Bah, forgot to change thread title. I started this thinking that crafted gear was inferior to looted. Now I'm just thinking that magic gear has a higher chance of good rolls than rare.
Ok, guys. I need some brainpower here.
Background
I've been wondering about how the game decides to roll affixes. It's a pretty complicated pursuit do the vast qty of affixes out there, so I wanted to start small and build up. I decided the easiest place to start would be single resists because they are very common. There were three questions I originally asked:
1 - Is the determination of which rank of affix to give completely random or is it skewed? For example, if the item is capable of rolling rank 52, 58, 60, 61, and 62, do each appear the same % or no?.
2 - Once a rank is determined, is the actual roll inside that min and max of that rank rolled randomly or is it skewed.
3 - Is the determination of the rank related to Act at all.
We started a project to log every piece of gear that has single resist on it. Basically, if you found any piece of gear that had a single resist, you'd just put the ilvl and resist amount into a spreadhsheet and we'd keep track of which ranks had been selected.
Looted: https://docs.google....FXWmN5YkE#gid=1
Pulling ilvl62-only pieces from the gear we looted and screen-capped:
Samples: 216
Samples that rolled Max Rank: 130 (60.2%)
Range with 95% confidence: 54% - 67%
Crafted:
Samples: 780
Samples that rolled Max Rank: 306 (39.2%)
Ranged with 95% confidence: 36% - 43%
So... I thought at first that this was indicating that crafted gear was not as good as looted gear. However, I then remembered that one of our contributors had been flagging rare drops vs magic. I went back and looked at just his rare drops, and THEY were ~40%, too. We're going to restart the research, tracking rare vs magic, but I feel strongly that Blizz gives magic items a better chance to get high rolls than rare items.
If anyone would like to help, I'd appreciate the additional datapoints.
I've setup a new sheet to track it, but ignore the charts for now because I haven't finished fixing them yet.
Damn.. interesting stuff. Have you seen anything about magic vs. rare affix quality discussed before?
The only thing I recall is that guy who said that magic items always have one max-rank affix - which is what prompted me to mark magic & rare separately on your spreadsheet, but which was swiftly disproved by a screenshot.
A worrying possibility is, of course, that 4-affix, 5-affix and 6-affix rares have different calculations. e.g. it could be something like the first affix generated has full chances, then each one after that is rolled from a worse table or penalized roll?
Damn.. interesting stuff. Have you seen anything about magic vs. rare affix quality discussed before?
The only thing I recall is that guy who said that magic items always have one max-rank affix - which is what prompted me to mark magic & rare separately on your spreadsheet, but which was swiftly disproved by a screenshot.
A worrying possibility is, of course, that 4-affix, 5-affix and 6-affix rares have different calculations. e.g. it could be something like the first affix generated has full chances, then each one after that is rolled from a worse table or penalized roll?
I believe there was something like that for legendaries, but I never heard anything about that for rare or magic items. I think the way it may be working is that the top rank has x% to be rolled. Failing that, all other ranks have equal chance. I think x% may be 60% for magic gear and 40% for rare. Will have to get more data, though.
that's a nice result. In the crafted gear, can you see whether items can have 5 or 6 max rank affixes? It might be difficult because of base stats adding up in a weird way, but maybe you can find something with single res + all res + life regen + ...
The crafted gear was all 6 affix.... perhaps I should also differentiate between 1-affix and 2-affix magic items.
Yep. I've set up the spreadsheet now to differentiate between 1 affix, 2 affix, and rare. We can later compare the "rare" to 6 affix and see if there's a difference, and we can compare 1 and 2 affix to themselves and to the rares. Unfortunately, the more we break up the data, the more datapoints we need, but I have a couple other people contributing and hopefully that number will continue to grow. I get about 20-30 on short runs, so it doesn't take long for a handful of people to come up with a couple thousand drops.
Yep. I've set up the spreadsheet now to differentiate between 1 affix, 2 affix, and rare. We can later compare the "rare" to 6 affix and see if there's a difference, and we can compare 1 and 2 affix to themselves and to the rares. Unfortunately, the more we break up the data, the more datapoints we need, but I have a couple other people contributing and hopefully that number will continue to grow. I get about 20-30 on short runs, so it doesn't take long for a handful of people to come up with a couple thousand drops.
