Smaller levels

[Philip]

Yeah, but then you still have unexpected drolems to deal with. Round a corner and your mage goes poof, pretty much whatever is there.

[fizzix]

I'm pretty sure that I think autoexplore is a bad idea. I was playing tome a bit last night, and realized that it's much more enjoyable to play without autoexplore for the vast majority of levels. And for the ones where it's necessary, it's due to awful level design. The standard maps with rooms and hallways are so much better without autoexplore. You can actually figure out tactical positioning and stuff.

Anyway, the more I think about it, the more I come to the conclusion that autoexplore is a bad idea. The only reason you'd want it is if you have a UI that requires minimal input, and angband is never going to be that.

I'll again bring up my proposal for a "smart run" command. Which goes N steps in a direction, stops at intersections or if monsters are in view, and is smart enough to find room exits if you start in a room. So it's just a minor modification to the run command.

The other reason to want autoexplore is for item collection. We don't have the item spam issues of Tome (100 items, 99 cannot be used by the class you're playing). But we might want to allow the player to select items to automatically pick up better than the current system (just all items, and all items matching inventory). Then you could have a second automated option which essentially is gather all visible items that are marked for pickup.

So basically the squelch menu would have a third setting for "always pick up" You can set it to always pick up weapons, and squelch non artifacts, then test-picking-up all weapons after clearing a vault/pit to find the artifacts is easy.

[bio_hazard]

While general auto-explore may not be so useful in Angband, I'd love a "path to nearest >" command once you've cleared out a level.

Also, I'm curious how different autoexplore would be compared to what the borg does.

[AnonymousHero]

While general auto-explore may not be so useful in Angband, I'd love a "path to nearest >" command once you've cleared out a level.

You'd still be at (however minimal) risk because monsters could spawn post-clear, i.e. before you reach the stairs.

Also, I'm curious how different autoexplore would be compared to what the borg does.

I guess an argument could be made for maintaining the Borg in-tree as a "variant" of auto-explore...

[fizzix]

Go-to is definitely something that should be added. With knowledge that it comes with risks. Even now, with dumb running, it doesn't take that long to traverse a level.

Getting the borg to work again is a significant project. It'll likely need to be rewritten from scratch after the UI refactor. Right now it is a series of successive patches and special cases that is extremely unwieldy.

[Magnate]

This is exactly the same as the squelch argument. I really didn't want to include it - it's a hacky workaround to item generation just not being very good. I suspect the proponents of autoexplore will win, just like squelch did. Someone just has to write a patch and everyone'll demand it.

I've spent a long time thinking about this - five or ten years, on and off - and I respectfully disagree.

Roguelike games are, in large part, about loot.

It seems to me that there are two choices with item generation:

1. You control it very carefully, so that there is no junk.

2. You accept that there will be lots of junk, leavened by the occasional whoop-out-loud item.

Almost all roguelikes opt for the latter option. Squelch is essentially just a UI solution to the visual clutter generated by that option. (In any Diablo clone the ground is always littered with white/non-magical items - some actually allow you to make them invisible, i.e. squelch them.)

For years I reflected on the position quoted above - that squelch would be unnecessary if item generation was working properly. (Leon Marrick, one of my favourite roguelike maintainers, held this view very strongly.)

I finally decided that I don't agree. For item generation to be that tight, you have to do one of two things:

(i) Keep variation so low that every outcome results in something useful. This takes you away from roguelike towards Colossal Cave style gaming, where there are very few items but they're all somehow useful.

(ii) Make the item generator respond to the race/class/skills/level of the character. (Diablo III's first expansion will do this, apparently.) It will probably take me another few years to articulate properly why, but this feels very, very, very wrong to me.

So I don't think there is a way to avoid junk. Once you're at DL98 wielding Ringil, everything that goes in the weapon slot is junk (bar about three - Deathwreaker, Doomcaller, etc.).

So I'm alright with squelch - but I'm with fizzix on autoexplore.

[Derakon]

Also consider that squelch is, functionally, purely an aesthetic streamlining. It hides stuff, but it does not mechanically change how the game is played. You can ignore squelch and the only difference is that items you "destroy" can be un-destroyed later.

Autoexplore has the player taking different actions from normal exploration, and building the game such that autoexplore is feasible requires significant changes in how the game plays even if you don't use it. So I don't think it's anywhere near as "safe" a change to make to the game.

[Scatha]

Magnate: You can avoid too much junk without avoiding junk altogether.

If you have a metric for item quality, here's a quick way to halve the number of drops without reducing useful drops by much:

Each time, you'd generate some items, generate only half as many (50% chance for a single item). Each time you generate an item, generate two and only produce the better one.

