- wowrank@thoth.purplefrog.com
Apologies Patch 3.0.2 changed a lot of stuff. I assume a lot of the formulas underlying my scoring functions no longer match the new rules in the game. Also, when Wrath of the Lich King debuted in mid-November, many gear rankings ceased to be relevant. I still have to redo a lot of the formulas for L80 values.
I have created a perl script which ranks an entire World of Warcraft character with his equipment. I was introduced to the concept by the Rawr tool for feral druids.
My vision is to apply the Rawr methodology of ranking an entire gear set to more than just feral druids. I also have the idea of a "pinnacle set" which is the set of gear which maximizes the score function for a certain tier of gear. My suspicion is that as gear from other tiers becomes avaialble, it changes the value of older gear. This is because many (if not all) classes will be ranked by functions with non-linear components (uncrittability, spell hit cap, armor-based mitigation).
experimental spreadsheet skeleton (far from complete)
I am also toying with the idea of just harvesting the scoring functions from existing spreadsheets. Finding the boundary between the scoring function and the basic stat computation would be a non-trivial job. I'm sure the theorycrafters would consider copying their spreadsheet formulas into my skeleton a similarly non-trivial job.
15 - I had not scanned all the items so some itemLevel tags were still showing up in the diffs. I suppressed them. More items lost their socketMatchEnchant. It appears there is now no socketMatchEnchant data in the database. Blizzard must have dropped it from their XML (how retarded!).
10 - It looks like Blizzard has added itemLevel to the XML (so the diffs were huge and I had to filter it out). A lot more items lost their socketMatchEnchant. Some items gained itemSource information. Some lost requiredPersonalArenaRating, or had it reduced (new arena season?). A couple of block values changed.
11 - There are lots of changes (probably all occurred with 3.2 and only now that I have been harvesting updated XML for a week do we have a decent picture). Lots of gear had its MP5 adjusted; block values were adjusted; itemSource areaNames were removed from some items (weak, Blizzard); Some DPS numbers appear to have changed by 0.01; a few set bonuses changed; and a few really obscure changes.
7 - Patch 3.2 hit last Tuesday. It's going to be a little while before Blizzard lets me download the new items' XML from the armory. In the mean time I have downgraded all emblem of heroism/valor/conquest gear to rank 7 (heroics) since heroics now drop Emblems of Conquest. I would not be surprised if some badge gear became top-of-slot even for Naxrammas.
I am not sure what to do about ranking 10-man heroic -vs- 25-man normal loot. A person in an elite 10-man guild might not have access to 25-man loot (because of his choice of play style) and a person in a mediocre 25-man guild might not be able to get loot from hard-mode 10-man raids. If I were not lazy I would have gear rankings for a separate 10-man track, a separate 25-man track, and a unified 10+25 track.
I have not incorporated the new dodge formula yet.
27 - The original item diffs for the 21st were 14 megabytes. The vast majority of it was an alteration of the XML to show the item as fully repaired instead of broken. I have tweaked the diffs so if that is the only change it will not be listed. This still leaves 2.5M of diffs which appear to be entirely socket bonuses being added to Burning Crusade gear (and a few WotLK items).
7 - My logs have been flooded with update messages for the past few days. It turns out that it's just a change in the <?xml-stylesheet?> tag at the beginning of the file. I have tuned the archive diff to suppress that difference.
30 - I started archiving copies of my item database. Now I am publishing automated diffs to detect changes in the armory XML.
12 - My first gear-refresh loop completed earlier this week. Now the first recent changes have been spotted: The damage on three fishing poles ( ids 636[567] ) has been increased.
2 - After the Harness of Carnal Instinct fiasco, and considering some of the Ulduar gear tweaks that are happening in various 3.1.x releases, I have finally coded a tool to refresh the item database. It is currently tuned to refresh the item database approximately once a week (19500 items, 30 second interval between items).
5 - The Harness Of Carnal Instinct was showing up as better than Ulduar gear. It turns out I had some ancient stats from the armory, and the armor value was over 800, which is not even remotely accurate these days. I wonder how many other items have stale armory XML.
21 - Evidently there's a way to paste a pawn config into the UI to speed things up. Maybe this works:
| ( Pawn: v1: "Igoar-Rawr": Agility=58.7, Stamina=37, BaseArmor=17.5, BonusArmor=3.5, DefenseRating=35, DodgeRating=54.8, RedSocket=1023, BlueSocket=1203, YellowSocket=863.7 ) |
17 - since Ulduar is out and Blizzard hasn't released all the gear, some of you bears out there might want numbers you can plug into Pawn.