Unfortunately, as I discovered from the run I just did, you really don't get very many 1-affix magic drops once you've got some NV stacks up. I just got eight blues with single resist - all of them 2-affix. Hardly saw any 1-affix items amongst all the other blues, either.
Yep. I've set up the spreadsheet now to differentiate between 1 affix, 2 affix, and rare. We can later compare the "rare" to 6 affix and see if there's a difference, and we can compare 1 and 2 affix to themselves and to the rares. Unfortunately, the more we break up the data, the more datapoints we need, but I have a couple other people contributing and hopefully that number will continue to grow. I get about 20-30 on short runs, so it doesn't take long for a handful of people to come up with a couple thousand drops.
Unfortunately, as I discovered from the run I just did, you really don't get very many 1-affix magic drops once you've got some NV stacks up. I just got eight blues with single resist - all of them 2-affix. Hardly saw any 1-affix items amongst all the other blues, either.
Yeah, that's gonna take a while. I'm finding same thing.
Ok, guys. I need some brainpower here.
Background
I've been wondering about how the game decides to roll affixes. It's a pretty complicated pursuit do the vast qty of affixes out there, so I wanted to start small and build up. I decided the easiest place to start would be single resists because they are very common. There were three questions I originally asked:
1 - Is the determination of which rank of affix to give completely random or is it skewed? For example, if the item is capable of rolling rank 52, 58, 60, 61, and 62, do each appear the same % or no?.
2 - Once a rank is determined, is the actual roll inside that min and max of that rank rolled randomly or is it skewed.
3 - Is the determination of the rank related to Act at all.
We started a project to log every piece of gear that has single resist on it. Basically, if you found any piece of gear that had a single resist, you'd just put the ilvl and resist amount into a spreadhsheet and we'd keep track of which ranks had been selected.
I also pulled data out of these threads:
http://www.diablofan...ula-to-compare/
http://www.diablofan...-craft-results/
http://www.diablofan...o-compare-them/
I happened to compare the results today.
Results:
Looted:
https://docs.google....FXWmN5YkE#gid=1
Pulling ilvl62-only pieces from the gear we looted and screen-capped:
Samples: 216
Samples that rolled Max Rank: 130 (60.2%)
Range with 95% confidence: 54% - 67%
Crafted:
Samples: 780
Samples that rolled Max Rank: 306 (39.2%)
Ranged with 95% confidence: 36% - 43%
So... I thought at first that this was indicating that crafted gear was not as good as looted gear. However, I then remembered that one of our contributors had been flagging rare drops vs magic. I went back and looked at just his rare drops, and THEY were ~40%, too. We're going to restart the research, tracking rare vs magic, but I feel strongly that Blizz gives magic items a better chance to get high rolls than rare items.
If anyone would like to help, I'd appreciate the additional datapoints.
I've setup a new sheet to track it, but ignore the charts for now because I haven't finished fixing them yet.
https://docs.google....RYUkF4amc#gid=9
The only thing I recall is that guy who said that magic items always have one max-rank affix - which is what prompted me to mark magic & rare separately on your spreadsheet, but which was swiftly disproved by a screenshot.
A worrying possibility is, of course, that 4-affix, 5-affix and 6-affix rares have different calculations. e.g. it could be something like the first affix generated has full chances, then each one after that is rolled from a worse table or penalized roll?
I believe there was something like that for legendaries, but I never heard anything about that for rare or magic items. I think the way it may be working is that the top rank has x% to be rolled. Failing that, all other ranks have equal chance. I think x% may be 60% for magic gear and 40% for rare. Will have to get more data, though.
The crafted gear was all 6 affix.... perhaps I should also differentiate between 1-affix and 2-affix magic items.
That could be illustrative, and unlike 4/5/6-affix rares, it's easy to distinguish between them.
Unfortunately, as I discovered from the run I just did, you really don't get very many 1-affix magic drops once you've got some NV stacks up. I just got eight blues with single resist - all of them 2-affix. Hardly saw any 1-affix items amongst all the other blues, either.
What you probably saw was 4 mods only, so one of them must have overlapped (like you got str, vit, and str+vit)
Yeah, that's gonna take a while. I'm finding same thing.