[V] Question on Extracting Monster Power

[Pete Mack]

Ancalagon is more dangerous than Sauron... assuming you decide to fight him.

Storm Giants are not more dangerous than quasits: you can teleport Storm giants away. It doesn't do much good with quasits.

And to answer the original question: Slay Orc has negative value, since a DROP_GREAT monster has a finite chance of dropping it rather than something actually useful.

[Magnate]

Storm Giants are not more dangerous than quasits: you can teleport Storm giants away. It doesn't do much good with quasits.

Am I the only one who doesn't understand this?

[buzzkill]

Am I the only one who doesn't understand this?

Well, I get it (I think). So yes, probably.

[nullfame]

Hmmm... well the quasit can pass walls so she can effectively cover more ground but has the same vision as storm giant so I'm not 100% sure I understand beyond that TO is less effective v. quasit.

That there is a debate tells me maybe the algorithm isn't altogether off. Death quasit is #458 on the list and Storm giant #461 (with Hand druj and Dark elven sorcerer in between them) so it's not like they are so far apart.

[d_m]

I don't have any comment on monster power, but I just wanted to make a public service announcement that the diffs for particular files can indeed be combined into one diff to rule them all.

On Linux, I do that via the following:

cd angband-head
*make some changes to various files*
svn diff > mywork.patch

This creates a patch called "mywork.patch" that has the differences I made to each file. The patches are fairly readable once you get the hang of them so I will often scan it to make sure I didn't accidentally pull in changes to other files that aren't relevant.

Hope this helps.

[Timo Pietilä]

Hmmm... well the quasit can pass walls so she can effectively cover more ground but has the same vision as storm giant so I'm not 100% sure I understand beyond that TO is less effective v. quasit.

Maybe that refers to surprise attack from the wall (invisible, pass wall) where you can't target them with TO.

[Magnate]

Hmmm... well the quasit can pass walls so she can effectively cover more ground but has the same vision as storm giant so I'm not 100% sure I understand beyond that TO is less effective v. quasit.

That there is a debate tells me maybe the algorithm isn't altogether off. Death quasit is #458 on the list and Storm giant #461 (with Hand druj and Dark elven sorcerer in between them) so it's not like they are so far apart.

I agree that the debate is a good sign ... but wow, was that an unfortunate last sentence. Both hand druj and (especially) dark elven sorceror are WAY more dangerous than either death quasit or storm giant.

I have never lost a character to either a storm giant or a death quasit: the former are dangerous way OOD but are avoidable until you can handle them. The latter are hard to avoid but more irritating than really deadly.

I have lost COUNTLESS characters to dark elven sorcerors: they are one of the first +20 speed monsters you meet, pretty tough to kill and with a massive arsenal of nasty spells. I am stunned that they are below #300, let alone #400. (You didn't mean dark elven MAGE, did you?)

Hand druj is the weakest of the three druj, is it? Again I'm surprised it's below #400 - but perhaps it doesn't move? I'm not so sure about this one, as I may be confusing it with nastier druj.

I am intrigued to fire up my dev box and apply your patch. Sadly it's a school night and it looks like I will be working most of the next two weekends ... but one day I will, honest ...

[nullfame]

I'm sorry, I was sorting ascending instead of descending (i.e., Morgoth was 616). Sorting descending Morgoth is #1, Storm giant #155, Death quasit #158.

Hand druj is the least of the three druj. I would have to double-check but I think it is penalized for not moving. In practice I don't find them very dangerous since they are easy to detect and trivial to avoid but when faced with one you may have some problems.

Storm giant > Dark elven sorcerer I'll look in to more. My guess is the sorcerer's speed advantage (1.5 times giant's) is just not enough to make up for the Storm giant's massive melee advantage (48dam*3speed = 144 for sorcerer, 160*2 = 320 for giant) even with magic. So that might even be right.

The real crime all of the above are rated weaker than Death mold, Black pudding, and Azog, King of the Uruk-Hai.

EDIT: math was incorrect for sorcerer melee, fixed now.

[d_m]

It's also true that monster power tries to encode two different things:

1. The chance that the monster will kill you even if you try to avoid it
2. The difficulty in killing the monster if you fight it

While they are harder to avoid, I do think Dark Elven Sorcerers are way easier to kill than Storm Giants.

[Derakon]

The real crime all of the above are rated weaker than Death mold, Black pudding, and Azog, King of the Uruk-Hai.

