Chromatic orbs are bullshit.

I view it more likely that the model is bad.

"

RogueMage wrote:

Two of the worst examples of this fundamental mismatch are daggers and evasive shields. Daggers are primarily melee weapons

Daggers are not primarily melee, they are crit based. They have both melee and spell caster mod pool and are associated with the Shadow tree which is hybrid dex/int. Pure evasion shields are in a weird position because you cant use it with bows. Daggers and shields only have 3 sockets, its not like trying to roll 6 off colors on a chest or 2h

"

JIIX wrote:

Since I was intrigued, I've made some Matlab simulations during my lunch break.

1. I made a N-by-6 matrix of random number between 0 and 1.
2. Value between 0.0 and 0.8 are B
Value between 0.8 and 0.9 are G
Value between 0.9 and 1.0 are R
3. Count the number of time 2R-and-whatever appears

Using N = 1000000, it gives a probability is 0.015277 i.e. 1/65 to have 4 off-color socket. So that math is definitively good.
Replicating 1000 times the 400 trials of the OP gives a probability of 0.002 of having 0 off-color so that math is also definitively good.


However, using the OP's data, the distribution of color is more like B=0.8358, G=0.0854, R=0.0788

Using 0.84-0.08-0.08 instead, the probability shifts to 0.0067978 i.e 1/147 and the probability of having 0 off-color in 400 chromatics is about 0.065 (32 times the previous value)

More data would give more precision but 0.84-0.08-0.08 and 1-in-15 bad luck seems more likely that 0.8-0.1-0.1 and 1-in-500 bad luck

if you use the same base asumtions ofc the math fits ^^

the adjusted probabilitys are certainly interesting - but even there any 4 off-colors would be 1 in 144. So OP was either very unlucky or there is a bias towards fewer off-colors.
Certainly needs more data.

@tikitaki: if you pay me you can call everything I do wrong for all I care :P
Anyway with only OP's data it seems the current accepted model for calculating color-chances is wrong. Or at least the chances have to be adjusted to 0.84 0.08 0.08.
However for 4-socket items the accepted systems seems to fit, so if there is a bias against 4+off-colors there are going to die a ton of chromatics to proof that.

"

tikitaki wrote:

I view it more likely that the model is bad.

What's wrong with 84-08-08?

"

Buchsbaum wrote:

"

JIIX wrote:

Since I was intrigued, I've made some Matlab simulations during my lunch break.

1. I made a N-by-6 matrix of random number between 0 and 1.
2. Value between 0.0 and 0.8 are B
Value between 0.8 and 0.9 are G
Value between 0.9 and 1.0 are R
3. Count the number of time 2R-and-whatever appears

Using N = 1000000, it gives a probability is 0.015277 i.e. 1/65 to have 4 off-color socket. So that math is definitively good.
Replicating 1000 times the 400 trials of the OP gives a probability of 0.002 of having 0 off-color so that math is also definitively good.


However, using the OP's data, the distribution of color is more like B=0.8358, G=0.0854, R=0.0788

Using 0.84-0.08-0.08 instead, the probability shifts to 0.0067978 i.e 1/147 and the probability of having 0 off-color in 400 chromatics is about 0.065 (32 times the previous value)

More data would give more precision but 0.84-0.08-0.08 and 1-in-15 bad luck seems more likely that 0.8-0.1-0.1 and 1-in-500 bad luck

if you use the same base asumtions ofc the math fits ^^

the adjusted probabilitys are certainly interesting - but even there any 4 off-colors would be 1 in 144. So OP was either very unlucky or there is a bias towards fewer off-colors.
Certainly needs more data.

@tikitaki: if you pay me you can call everything I do wrong for all I care :P
Anyway with only OP's data it seems the current accepted model for calculating color-chances is wrong. Or at least the chances have to be adjusted to 0.84 0.08 0.08.
However for 4-socket items the accepted systems seems to fit, so if there is a bias against 4+off-colors there are going to die a ton of chromatics to proof that.

Not *that* bad, still a 6% chance of no 4-off-colors.

"

Navac wrote:

Daggers are not primarily melee, they are crit based. They have both melee and spell caster mod pool and are associated with the Shadow tree which is hybrid dex/int.

That doesn't contradict the point I was making about daggers. Socket color bias is based solely on item attribute requirements. Daggers require Int and/or Dex, and that gives them a strong bias against red gems, which is where most melee skills and supports are found. IMO, color bias should not be so extreme that it blocks you from using melee gems on a melee weapon.

"

RogueMage wrote:

"

Navac wrote:

Daggers are not primarily melee, they are crit based. They have both melee and spell caster mod pool and are associated with the Shadow tree which is hybrid dex/int.

That doesn't contradict the point I was making about daggers. Socket color bias is based solely on item attribute requirements. Daggers require Int and/or Dex, and that gives them a strong bias against red gems, which is where most melee skills and supports are found. IMO, color bias should not be so extreme that it blocks you from using melee gems on a melee weapon.

From my count, 'most' is a gross overstatement. :p

Str gems: Cleave, Dominating Blow, Glacial Hammer, Ground Slam, Heavy Strike, Infernal Blow, Leap Slam, Lightning Strike, Shield Charge, Sweep.
Dex gems: Cyclone, Double Strike, Dual Strike, Ele hit, Ethereal Knives (Melee or not? Short range but still ranged; usable with melee), Flicker Strike, Frenzy, Puncture, Viper Strike, Whirling Blades.
Total count: 10:10. 11 once Reaver comes out if it's green.

You're also forgetting that daggers roll mixed stats on weapons as well. They're not pure damage weapons - they can roll spell damage, spell crit, etc. Also the whole using your main attack on a one handed weapon is a terrible idea. Only 2 supports? My grandma has more supports than that. :p


edit - GUYS, remember this is Beta Feedback.. Whether you agree or disagree with something, say so and explain your reasoning. Don't say "because it is" as a defense. If the founding fathers were just like "Bah it's just the way it is. Praise King Henry!", well, we'd all be eating crumpets and drinking cups of tea.

"

RenoR wrote:

I heavily doubt that GGG thinks: "Oh, 6 blue sockets on a astral plate is OP, so lets make the chance 0,000005%"
Nonsense.

I have also seen this argument recently. People saying that rolling off colors should be hard because its supposedly better or something. There is nothing inherently OP about off colors. They just provide more diversity. I really don't see why it should be so hard to get them.

I remember having a sick (read: pretty shitty) evasion chest in CB. It was 5L5S linked and I needed gggrr. Not a very hard combo to get really. What was really cool though was I got bbbgr on the 4th chrom. Astronomically small odds of this happening, yet it did. I still have a ss cause I was pretty new and I thought that was so cool.

"

iamstryker wrote:

I have also seen this argument recently. People saying that rolling off colors should be hard because its supposedly better or something. There is nothing inherently OP about off colors.

This is a RPG (role playing game). Things like this make it thematic and interesting.

"

iamstryker wrote:

They just provide more diversity. I really don't see why it should be so hard to get them.

It really isn't that hard. Unless you need 6 blues on an armor piece you will be able to get it pretty easily. Sometimes it takes a very long time, but that is how RNG works. If you don't like RNG you probably will get frustrated very quickly with a game that is almost entirely RNG based.