14 - Defense rating was egregiously overvalued because some expressions were still using L70 math. Look for rearrangements in tank gear.
5 - I am no longer applying the bear armor bonus to necklaces and cloaks. When I filtered out low-level items from the pinnacle computations I accidentally filtered out some quest rewards that would have been appropriate; fixed.
25 - I have altered the bear armor bonus to not apply to rings, trinkets, and weapons. The actual change by blizzard will probably be much messier with necklaces and cloaks being affected and I am fearful that they will not adjust the item XML to have usable information.
21 - I have some new druid gear rankings for level 80. Some of the gear is probably in the wrong rank/tier. I'm sorting this out as I notice it. There are also inevitably mistakes in the math, and talents I have overlooked.
1 - The 3.0.2 patch changed a lot of gear, and some of the rules. I have downloaded all the new item XML (which is now lacking the arena rating information, bugger) and am calculating new gear rankings. They will be approximate until I find all of the rules that changed. I probably will not get around to changing everything before Wrath of the Lich King arrives and makes it all useless (again).
3 - I added a gear ranking function for balance druids based on spamming wrath. I discovered a mistake in my mana recovery math and this will affect the resto gear ranking too.
28 - I rehacked the damage calculations to take into account block value. My placeholder value of 50% damage block was far out of line with a boss fight. It was seriously screwing up the warrior gear rankings. I am recalculating all the tiers.
20 - I have added support for socket bonuses. I need to do a recalculation. I am slowly creeping toward on-demand gear rankings.
6 - I have separated Tier 6 from Sunwell. Sunwell is now "6.5". I realized that Sunwell drops include a lot of items that don't involve Yrma. Luckily Wowhead has a list of "zone drops" I can harvest.
17 - I took a first stab at a holy light spam paladin gear ranking.
10 - added the Brutal Gladiator (season 4) gear.
11 - I did some tier 6 rankings last night. The bear recommendations have an 0.2% crit vunlerability at Tier 6, weak. I have harvested some data about sunwell items, but the majority of their stats are unavailable from the armory.
10 - the reranking completed last night. There are some clear T5 upgrades for bear tanks. The resto druid gem insanity indicates that the T5 gear is capable of dancing across the OOM line in a 9-minute fight (as evidenced by all the (540/2) appearances in the explanation lines).
9 - Blizzard's armory is now divulging stats of the new Shattered Sun badge loot. It should appear in the next rerank. I am also scraping the loot/vendor table of a few NPCs to provide more meaningful "source" information.
8 - I'm now caching the computations of the pinnacle sets so that rerankings won't be upgrading as many (if any) pieces of gear, and reranks should run a lot faster.
27 - I cobbled together a function for scoring rogues. It is based on white damage, which is rather absurd, but better than nothing. I added a section to the ranking page for melee attack table. I added a section to the bottom of the ranking page for per-stat point value.
24 - I reworked the gem readout to score each gem, including gems that are clearly inferior to other gems (clearly inferior gems have been excluded from pinnacle set calculations).
17 - Meta gems are now placed in head sockets, but their activation conditions are not accurately modelled.
5 - I threw together a "holy paladin" scoring function based on spamming Flash of Light for 9 minutes (a semi-average berserk timer). I noticed that on-use trinkets were having their bonuses scored as on-equip. I'll have to fix that.
31 - I'm much more confident in the tier assignments for items now that I've correlated them against the Atlas Loot categories.
28 - I have a big list of item ids from AtlasLoot and will be incorporating them into my database slowly.
27 - I spent half an hour putting on one piece of gear at a time onto my druid and encoding the character pane into unit tests. My computations now accurately predict the 5 core stats and do an OK job with armor, defense, and resilience, but are still quite flaky (off by +-0.03%) for dodge. I also don't handle meta gems, gem socket bonus, or set bonus. I need to work on other classes too.
23 - shadow priest, done poorly. Hunter, kind of like white damage, but probably wrong. woo
22 - I currently have done a sloppy job of assigning items to the following tiers: solo, group, raid, heroic, T4, T5, T6. The distinction between group and heroic is unlikely to be correct.
12 - The underlying math is bad, but the annealing function mostly works. I have an approximation of the Rawr bear scoring function.