Randarts...

[Derakon]

That's something that bugs me about the power ratings, actually. They're independent of the character being played. Half the classes don't care about INT; two-thirds of the classes don't care about WIS. But for the classes that do care, those stats are hugely important! In other words, player power ratings for artifacts don't match the assigned power ratings because the player is giving ratings based on the practical utility of the item for the character in question, while the game is trying to generate a power rating that's independent of character.

What do you think of the idea of making randart generation dependent on the character? In other words, you'd have a set of power-rating tweaks based on class/race: a priest will value +WIS more than a warrior will, but is less interested in +INT than a mage. Most powers would be of equal utility for all classes, so it's only a few outliers that would need this. I'll grant this is really weird, but given the potential for randarts to sometimes wildly favor one class over another (unlike standarts, which tend to be less extreme), I think it'd serve as an effective nerf of these only-useful-to-some randarts. It'd also make priest randart games a bit less of a gamble, since non-blessed sharp weapons would get a power rating nerf.

Heck, you could base the power ratings on the character's base stats, too. "Oh, I see you're a gnome mage. Whelp, this +6 CON shield is now too powerful; guess I'll have to reroll it."

[fizzix]

What do you think of the idea of making randart generation dependent on the character?

I don't like this approach if only for your gnome mage example. It would suck to play a mage and have all the randarts get huge Wis bonuses just because the power is then low enough.

@Magnate

214 is low. Boots with nothing but +8 speed should be about 250. I think you're oscillating around the correct point with how much to value speed. You've overshot on the latest adjustment undervaluing it.

@rest of thread:

Something does need to be done with fudging the Int and Wis power. My last game had tons of huge Wis bonuses but no big Int bonuses. That seems wrong. I think it should be split more 50-50. This is begging for multiple pvals though, so we may have to wait for that. Then we can soft cap stat bonuses to some reasonable level, like +5, and still allow speed to be +10.

In reality I'd like to see Int and Wis combined and Charisma removed altogether, but that's fantasyland.

[Magnate]

I don't like this approach if only for your gnome mage example. It would suck to play a mage and have all the randarts get huge Wis bonuses just because the power is then low enough.

@Magnate

214 is low. Boots with nothing but +8 speed should be about 250. I think you're oscillating around the correct point with how much to value speed. You've overshot on the latest adjustment undervaluing it.

@rest of thread:

Something does need to be done with fudging the Int and Wis power. My last game had tons of huge Wis bonuses but no big Int bonuses. That seems wrong. I think it should be split more 50-50. This is begging for multiple pvals though, so we may have to wait for that. Then we can soft cap stat bonuses to some reasonable level, like +5, and still allow speed to be +10.

In reality I'd like to see Int and Wis combined and Charisma removed altogether, but that's fantasyland.

Wow, lots to pick up here.

First, I haven't actually changed the power ratings for speed for a very long time. My last change was to make speed more common on boots in higher quantities, because there were too many randart sets with no boots better than +4 speed, and too many winners wearing straight +9 speed boots instead of artifacts. I may have gone too far, but it's too early to tell whether fizzix's set is an outlier.

Secondly, I don't have a clear idea of the difference between 214 and 250, so I'm curious as to how anyone else can. I don't mean to be rude, I just don't understand. The only difference I can think of is price, and speed items have always seemed underpriced because of the psychological effect of their previous massive prices. That's not to say that they're now correctly priced, of course - it's possible that speed is still undervalued so I'm happy to revisit that with the next set of changes (real soon now ...).

Thirdly, I'm open to the idea of allowing object_power to make reference to p_ptr->pclass in its calculations. It's a complete change from the current paradigm, and will further break comparability with standarts, but neither of those issues is insurmountable. Grateful for further views on this. (I don't think the gnome mage example would be a problem, as referring to class would actually help *reduce* the number of junk randarts.)

Finally, I'm really thinking about multiple pvals quite seriously now, as it's something I know Takkaria would support. I've got two quite big things to do first (pref files and fractional blows), but then I think a return to artifacts would be helpful. I want to do a complete rewrite of both obj-power.c and randart.c, and both fractional blows and multiple pvals are necessary before I can do that properly. (So are three other things: getting rid of the limitations around "special" artifacts, sorting out the new curses, and revising the monster power algorithm. Fortunately someone (nullfame, IIRC) has taken on the latter.)

