09-15-2005, 05:28 PM
Nice post, that.
Hyzenthlei is obviously our kind of guy :)
I'm not trying to nitpick his work, but to be useful to us statfreaks here, one could suggest a few
improvements.
For one, the weighting base for boni should definitely be STAT-boni. Not only because the
numbers will be a little nicer (see below), but also because to be scientifically sound in a
comparative measurement, one needs to use the most-measured value as the meter for
all others. +stats is of course the most frequent bonus on items. If one wanted to be
100% correct, one should probably base on +Strength, as Blizzard MIGHT change the
relationship between stats (e.g. most players consider spirit vastly inferior to other stats,
perhaps Blizzard does listen and gives spirit an 0.5 SEB or something). For the moment,
this doesn't happen though and it's much easier/better/more precise to use the additional
measurement data one gets by assuming all stats have a 1.0 SEB.
Here is Hyzenthlei's weighting list with an SEB (stat-equivalent bonus) unit.
Strength = Agility = Stamina = Intellect = Spirit = 1.00 SEB
Attack Power = 0.50 SEB
Ranged Attack Power = 0.40 SEB
+%To Hit = 9.5 SEB
+%Crit = 13.9 SEB
Armor = 0.096 SEB
Defense = 1.00 SEB
Damage Shield = 3.1 SEB
+%Dodge = 10.9 SEB
+%Parry = 15.7 SEB
+%Block = 5.7 SEB
+BlockValue = 0.65 SEB
x/5 hp regen = 2.8 SEB
x/5 mana regen = 2.4 SEB
+%SpellCrit = 11.3 SEB
+%SpellHit = 10.9 SEB
+AllSpells = 0.83 SEB
+Healing = 0.43 SEB
+Fire = +Frost = +Shadow = +Arcane = +Nature = 0.69 SEB
+Holy = 0.91 SEB
Fire Resist = Frost Resist = Nature Resist = Arcane Resist = Shadow Resist = 1.00 SEB
I got this list simply by dividing his values by 230 (and rounding to two significant figures).
Note that I do NOT know how many samples he has for each value, therefore I don't
know how to round exactly. Someone would need to run the numbers themselves with
this scaling to find out, for example, if it's really 0.096 for armor or simply 0.1 (my guess).
Hyzenthlei didn't mention what he used to get this data (even though someone actually asked
this in the thread :), but I assume a programmer could do this with a little script-programming
(perl, I guess) and the data from thottbot.
One would need a LOT of data to try and find out the actual rounding employed by the
Blizzard item creator program. OTOH, it seems as though Blizz sometimes changes items by
hand, which would make this impossible.
Oh, I should mention that with the given SEB values, the ival is different from the one given
by Hyzenthlei.
ival = || stats_i ... || _1.5 = Sum_i( statbonus_i * statmod_i ^1.5 ) ^ (1/1.5)
Therefore, his prediction formula needs to be changed, as well:
ival_prediction_green= (ilvl * 0.53 - 4.3) * slotmod [SEB]
ival_prediction_blue= (ilvl * 0.62 - 1.83) * slotmod [SEB]
ival_prediction_purple= (ilvl * 0.71 + 4.9) * slotmod [SEB]
As an example, the Circle of Applied Force would have an actual ival of 30.8 SEB and a predicted value
of 31.4 SEB.
Later,
ink
Hyzenthlei is obviously our kind of guy :)
I'm not trying to nitpick his work, but to be useful to us statfreaks here, one could suggest a few
improvements.
For one, the weighting base for boni should definitely be STAT-boni. Not only because the
numbers will be a little nicer (see below), but also because to be scientifically sound in a
comparative measurement, one needs to use the most-measured value as the meter for
all others. +stats is of course the most frequent bonus on items. If one wanted to be
100% correct, one should probably base on +Strength, as Blizzard MIGHT change the
relationship between stats (e.g. most players consider spirit vastly inferior to other stats,
perhaps Blizzard does listen and gives spirit an 0.5 SEB or something). For the moment,
this doesn't happen though and it's much easier/better/more precise to use the additional
measurement data one gets by assuming all stats have a 1.0 SEB.
Here is Hyzenthlei's weighting list with an SEB (stat-equivalent bonus) unit.
Strength = Agility = Stamina = Intellect = Spirit = 1.00 SEB
Attack Power = 0.50 SEB
Ranged Attack Power = 0.40 SEB
+%To Hit = 9.5 SEB
+%Crit = 13.9 SEB
Armor = 0.096 SEB
Defense = 1.00 SEB
Damage Shield = 3.1 SEB
+%Dodge = 10.9 SEB
+%Parry = 15.7 SEB
+%Block = 5.7 SEB
+BlockValue = 0.65 SEB
x/5 hp regen = 2.8 SEB
x/5 mana regen = 2.4 SEB
+%SpellCrit = 11.3 SEB
+%SpellHit = 10.9 SEB
+AllSpells = 0.83 SEB
+Healing = 0.43 SEB
+Fire = +Frost = +Shadow = +Arcane = +Nature = 0.69 SEB
+Holy = 0.91 SEB
Fire Resist = Frost Resist = Nature Resist = Arcane Resist = Shadow Resist = 1.00 SEB
I got this list simply by dividing his values by 230 (and rounding to two significant figures).
Note that I do NOT know how many samples he has for each value, therefore I don't
know how to round exactly. Someone would need to run the numbers themselves with
this scaling to find out, for example, if it's really 0.096 for armor or simply 0.1 (my guess).
Hyzenthlei didn't mention what he used to get this data (even though someone actually asked
this in the thread :), but I assume a programmer could do this with a little script-programming
(perl, I guess) and the data from thottbot.
One would need a LOT of data to try and find out the actual rounding employed by the
Blizzard item creator program. OTOH, it seems as though Blizz sometimes changes items by
hand, which would make this impossible.
Oh, I should mention that with the given SEB values, the ival is different from the one given
by Hyzenthlei.
ival = || stats_i ... || _1.5 = Sum_i( statbonus_i * statmod_i ^1.5 ) ^ (1/1.5)
Therefore, his prediction formula needs to be changed, as well:
ival_prediction_green= (ilvl * 0.53 - 4.3) * slotmod [SEB]
ival_prediction_blue= (ilvl * 0.62 - 1.83) * slotmod [SEB]
ival_prediction_purple= (ilvl * 0.71 + 4.9) * slotmod [SEB]
As an example, the Circle of Applied Force would have an actual ival of 30.8 SEB and a predicted value
of 31.4 SEB.
Later,
ink