage (of the shadow wrath, of the fiery wrath, ...).

And the interesting thing is, when they "fixed" it, the existing items were not changed. From that patch on, there where two items (for every wand they changed) with the same name in the game where the old had about twice as much shadow damage on it than the new one. Only the new one dropped from this patch on, of course.

That might be relevant to this thread because it shows that they leave (or left) items in the game which were wrong itemized.

I still have this wand in my bank and it still only contributes to shadow damage. This item was not changed to spell damage with the WotLK patch.

And out of curiosity: When Hyzenthlei wrote his "Item Balance - From Epics to Errors" he mentioned that there are some items which are too good but he did not mention which items. Do we actually know which items he was talking about? I always wanted to know.

 

Jerry

From an item design standpoint, item level is the first known quantity. It's from ItemLevel that the item budget will be decided (based on slot and quality), and then finally this budget will be consumed on stat, sockets or procs/On use spells. As such, an item can be ender- or over-budget, but it can't have a "lower/higher itemlevel". That just doesn't make sense, to me at least.

One thing that the current model for budget allocation does not consider is how is determined the number of different stats that may get a part of the budget and the amount of budget allocated to each stat. The formula:



where do not help to explain why there is no item which give budget to 10 or 12 individual stats, or why there is no item which gives 100 points to one stat and only 10 points to another.

I think that the real model for budget allocation is simpler but more restrictive than the given formula. Something like, for a given quality, the designer can choose between a small number of "templates" of allocations, each template giving the number of stats and the fraction of the total budget given to the stat.

 

Prinsesa

Correct me if I'm wrong, but while 3.0 did change all +healing items into +spell power items, it did not change all school-specific spell power items. Only special items such as Frozen Shadoweave and Spellfire got changed into generic spell power, but school-specific random enchantments still persist to this day among low-level greens.

"We do want Sanctuary to be the tanking seal"

- Ghostcrawler

 

Thorgred

What i mean to say with the lower/higher thing is that an item is most probably designed with some kind of slider bar, or % template as you suggest.

If their % template is followed exactly, splitting the budget into percentages, a few items might come out as say:

55.7 Stamina
67.6 Str
29.6 Crit rating

which is then rounded up to 56/68/30. And so when we reverse-engineer the formula, the item appears 'over budget'


What you are saying about certain templates - I agree to an extent (and look up the 'item random suffix stuff on wowwiki) but it all does fit to the formula. The reason we don't see items with 100 / 10 or 5/5/5/5/5/5/5/5/5/5 is that no one wants those items. No one uses the straight +Str greens. And there are a few low level items with 5+ to 6 stats, and they are not used at all either.

 

ildon

There's some pretty strong evidence that randomly generated stats are "baked into" the item at creation, so it'd be a lot more work to change all the "of shadow wrath" items already in the game, because it involves parsing the entire item store of every character, instead of simply changing the base item in the main database (and the tables from which randomly generated items draw their stats). This would also pretty well mirror Diablo 2's item creation (which would make some sense, seeing as both were made by the same company).

Plus "fixing" old randomly generated items would likely cause a lot of ill will with players who possibly paid several hundred gold for the item in the AH or whatever.

Anyway, I was talking about spell school specific damage on random greens, and its associated too-high stat allocation. See my second post on the subject a little further up. I don't believe it had anything to do with "budgeting school specific damage on wands differently", at least not on purpose, and everything to do with simply having the wrong data in the table from which randomly generated wands pulled their stats, which was my original point.

 

Mman

If the items were rounded up though I don't think that would make an item over budget. Rounded to the closest integer the item would have the proper ilevel. If that is not the case their their "template" is incorrect. It also doesn't answer the question why some items are rounded up and other are rounded down. On a lot of the items that are rounded up to the iLevel you have room in the budget to actually increase the stats (even if it is only a single point) without bringing it over budget.