Ho hum. Shame I have to work, really.

[EpicMan]

The problem with tying the powers of stat boosts to player class is that priests would find lots more artifacts with int bonuses while mages would see lots of useless wis bonuses and fewer int bonuses. Even in the most powerful arts, the opposite stat would be more likely to be added because it is cheaper and more could be fit in than the higher-valued main stat.

The upshot of this is that the best mage randarts would be found in priest/paladin games and vice-versa.

Then again, the RNG has a tendency to be sadistic, so perhaps this would fit right in after all.

[Magnate]

The problem with tying the powers of stat boosts to player class is that priests would find lots more artifacts with int bonuses while mages would see lots of useless wis bonuses and fewer int bonuses. Even in the most powerful arts, the opposite stat would be more likely to be added because it is cheaper and more could be fit in than the higher-valued main stat.

The upshot of this is that the best mage randarts would be found in priest/paladin games and vice-versa.

Then again, the RNG has a tendency to be sadistic, so perhaps this would fit right in after all.

I am pretty confident that we can code around that problem, if we choose to introduce class-specific randart generation.

[camlost]

I am pretty confident that we can code around that problem, if we choose to introduce class-specific randart generation.

Or you can just add a 50% chance of swapping Int for Wis or vice versa so that the player is happy. At the end of the day, randarts that are useful are more important than strict balance.

[Tiburon Silverflame]

A side comment on the effect of Int/Wis...

Even for the classes where it's very important, one thing to note is, the score in question has a very high maximum before equipment...typically it's gonna be at least 18/130, and we could probably settle on, say, 18/150 as a 'typical' max. Therefore, there's absolutely no difference between +7 and +9...and if you've got any other desirable piece of gear that gives, say, +3...there's no difference between +4 and +9.

In the latter case, sure, there could be a different equipment set where it matters, but the value and therefore price should IMO be based on the more common scenario...and most of the time, especially for, say, DL 50+ artifacts/items, you'll have something else giving some decent boost.

That suggests, if it's not being done now...the value of a stat booster probably should be capped, or at least start rolling off drastically, at/after some point. +4? +5? I'm not sure.

Note, too, that this holds for any stat. +9 Con sounds awesome, but again, you're getting Con from more than one source...I often find that my endgame Con would show as 18/270 or the like.

[Magnate]

A side comment on the effect of Int/Wis...

Even for the classes where it's very important, one thing to note is, the score in question has a very high maximum before equipment...typically it's gonna be at least 18/130, and we could probably settle on, say, 18/150 as a 'typical' max. Therefore, there's absolutely no difference between +7 and +9...and if you've got any other desirable piece of gear that gives, say, +3...there's no difference between +4 and +9.

In the latter case, sure, there could be a different equipment set where it matters, but the value and therefore price should IMO be based on the more common scenario...and most of the time, especially for, say, DL 50+ artifacts/items, you'll have something else giving some decent boost.

That suggests, if it's not being done now...the value of a stat booster probably should be capped, or at least start rolling off drastically, at/after some point. +4? +5? I'm not sure.

Note, too, that this holds for any stat. +9 Con sounds awesome, but again, you're getting Con from more than one source...I often find that my endgame Con would show as 18/270 or the like.

Interesting. I have exactly the opposite view: the value of +8 or +9 is precisely that you *don't* need any other item to boost that stat - making the choices when combining items considerably easier.

[Derakon]

I'm actually not so much interested in the INT/WIS conundrum (which standarts "solve" by making it so that most artifacts that boost one also boost the other) as I am in the blessed weapon issue. Nobody except for priests cares about the blessed flag, but it is seriously important for priests (in the late game, you can sometimes get away with a sharp weapon if you don't mind the failure rate increase on your higher spells).

Then again, we might want to solve this by simply getting rid of the sharp-weapon penalty entirely. Give them (and by extension, paladins) the same gloves penalty all the other casters have instead.