That would still let weak items drop right through the game, if the current code allows it, but they get less frequent later, while stronger items are about as frequent as they always were.

You can easily adapt this to a more extreme version, for instance generating one-fifth as many items but having each be the best-of-five. And if there are differential amounts of junk at different points in the game, you could vary that parameter accordingly.

You might want to do this separately for different broad item types (e.g. potions, helms, swords, etc.; or even just weapons, armour, consumables ), so you don't have the effect of e.g. consumables rarely dropping in the late game because the high-end weapons are rated as higher quality.

Edit: Not sure if that was clear, but I agree with Takkaria that squelch is a hacky solution.

[AnonymousHero]

Magnate: You can avoid too much junk without avoiding junk altogether.

Each time, you'd generate some items, generate only half as many (50% chance for a single item). Each time you generate an item, generate two and only produce the better one.

I think the problem is "produce the better one". If you knew how to do that, you'd already be golden -- the problem is the "the better one" may depend on lots and lots of things.

That's not to say that I necessarily agree with Magnate on the impossibility of doing such (or succumbing to a game-specific mechanic). It'll be hella difficult though -- and it would require some sort of "all-items-are-usable-for-all-characters" mechanic which I can't imagine would be easy to balance.

[AnonymousHero]

So I don't think there is a way to avoid junk. Once you're at DL98 wielding Ringil, everything that goes in the weapon slot is junk (bar about three - Deathwreaker, Doomcaller, etc.).

I don't necessarily agree/disagree with you, but allowing randarts and standarts means that you could actually find an artifact better than Ringil. I don't think you need infinite randarts as in e.g. ToME 2.x, but you could generate both a standart set and a randart set at the start of the game -- it'd be interesting to see how such a game plays. You'd probably want to tone overall artifact power down a little, but...

[buzzkill]

All this talk about squelch. How about Smart-squelch, or as it might more aptly be called, PSSS, Pandora-style soft-squelching. Pandora learns what music you like by (you pointing it in a direction and then) what songs you choose to listen to and what songs you skip over. It's more complicated than that, but I think that's the basic theory IINM. In the end, it does a fine-ass job of giving you pretty much only music you want to hear. It's uncanny how well it works. In this same spirit, let the game engine learn what the player considers useful by the tracking the players interaction with the various items.

New game. Start every item with a PSSS rating of 50. A rating of 50 or higher mandates that an item not be squelched. For every point below 50, there is a 2% chance of PSS occurring (instantly in the background upon item generation). If the player uses an item the rating goes up a point. If the player doesn't use the item, the rating goes down a point. Buy an item, rating +5. Sell an item, rating -5 (maybe...). Destroy an item, rating -10.

I call it soft-squelching because unused items are gradually phased out. Just because an items was squelched this time, doesn't necessarily mean it will be squelched next time. This could easily coexist with current hard squelch settings. That's it in a nutshell.

[Patashu]

The main problem with 'smart automatic squelching' is that an item could theoretically be worthless early on, but useful later on (I'll use an example from Sil - Potions of Voice are worthless if you don't have any songs, but you can pick up songs any time in the game you like). Or you could be neglecting to pick up more of an item because you don't need any more of it at the moment (full potions or full ammo for example).

[Therem Harth]

I don't necessarily agree/disagree with you, but allowing randarts and standarts means that you could actually find an artifact better than Ringil. I don't think you need infinite randarts as in e.g. ToME 2.x, but you could generate both a standart set and a randart set at the start of the game -- it'd be interesting to see how such a game plays. You'd probably want to tone overall artifact power down a little, but...

Yes yes yes. I'll also add infinite randarts to Neoband, as soon as I can figure out how in the heck it's done. :/

[AnonymousHero]

Yes yes yes. I'll also add infinite randarts to Neoband, as soon as I can figure out how in the heck it's done. :/

I think maybe you are (or I am!) missing a negation. I don't think I missed any .

The way Angband does it is essentially to "clone" the vanilla standarts and shuffle around attributes (see, primarily, rantart.c), but I really have no idea how to retrofit that into a ToME 2.x codebase at the moment.

[Scatha]

I think the problem is "produce the better one". If you knew how to do that, you'd already be golden -- the problem is the "the better one" may depend on lots and lots of things.

Your quality metric doesn't need to be perfect, so long as you're still generating the lower-quality items sometimes (at a lower frequency). It will end up slightly changing the distribution of items a character gets and uses, but that's okay.

I don't think it's too hard to produce a quality measure which gets things roughly right. Some value for the base item type, some values for modifiers. We did this in Sil, for smithing difficulties -- we ended up using it to improve artefact balance too.