Talk:Attribute

How attributes are actually calculated:

• Take the 7 XXX_ATT_RANGE values from the raws as A:B:C:D:E:F:G
• Pick a random number N between 0 and 11999, inclusive
  • For N in 0-1999, calculate X = (1999 - N) * A and Y = (N - 0) * B
  • For N in 2000-3999, calculate X = (3999 - N) * B and Y = (N - 2000) * C
  • For N in 4000-5999, calculate X = (5999 - N) * C and Y = (N - 4000) * D
  • For N in 6000-7999, calculate X = (7999 - N) * D and Y = (N - 6000) * E
  • For N in 8000-9999, calculate X = (9999 - N) * E and Y = (N - 8000) * F
  • For N in 10000-11999, calculate X = (11999 - N) * F and Y = (N - 10000) * G
• The attribute value itself is initialized to (X + Y) / 1999
• The attribute's maximum value is set to either (value * XXX_ATT_CAP_PERC) / 100 or value + D, whichever is higher

Exactly how this should be translated into a readable form for the article, I'm not sure, but hopefully somebody will be able to manage. --Quietust (talk) 20:22, 30 May 2025 (UTC)

What's the distribution of N? I've understood the attribute range system as being a bunch of linear buckets that approximate a normal distribution. This sounds like it might actually be distributed normally? DPhKraken (talk) 00:17, 31 May 2025 (UTC)

N is calculated as trandom(12000), so every result should have (more or less) equal probability. Basically, each of the 6 buckets has equal probability (i.e. a 1/6 chance for each one), and the values you get within those buckets are also equal probability (i.e. within the F-G bucket, you have an approximately equal chance of rolling the lowest value, the highest value, or any value in between). I ran a script to calculate all of the probabilities, and the very first entry in each bucket ended up having a slightly higher change (e.g. 11/12000 instead of 10/12000) and the very last entry had a slightly lower chance (e.g. 9/12000 instead of 10/12000), but that's fairly insignificant overall. --Quietust (talk) 00:30, 31 May 2025 (UTC)