- Past discussions archived to...
- ...Talk:Item level/Archive_1 Archived 04:18, 1 March 2009 (UTC)
General Discussion Edit
This section is for the article, not the contents. E.g. how to break up sections, how to organize the page, etc.
What is this tendency to write iLevel instead of ilevel? The "i" of ilevel has nothing to do with the Apple meme of putting a lowercase "i" followed by a capital letter in the name of their products. —This unsigned comment is by Άλεχανδροσ (talk • contribs) 20:40, November 28, 2011. Please sign your posts with ~~~~!
- People use "iLevel" to distinguish between "i"tem and "L"evel, not any connection to an Apple meme. -- (talk · contr) 1 Dec 2011 6:00 PM Pacific
Stuff to rework Edit
Calculating Item Level Edit
A plot of item budget vs ilvl for WOTLK armor epics with a slotmod of 1 (ilvl 200, 213, 226) gave a best fit of ItemSlotValue = 1.6*ilvl-152.8 with R^2 of 0.99174, or ilvl = ItemSlotValue/1.6 + 95 Hunding (talk) 09:33, 8 March 2009 (UTC)
Currently pre-3.0, so needs updating for WOTLK (spell damage/healing -> spell power change, maybe other changes). I can easily get most of the values from ItemRandomSuffix. Hunding (talk) 23:26, 7 March 2009 (UTC)
DPS Sacrifice Edit
Logarithmic Equation for Item Levels Edit
After a discussion with Falk on the Elitist Jerks message board, I came up with these formulas to calculate item levels for epics and rares. For epics:
iLvl = 105.92 * Ln(itemSlotVal) - 342.12
and for rares:
iLvl = 97.632 * Ln(itemSlotVal) - 287.14
These equations give correct epic item levels from Naxxramas to the newly discovered iLevel 239 and 252 PvP epics. The equation for rares matches rares from AQ20 and ZG up to the iLevel 200 rares from heroics. Adriathys (talk)
- I'm working on doing some linear fits to get better coefficients, but this definitely fits epics from late sunwell to ilvl 252 PVP gear not yet released, when ignoring weapons, where I am not 100% confident on the coefficients. Also, it seems to me that having the formula in the form itemSlotVal = A · e ilvl/b is a more logical way to represent it, and also more likely that that is how Blizzard does it internally, going in the direction of ilvl to budget.
- The fit I'm doing is first a fit on stat coefficients, then a fit for itemSlotVal(ilvl). I get the first fit by taking , where s is the value of the stat and m is the statmod coefficient for that stat and doing a multivariable fit with the data , where the powers are done to each element, and b=mp, and K is a problem-specific parameter for each fit on each ilvl. I set the coefficent for stats and ratings to be 1Hunding (talk) 20:46, 9 March 2009 (UTC)
Weapon DPS and Armor Value Edit
Doing some simple calculations I found that weapon DPS seems to closely follow a quadratic function of itemLevel.
http://elitistjerks.com/f15/t44718-item_level_mechanics/p3/#post1152313 Thorgred (talk) 13:17, 17 March 2009 (UTC)
My ideas for sockets seem to have become 'fact'. Truthiness ftw. Is anyone able to look into sockets and see whether the 'subtract stats to make a socket' is still a good idea?
Slot mod problems Edit
12/11/12 - Despite the hard work put into this by the original author, this is wrong (using the Vanilla stats). Looking at multiple items, I can successfully say the slot mods are incorrect. Slot mods appear to be class specific. Multiple examples of this are available, and apparently in evaluating "several thousands of items", they did not look at items that were obviously similar.
-Belt of Valor - Lightforge Belt - Girdle of the Dawn - Girdle of Uther. All 4 are iLevel 58. Using this formula, the last 3 are about ilevel 58, but the Belt of Valor is at 48.25, clearly being wrong.
-Comparing all of the DS1 Belts (save Beaststalker, due to comparing a stat not on any of the other pieces) yields the same issue - Valor is weak by comparison.
-Field Marshal's Satin Vestments vs Field Marshal's Dreadweave Robe - the priest loot has the same stamina and additional +damage and healing, armor, intellect and mp5. At no point does the Warlock loot have more stats than the Priest's. This dictates it must in fact be different between classes. -Highlander's Lamellar Spaulders vs Highlander's Plate Spaulders - using the formula, they do not calculate to 65, but much more. It also shows Stamina does not have the same stat mod as Intellect between the loot/classes, but that intellect on paladin loot is more valuable than stamina.
I attempted to follow in the original author's footsteps, looking to see what could be wrong in the formula. I tested out the power of log(2)/log(1.5) to see if that was accurate, using the Highlander's Leather Shoulders and Lizardhide Shoulders - it was in fact accurate between Agility and Intellect.
I tried setting up the first example listed below: the formula did not yield 74.4, nor anywhere near it, for the Circle of Applied Forces. I checked if it were an issue of different stat mods for epics instead of rares, also did not check out. I even checked if Belt of Valor were mistakenly itemized under the epic formula by Blizzard instead of the rare formula. Surprisingly, this is in the realm of possibility, as it yields similar results to the other DS1 Belts. I then noticed the Priest and Warlock difference listed above. Currently, I'm testing out what values each class must yield. One thing I noted was the Sergeant Major's Bracers. Each of them carry the same stat numbers: 17 stamina and 7 of strength/agi/intellect.
Strength: Warrior/Paladin Agility: Rogue Intellect: Druid/Mage/Priest/Warlock/Hunter/Shaman
This confirms the previous find of agility being equivalent to intellect between a rogue and druid. This also gives me a headstart on the values of each piece. Stamina must be equivalent between classes, and the other stats are equivalent to each other. This brings us back to the Arathi Basin Plate Shoulders.
Considering that the item levels are too high, I calculated that, if Intellect = 1, Stamina = ~0.869318. I will continue editing this bracketed as I progress.
The sheer size of the equation required to solve the item level formula is a major pain in my back (or maybe it's this broken chair). It appears I'm going to have to make a gigantic set of equations to solve this issue. Each class has its own set of statmods. Fortunately, I'm only looking at classic stats for the time. Considering there are 9 classes, with about 25 total stats to look at, i'm looking at a 225 variable set, with at least as many items.
Unfortunately, my usual online math tools don't typically accept 10 page formulas.