It sounds to me like you need to apply a major power penalty to nonmagical sessile monsters. Death molds are completely harmless so long as you don't step next to them, and anyone with a radius-2 light source and the "don't cut corners" option set should never accidentally step next to a death mold.

On the flipside, a sessile monster with frequent spellcasting is actually more dangerous than an equivalent motile monster because you have to come to it, instead of letting it come to you. Of course, you can generally avoid it if it's not in the way, but if it is then destroying it can be a major pain.

As for puddings and Azog...yeah, I got nuthin' there.

[nullfame]

It sounds to me like you need to apply a major power penalty to nonmagical sessile monsters.

Indeed. Such a penalty exists. I will look at it again when I contemplate Magnate's ideas further but it probably isn't enough of a penalty. OTOH death mold is twice as fast, does more damage, and has more hitpoints than a storm giant so it will probably be an outlier in any event. If you cut them in half it is still rated higher than a Spirit troll (maybe PASS_WALL is undervalued but we can play that game all day), cutting it in third means it is still more powerful than a Colbran, another legit killer.

As for puddings and Azog...yeah, I got nuthin' there.

FRIENDS and ESCORTS too powerful. Should be fixed if I can implement Magnate's formulaic power idea instead of the constant multiplier we use now.

[Magnate]

Indeed. Such a penalty exists. I will look at it again when I contemplate Magnate's ideas further but it probably isn't enough of a penalty. OTOH death mold is twice as fast, does more damage, and has more hitpoints than a storm giant so it will probably be an outlier in any event. If you cut them in half it is still rated higher than a Spirit troll (maybe PASS_WALL is undervalued but we can play that game all day), cutting it in third means it is still more powerful than a Colbran, another legit killer.



FRIENDS and ESCORTS too powerful. Should be fixed if I can implement Magnate's formulaic power idea instead of the constant multiplier we use now.

But isn't power still tied to native depth? Azog is one of the toughest monsters there is at native depth, so I'm quite happy with his rating. This is my problem with storm giants really - yes they are tougher than DE sorcs, but not relative to their native depths. DESs appear well before they are easy to kill, storm giants usually well after.

Death molds are a special case - let's not worry about them.

I think the issue may be that melee contributes too much to power and spells not enough. But that's just a hunch which needs further exploration.

Nullfame - since I'm working all this weekend I decided to take tonight off and update my dev box. With luck I will try your patches in the next few hours.

[nullfame]

diffs for particular files can indeed be combined into one diff to rule them all.

Thanks for this tip. Here it is as one (unified) patch.

[Magnate]

Thanks for this tip. Here it is as one (unified) patch.

Excellent - applied and compiled successfully (one warning about C++-style comments but nothing major). Will have a tinker tonight.

[nullfame]

What is the difference between ESCORT and ESCORTS?? And FRIENDS, for that matter?

I had to look this up while I was tinkering but someone more familiar can correct me...

FRIENDS means when generated the monster is generated with several others of the same type around it if there's space (hounds, orcs). FRIEND is an unused tag, though some variants use it for monsters that appear in pairs or the like. The power calculation treats it this way though I could make a case it shouldn't (a monster.txt entry can carry an errant friend flag which does nothing but increase power rating).

ESCORT means the monster when generated has several other creatures of the same symbol around if there's space (orc captians). ESCORTS means the same thing but as each escort is placed it is treated as if it has the FRIENDS flag. That is why kobolds and yeeks don't come as packs but their uniques have massive escorts (ESCORTS flag) compared to Ancalagon or Gabriel who just have ESCORT. The unique balrogs have ESCORTS.

But isn't power still tied to native depth? Azog is one of the toughest monsters there is at native depth, so I'm quite happy with his rating.

Not in my current patch, though I still save the power rating relative to depth. If my analysis of the code is correct it is still is needed to calculate rarity when the rebalance flag is set. I have probably broken that but it would be a one-line fix. If I have an hour or two tonight I will take a look at this and a few other things.

I think the issue may be that melee contributes too much to power and spells not enough. But that's just a hunch which needs further exploration.

I think this is right, though I sought to partially address this by adding spell and melee power. I believe a lingering problem is spell power is scaled by frequency but melee is not. E.g., a 1_IN_3 monster's spell power is multiplied by 1/3 but her melee is not multiplied by 2/3 and it probably should.