Removed artifacts used as randart templates

[jens]

Well, I guess it's no big deal, but it's wrong :-)

If you start a randarts game, then check the log you will see that the Palantír, which has been removed from standart games, is still used when creating randarts.

[Magnate]

Well, I guess it's no big deal, but it's wrong :-)

If you start a randarts game, then check the log you will see that the Palantír, which has been removed from standart games, is still used when creating randarts.

Why is this wrong? I rather like it.

[jens]

Well it might be fun, but in the long run it can unbalance things. I mean, when me have made some crazy changes to artifacts, and left lots of halfway deleted ones behind, and rebalanced the game to fit the new set of standart, then the randarts game will be unbalanced. And also for the newbies who don't know about there ever having been an artifact called Palantír, will be confused when they start checking the rantarts log...

I know, the reasons are somewhat unlikely scenarios, but they do point the way to sound practises.

[Magnate]

Well it might be fun, but in the long run it can unbalance things. I mean, when me have made some crazy changes to artifacts, and left lots of halfway deleted ones behind, and rebalanced the game to fit the new set of standart, then the randarts game will be unbalanced. And also for the newbies who don't know about there ever having been an artifact called Palantír, will be confused when they start checking the rantarts log...

I know, the reasons are somewhat unlikely scenarios, but they do point the way to sound practises.

You are conflating a couple of issues here.

Not wanting to Confuse newbies is a legitimate aim - but newbies by definition aren't playing with randarts. If they know enough to decide to turn on randarts then they know enough about roguelikes and/or CRPGs in general, even if they're new to Angband.

Balance is a whole separate issue. When I started work on randarts ten or twelve years ago, with a guy called Chris Robertson, we had a very strong principle of parity with standarts. It was really important to us that randart games could be compared like-for-like with standart games.

Over the past decade that principle has been steadily eroded, for several reasons. The main one is that it's just not possible, and it constrains good randart design to attempt it. There can be no replication of the way in which certain combinations of standarts fit together for endgame kit - all we can do is attempt to create the same power curve (i.e. same findability of power in randarts over the dungeon). That's predicated on the assumption that you can measure the power of an object, which is disputed.

The second reason is that it's more fun to allow permutations of stuff which don't occur on standarts (yet).

Another reason is that randarts, by definition, will sometimes be rubbish (e.g. weapons with <+9 to_d). So we have to either not generate them or adjust them, both of which breaks the parity.

In summary, I've become quite relaxed about the parity issue, and more focused on making the randart generator come up with interesting but individually balanced randarts. I will rely on the stats module to tell me that the overall power curve is a decent match for standarts - beyond that, I don't think it's necessary to limit randarts to the standart templates.

[Nomad]

Oddly enough, I agree in general that removed artefacts should be fine to use as templates, but think that the Palantír might be a special case that shouldn't be used.

The reason being that, as a special artefact, it has its own unique tile in tile sets, and while the current ones obviously still have a tile assigned, future sets might not keep support for object types that appear to have been removed from the game.

[Derakon]

The Palantir is also typically unbalanced as a randart because it keeps its old activation for Clairvoyance, but doesn't typically get aggravation. That pretty much immediately makes it better than the standart Palantir regardless of what other abilities it has.

[Magnate]

The Palantir is also typically unbalanced as a randart because it keeps its old activation for Clairvoyance, but doesn't typically get aggravation. That pretty much immediately makes it better than the standart Palantir regardless of what other abilities it has.

But on the other hand it may well not be better than the randart arkenstone ...

@Nomad: you make a good point - we're going to have to come up with some guidelines for tilesets which include this detail.

[jens]

Not wanting to Confuse newbies is a legitimate aim - but newbies by definition aren't playing with randarts. If they know enough to decide to turn on randarts then they know enough about roguelikes and/or CRPGs in general, even if they're new to Angband.

Yeah, I know, even more so if they would notice that a randart is based on an non existing artifact. Wrote it mostly for fun :-)

There can be no replication of the way in which certain combinations of standarts fit together for endgame kit

I do not agree. It might be harder to achieve than what we have today, but it's just a question of finding the right algorithm. Now I have never looked into randart generation at all, so I don't know much about this, but I can think off a couple of approaches that could achieve this.

In summary, I've become quite relaxed about the parity issue, and more focused on making the randart generator come up with interesting but individually balanced randarts. I will rely on the stats module to tell me that the overall power curve is a decent match for standarts - beyond that, I don't think it's necessary to limit randarts to the standart templates.

One thing I was about to suggest in my previous post is this: How about using a list of power numbers, and what else you need as the base for randarts, instead of making complicated calculations for each new set just to achieve the parity that you don't really feel is neccessary any more.

The gains would be that generation of randarts would go quicker, and we would get a bit more control on the base used. The starting point would of course be to once generate this list from standarts, but then we could modify this as we saw fit.

[Magnate]

I do not agree. It might be harder to achieve than what we have today, but it's just a question of finding the right algorithm. Now I have never looked into randart generation at all, so I don't know much about this, but I can think off a couple of approaches that could achieve this.

I don't think we're talking about quite the same thing. My original statement was about reproducing the ways standarts fit together - that's by definition impossible, because the randarts will be different.

I didn't say that it wasn't possible to make randarts that complement each other well - that is something I intend for the next generation of randarts. Being a Linux geek I guess I ought to call this "randarts-ng".

One thing I was about to suggest in my previous post is this: How about using a list of power numbers, and what else you need as the base for randarts, instead of making complicated calculations for each new set just to achieve the parity that you don't really feel is neccessary any more.

The gains would be that generation of randarts would go quicker, and we would get a bit more control on the base used. The starting point would of course be to once generate this list from standarts, but then we could modify this as we saw fit.

I have said quite a few times that I will be completely rewriting the randart generator for a future version. This has taken a lot of prerequisite work refactoring monster spells, breaths/projections, object flags and object power, but that's now all done, so 3.4 is looking enticing.

As you note, my more relaxed attitude to parity means I was indeed thinking along precisely these lines - using a minimum-restriction approach to generation. Quicker randart generation is also a very high priority, because at the moment it slows down stats generation by a factor of about 20 (though I've not tried it with logging off - that would speed it up a lot).