 

Ukerric

I would not refer too much to "templates". Put all the items in a spreadsheet, and you'll quickly see that there's no real template that uses a given proportion "times" the item's ilvl. Even if you try to find all, say "priest helm - high spi, medium int and crit, lower sta", you don't find a family with each item being a direct interpolation of each other, modified by the ilvl. All helms with higher spi have varying relative amounts of int and crit. If there's a template there... it's been used only once!

Except one: The PvP items. Each successive incarnation of PvP items is close to a linear 10% (modulo rounding effects) increase in each stat, and each family of items have exactly the same repartition in stats.

For all other items, it appears that they were individually designed. Sure, the designer had an "item bible" to work with (such as vanilla "thou shalt not put STR and AP on the same item"), but I don't think there's a spreadsheet that spits ready-made items ("I need a discipline priest wrist, epic ilvl 226"). In fact, they told us that item were deliberately designed not to be 100% optimal this early in the expansion, to have room for improvements.

 

NemoX

I made some modifications to an old addon "ItemTips" that I found on curse using the formulas as they're currently understood. The addon adds a line into the item tooltips with the real item level, a calculated item level, and the item's ID. For the most part, the formula seems to be correct and the values are within a level or two of the expected values.

For example, [Belt of the Tortured] and [Boots of Persuasion] both calculate out to a score approximately .5 from the expected, but [Sash of Solitude] comes out significantly lower at 209.4. I'm sure this has something to do with how the socket values are being calculated, but I'm not sure what I'm doing wrong. I've tried several different strategies for calculating the values, and the closest formula I've found is to take each socket on its own, find the corresponding maximum primary stat, subtract the primary stat's value N^ln(2)/ln(1.5) and add back in the stat as if the gem's value was not taken away.

A meta is defiantly undervalued at 18, since every helm with one comes up short. I also apparently managed to break the pre-BC calculations for epics, despite my attempts to preserve it.

The addon (in its current state) can be found at: ItemTips

 

◊ zeidrich

The itemization formula is largely arbitrary. Designers don't have to follow it, as evidenced by many items that were "misbudgeted".

The presence of poor or minor stats (like Ar.Pen, Parry Rating, Haste Rating) gives the designers a lot of knobs to turn to vary items of the same item level in power. This is important because as I'll try to demonstrate, there's a lot of potential variance in similar ilvl items; so much so, that if you were to itemize naxx gear perfectly, ulduar gear would have to be itemized perfectly as well, or it would be useless. The graduation of stats helps ensure that future gear will still be more than a marginal upgrade.

Take for instance a tank piece. Assume that part of the budget has been used up by stamina, and strength, and we're left with 100 points to spend on avoidance. Namely, dodge and parry rating. Now parry rating is, for all intents and purposes a "Junk Stat". Dwarfed by both dodge rating and defense rating. But due to the way the item level formula works, it's still useful. Take the following table:

%parry	ip	% Av	% Av @ 100	Factor
0	100	2.541296061	2.541296061	1
5	95.36159474	2.515898602	2.638272367	1.048640181
10	91.2245982	2.490501144	2.730076309	1.096195565
15	87.53609273	2.465103685	2.816099746	1.142385922
20	84.29377878	2.439706227	2.894289783	1.18632717
25	81.50852758	2.414308768	2.962032121	1.226865495
30	79.19640904	2.388911309	3.016438925	1.262683513
35	77.37489396	2.363513851	3.054626288	1.29240888
40	76.06029993	2.338116392	3.07402994	1.314746327
45	75.26581596	2.312718934	3.072734819	1.328624406
50	75	2.287321475	3.049761967	1.333333333
55	75.26581596	2.261924016	3.005247452	1.328624406
60	76.06029993	2.236526558	2.940465078	1.314746327
65	77.37489396	2.211129099	2.857682882	1.29240888
70	79.19640904	2.185731641	2.759887307	1.262683513
75	81.50852758	2.160334182	2.650439465	1.226865495
80	84.29377878	2.134936723	2.532733441	1.18632717
85	87.53609273	2.109539265	2.409907958	1.142385922
90	91.2245982	2.084141806	2.284627006	1.096195565
95	95.36159474	2.058744348	2.158882046	1.048640181
100	100	2.033346889	2.033346889	1