[Tiburon Silverflame]

Magnate, your point's valid...you don't *have* to look for anything else with an Int boost, when you have something giving +8. BUT, you probably find stuff with key things that you want, that *also* has a nice Int boost as an incidental aspect. While this won't be true all the time, I suspect it will be true most of the time. Look at your late-game gear; how many of those late-game items are stat boosters, and how many of your stats are being boosted by at least 2 items? Then, on the assumption that you didn't need anything else to boost Int (or whatever)...how many of those items would you swap out for something else, strictly based on the fact that the Int bonus is now meaningless?

@Derakon: IMO the priest's edged-weapon rule is awful. It's a carryover of a half-assed 40 year old notion, that has been steadily eroded ever since Gygax's whims were no longer allowed to dominate. And it's *by far* the most restrictive rule...arcanists can use +Dex OR FA gloves, and those exist as ego items and standarts both. But an ego weapon that's Blessed, only has something else if it's a Holy Avenger, and damn few higher-end standarts are Blessed.

And yes, I *much* prefer the notion that Magic Is Magic. I hate "divine magic" or "arcane magic" or most of the artificial distinctions. If gloves interfere with casting, do it for *everyone*. If a wizard can't wear metal armor because it interferes, then neither can a priest.

[Timo Pietilä]

The problem with tying the powers of stat boosts to player class is that priests would find lots more artifacts with int bonuses while mages would see lots of useless wis bonuses and fewer int bonuses. Even in the most powerful arts, the opposite stat would be more likely to be added because it is cheaper and more could be fit in than the higher-valued main stat.

The upshot of this is that the best mage randarts would be found in priest/paladin games and vice-versa.

Then again, the RNG has a tendency to be sadistic, so perhaps this would fit right in after all.

I'm thinking that randart code should take a small peek what has already been created before creating another one, so that items do not get too similar. A bit more not-so-random set of random artifacts? Make it create only a couple sets of boots with very high speed bonus and then stop creating them because they are there already. Same with WIS and INT bonuses, not base them on class (because like you said it would really suck to get INT-bonuses with priest, and WIS with mage).

Give enough variability that they can still be called "randarts", but not allow 10 sets of boots with speed bonus. And after creation make sure that there is at least one "great" one for each equipment slot.

Might be quite hard to code though.

[Magnate]

I'm thinking that randart code should take a small peek what has already been created before creating another one, so that items do not get too similar. A bit more not-so-random set of random artifacts? Make it create only a couple sets of boots with very high speed bonus and then stop creating them because they are there already. Same with WIS and INT bonuses, not base them on class (because like you said it would really suck to get INT-bonuses with priest, and WIS with mage).

Give enough variability that they can still be called "randarts", but not allow 10 sets of boots with speed bonus. And after creation make sure that there is at least one "great" one for each equipment slot.

Might be quite hard to code though.

Funnily enough I was thinking along exactly these lines in response to Tiburon's last post: yes I do see that one is likely to have incidental stat bonuses on items one wants to wear anyway - we could in fact check to see if that's true (i.e. look at what is available) when determining the value of the +8!

I don't think it would be too hard to code. I've been thinking for a long time that randart.c needs to maintain a stats array, so it would be easy to peek into that during generation.

I also quite like the idea of calling them semirandarts. True randomness does lead to some undesirable results ...

[Derakon]

Call them prandarts -- pseudo-random artifacts.

[PowerDiver]

I also quite like the idea of calling them semirandarts. True randomness does lead to some undesirable results ...

It is feasible to win without any artifacts. Is a non-random approach really worth the effort?

[Tiburon Silverflame]

I don't know the steps involved in creating a randart for sure, but from comments, it seems like it's

a) create randart
b) determine power, and therefore depth/rarity

IF this is the case, might it not make sense to work this a bit differently? I'm thinking, plan the randart set based on depth. START by assigning depth, or a small range for depth, and deriving a power range from that. Then build the artifacts to the appropriate power.

I think this would be complementary to keeping an internal table of all artifacts created so far. As to the difficulty...the problem may be that it's a multi-dimensional problem. One might have a relatively low-power artifact that is the only source of, say, rConf, or maybe more likely, there are other artifacts that have rConf, but they all get massively trumped by others in their item class.

Finally, of course, depth and rarity have to play a big role, because that impacts the notion of 'available.' We all look over the randart posts and drool at The Big One I Didn't Find.