[1.0.4] Reduce Howl tactical value to 1 or 2

[jenx]

The warg talent Howl has ATTACK = 3 for tactical ai.

This is too high, and results in rares/boss wargs using howl before other much more damaging attacks.

I suggest it be changed to 1 or 2.

[Doctornull]

IMHO a wolf howling before it attacks is quite thematic, even if it is sub-optimal on the part of the wolf.

[jenx]

that's true. perhaps that explains DG's logic on this one !

[lukep]

There's a whole bunch of wonky AI values. If you haven't yet, check out my abandoned addon trying to do much the same for tactical AI (here).

[jenx]

That's very helpful lukep. I'll look into it

[jenx]

There's a whole bunch of wonky AI values. If you haven't yet, check out my abandoned addon trying to do much the same for tactical AI (here).

heh lukep - your addon still works for 1.0.4, and is better than the one I've been developing. That's good (except that I wasted lots of time!).

Are you planning to develop it further, and to add the new oozemancer talents etc?

Or are you happy for me to tweak it and release an update?

[lukep]

I don't plan on updating it anymore, so if you want to tweak it/take it over, that would be great. I think that I showed most of my reasoning, but I can (probably) answer any questions you have about it.

EDIT: if you haven't yet, check out this post too.

[jenx]

Thanks lupkep. I"ve already started tweaking your code. Several of the improvements I had made can be integrated into your approach.

One line though puzzles me totally:

Code: Select all

local effectiveness = ((act:canBe(damtype) and 1 or 0) + (act:canBe(damtype) and 1 or 0) + (act:canBe(damtype) and 1 or 0) + (act:canBe(damtype) and 1 or 0) + (act:canBe(damtype) and 1 or 0) + (act:canBe(damtype) and 1 or 0) + (act:canBe(damtype) and 1 or 0) + (act:canBe(damtype) and 1 or 0) + (act:canBe(damtype) and 1 or 0) + (act:canBe(damtype) and 1 or 0))/10

can you explain?

[lukep]

That replaces the old, binary check for status effects. It's a bad kludge, but what it tries to do is get a better value for immunities. In the old system, if the target had 60% immunity (let's say to stuns), 6 turns out of 10, the talent would have a DISABLE rating of 0, and 4 turns out of 10 it would be 1 (or whatever it is).

What that line does is change it to always have a value of (roughly) 0.4 in that situation.

I had planned on replacing it with a real check of the target's immunity, but IIRC there were complications with that.

[jenx]

That replaces the old, binary check for status effects. It's a bad kludge, but what it tries to do is get a better value for immunities. In the old system, if the target had 60% immunity (let's say to stuns), 6 turns out of 10, the talent would have a DISABLE rating of 0, and 4 turns out of 10 it would be 1 (or whatever it is).

What that line does is change it to always have a value of (roughly) 0.4 in that situation.

I had planned on replacing it with a real check of the target's immunity, but IIRC there were complications with that.

I still don't quite get it. It looks like the same calculation 10 times added, then divided by 10, so wouldn't that just be the calculation by itself?

[lukep]

The old system was like this, while the one that replaces it is like this(except divided by 10), so not quite the same as before.

[Doctornull]

lukep - it looks like you've basically got a system where you take a hidden variable of 0.60, measure that variable as a 1 or 0 based on a weighted dice roll, then put that dice roller through a stochastic process to get back the variable 0.60.

Why not just pull the 0.60 out and use it directly?

Is it for round-to-round utility variance or something clever like that?

[jenx]

lukep - it looks like you've basically got a system where you take a hidden variable of 0.60, measure that variable as a 1 or 0 based on a weighted dice roll, then put that dice roller through a stochastic process to get back the variable 0.60.

Why not just pull the 0.60 out and use it directly?

Is it for round-to-round utility variance or something clever like that?

I've been debugging it, and the value is usually 1 or 0.5. So I'm not even sure it is working as intended.

[lukep]

lukep - it looks like you've basically got a system where you take a hidden variable of 0.60, measure that variable as a 1 or 0 based on a weighted dice roll, then put that dice roller through a stochastic process to get back the variable 0.60.

Yup, that's the goal.

Why not just pull the 0.60 out and use it directly?

It's been a while, but I seem to recall it having problems when dealing with damage types like FIRE or WEAPON, as opposed to STUN or CONFUSION. It's probably fixable, but it was just a quick hack.

I've been debugging it, and the value is usually 1 or 0.5. So I'm not even sure it is working as intended.

Hmm. I didn't debug that part at all, so it could very easily be broken.

[jenx]

With the release of my new better tactical ai addon (http://forums.te4.org/viewtopic.php?f=50&t=38820), the issues in this thread have been addressed. Enjoy!