(edit, fixed table)

The first column is the proportion of parry rating on the item, the second column is the item points that would be used if exactly 100 rating points were split under that proportion, (IE: 60 Parry Rating / 40 Dodge Rating would cost 76.06 item points) The third column is how much avoidance you would get if you were to use that proportion. The 4th column is the avoidance you would gain for using up exactly 100 item value was spent on that proportion. And the final column shows how many rating points you get in relation to if you were just using a single stat.

So the "sweet spot" so to speak is at 60% dodge, 40% parry. There, you can get ~78 dodge and ~52 parry and get ~3.07% avoidance. All at the same price as a straight 100 dodge (2.54% avoidance) or 100 parry (2.03% avoidance)

The point is, having a "throwaway" stat like parry interacting with dodge still adds value to your items. In fact, about 20% more avoidance than if you had straight dodge for the same item level. And having both stats there gives means that there's a giant range between the worst configuration and the best configuration. For 100 item value, the best config (78 dodge, 52 parry) is about 150% as powerful as the worst config (100 parry).

This means the designers have a lot of room to range within their item value formula. If parry was equal in value to dodge, the spread between the best (66 parry, 66 dodge) would be only be 133% higher than the worst (100 parry or 100 dodge). But the bigger problem is the gap between the best config and the best single stat. Between the old parry and dodge, it's a 20% spread if you choose just dodge. If they're both equal, it's 33%. This means that any gear with solely dodge on it is in danger of being weaker than any gear with both stats.

If that were the case, either no items could be created with both dodge and parry on them. Or else, if they were, they would be de facto stronger than any item with just one of the two. But using differing avoidance conversions it allows them to create equal power items with different flavor. IE: 100 dodge is about equal to 94 parry, 23 dodge in both item level and % avoidance.

You can see this sort of sentiment expressed in Strength and AP items. As, in general (not counting scalars like kings and divine spirit) 1 strength is about equal to 2 AP, and they, for the most part serve exactly the same purpose, and the same stat budget. If you were to put both on an item it would be categorically better than an item that just had one or the other.

That's of course why gems seem to allow you to bump up the "effective ilvl" of your gear by stacking stats on top of eachother, which wouldn't be "allowed" under typical item creation rules, making them overbudget. However, just because the effective item points are greater doesn't necessarily mean that the power of the item is greater. Take the parry/dodge example again. If you had a 94 parry, 23 dodge item with a socket, the way to push it's "effective itemlevel" as high as possible would be to put a parry gem in. But anyone could tell you that a dodge gem would easily give you better returns.


Very TL;DR:

Item Level mechanics are interesting. However, since the actual stat allocation is really abitrary, and the item level is used solely as a guideline, and due to the fact that there is a massive amount of variance in what sort of stats and efficacy can be put into identical ilvl items, item level should never be used as a metric as to how strong an item is.

In Vanilla, when ZG and the Dungeon 2.0 sets came out, Blizzard started putting in "epics". These epics had purple names but were lower item level than many blues. ZG dropped ilvl 71 blue items, but ilvl 65 epics, which meant they were very closer in power level.

People picked up on this and were upset by it. So now Blizzard has stopped messing around with item levels, but has instead put in a number of minor stats with variable efficacy that they can use to raise or lower the power of items. And it's just arcane enough to keep people from feeling cheated unless they make an item really bad.

But as always, to evaluate items, it's infinitely more important to look at the value of the stats on them items than it is to look at the item level.

