Enemy Action Pattern/Rating

Am I getting it right?

Suppose I have 4 different actions set up for an enemy.

Attack A: Rating 5

Attack B: Rating 4

Attack C: Rating 3

Attack D: Rating 2

The tooltip says that "the one within 2 rating points of the standard will be used".

1 point away: 2/3 of the time.

2 points away: 1/3 of the time.

Does that mean that Attack A will be used at a chance of 50%, while Attack B will be used at a chance of 50%*2/3 (about 33%) and Attack C at a chance of 50%*1/3 (about 17%)?

Or how does it work?

In that specific case your % values are correct - but the calculation is a bit more complex when several attacks have the same ratings, because the calculations don't use a 50% base like you did and that base only works into the same numbers in this special case with three skills.

Ah, okay. So ... let's see ...

Attack A: 5

Attack B: 5

Attack C: 4

Attack D: 3

Attack E: 3

Then A and B combined are 50%, so 25% each?

And C, D, E combined are also 50%?

Attack C would be ~33% and Attack D and E combined are the ~17% that are left, so ~8.5% each?

no - I don't have the time to seek the code in the engine (that's why I didn't give you a full answer before) but most probably they will use weighting to calculate probability, and that never translates to readable percentages.


If I'm correct, then attacks A and B will have a weight of 3 each, attack C one of 2 and attacks D and E one of 1 each.


That is a total weight of 3+3+2+1+1 = 10 - you're lucky, in this case it does translate to reasonable percentages (it wouldn't have if you had an attack F with rating 3 again, because that would have added to weight 11)


So attack a has 3/10, B has 3/10, C has 2/10 and D and E had both a chance of 1/10.


If that hypothetical Attack F existed, then the results would have been A 3/11, B 3/11, C 2/11 and D,E,F each 1/11.


(this is without any guarantee, I never checked the real coding in the engine, but you can't work with percentages in an open list of selections, and that would be the most effective way of programming in my opinion)


Edit:


And the best proof that your calculations are wrong is probably that in your last example, attack C (33%) would have a higher chance than attacks A and B (25% each), I really doubt that the engine would work that way...

You're right with your last paragraph. And probably also with the rest. :) Thank you!

EDIT:
I'm confused in the other thread they said that the standard (highest rating) would be given 50% so if two skills are standard then they would be 25% each and no sum of rating is considered in the calculation