 

Buttons245

not sure if it has already been mentioned, but was wondering about the fact that ilvl200 trinket "Mirror of Truth" has 84 Crit rating, and ilvl213 "Rune of Repulsion" has also got 84, but parry rating instead. Does the equation find parry rating to be ranked higher than crit? Be nice if you could shed some light

EDIT: Had a read of the bloke aboves post, if it is indeed a junk stat and holds less value on other stats, why does the higher lvl item have less stats?

 

Petrus

Rune of Repulsion actually has 95 parry rating, not 84. Valor Medal of the First War, the badge dodge trinket, has 84 dodge rating, however.

I'd like to go back and touch on the Block Rating/Value itemization. For some reason, only those four items - Wall of Terror, Inexorable Sabatons, Shadow of the Ghoul, and Fleshless Girdle - are short on stamina. It isn't clear if it's block rating or block value, but those four seem to have been budgeted at the incorrect item level.

From a discussion on this at Maintankadin: Maintankadin :: View topic - Question on blocking gear Ilevel

Honestly, it seems a bit silly for those FOUR items to have lower stamina (and thus seem underbudgeted). I compiled a list of block items iLevel 213 or higher (as those are the only ones affected) for comparison, and the correctly budgeted pieces still outnumber the underbudgeted ones 3:1. iLevel 200 block items seem just fine, with proper amounts of stamina on all pieces.

Here's the full list of items affected by this misitemization:

Fleshless Girdle (75 stam instead of 100 + a socket)
Shadow of the Ghoul (56 stam instead of 84)
Wall of Terror (63 stam instead of 94) (226)
Inexorable Sabatons (63 stam + socket instead of 111)

Unaffected items (i.e. they have block stats but have stamina that is roughly equal to other pieces of their iLevel):

Nexus War Champion Beads (94 stam, 226)
Bindings of the Hapless Prey (84 stam, 213)
Burdened Shoulderplates (112 stam, 213)
Chestguard of the Exhausted (126 stam + 2 sockets, 213)
Drakescale Collar (72 stam + socket, 213) [Edit: it looks as though this piece too may be underbudget]
Gauntlets of the Disobedient (112 stam, 213)
Greaves of Turbulence (150 stam, 213)
Platehelm of the Great Wyrm (150 stam, 213)
Warrior/Paladin T7 Chest (127 stam + 2 sockets, 213)
Warrior T7 Shoulder
Paladin T7 Legs
Paladin T7 Helm


And here's a link to the official Tanking forums post on the subject: World of Warcraft - English (NA) Forums -> Block rating items severely underbudgeted?

 

Thorgred

As said before blizzard has ways of artificially making items 'bad'. They've admitted to spending item points on sub-optimal stats on tier7 gear so that tier 8 will be better.
For example, as you said, they can spend 100 points on parry, or 60/40 on Defense/dodge, for the same ilevel. They spend points on haste or armor pen, or alter weapon speeds.

I think this would all be a purely academic exercise if we didnt know that some items in game are severely under-budget. Ie this one Ghostflicker Waistband. but seeing as we do know that some *are* actually underbudget (not just badly itemised) I'd like to be able to prove that it is a bug and needs to be fixed. Or, to be able to hunt down those pieces that are slightly over-budget and use them to my advantage

 

Scrufola

They changed all school-specific green TBC items into generic spell power items and even renamed them. "X of shadow wrath" and "X of healing" both became a "X of spell power". I haven't seen a single green TBC item which kept its school-specific damage. This is also true for at least some vanilla items.

Enchants (weapon and gloves) staid school-specific.

All "X of shadow wrath" items my shadow priest had equipped during TBC were changed with WotLK to "X of spell power". I used at least the off-hand and a waist "of shadow wrath" and both were changed while he was wearing them. (That doesn't mean your theory is wrong.)

 

bdew

This is false. All the random sufixes are stored centrally with your item having only the ID of the suffix. On the client side this is in ItemRandomSuffix.dbc