Variable length branches

[GermanJoey]

This is an idea I had due to the discussion from my other thread, about crawl's midgame being too long (viewtopic.php?f=8&t=617), prompted by people correctly (IMHO) pointing out that part of the problem with midgame is that the current optimal branch-order usually involves skipping dangerous branch-ends until much later.

So, my idea is that Instead of just straight chopping off floors from branches, what if they were made variable instead? For instance, take Snake: instead of being five floors, including the branch-end, could be made 2+1d3 (average 4) floors, or perhaps, more conservatively, 2+1d4 (average 4.5) floors. The effect would be:

1.) The mid-game, overall, would be slightly shorter. (see the other thread linked above for pros and cons of this)
2.) Magic Mapping scrolls would be now worth a whole lot more in the mid-game, rather than just as a resource to hoard for the end.
3.) The current optimal strategy of skipping branch ends would have a big ole' wrench thrown into its gears, as:

• Exploring only the first two "safe" floors of the post-orc branches would drastically shorten a player's bearings-gaining midgame.
• You might not know if you're in a branch end until you're halfway through it, at which point you'll be more likely to either say, "hey, yeah, I can do this...", or start freaking out because you're sure you're about to die, or maybe one after the other.

A bit of uncertainty in a player's long-term plans seems like the rougelike way to do it, at any rate!

[galehar]

It would be hard to balance, because so much XP depend on a single roll. If you're unlucky and keep getting short branches, you end up crippled. Also, it would encourage spoilers to be able to recognise end vaults as soon as possible. Would also encourage wearing stasis all the time, because one of the nastiest thing that can happen to you in a branch end, is stepping on a teletrap and ending (!) in the middle of the vault.

[dpeg]

galehar: The same could be said about certain vaults, however. We probably all know the early gnoll/kobold castles -- getting one of those means significantly more xp than you'd have gotten otherwise.

Regarding the global structure, I believe in this:
0. Keep cutting as before, in small steps.
1. Better grip on monster generation. Use this to make monster progression smoother.
2. The dungeon builder should accept target sizes smaller than 80x72.

[GermanJoey]

It would be hard to balance, because so much XP depend on a single roll. If you're unlucky and keep getting short branches, you end up crippled. Also, it would encourage spoilers to be able to recognise end vaults as soon as possible. Would also encourage wearing stasis all the time, because one of the nastiest thing that can happen to you in a branch end, is stepping on a teletrap and ending (!) in the middle of the vault.

You would not end up crippled. Even if both, say, Shoals and Snake, rolled 3 floors a piece, you would be deprived of at 10-15k total lost XP as compared to 5 floor branches. Other branches in fact could roll longer. That much XP could be easily made up by even a single longer floor of the Vaults. Roguelikes depend on randomness to keep them interesting. If I start a mummy, at what point do I consider my character crippled if I have not yet found a wand of healing? Or how many spellbooks does Sif Muna have to gift my Ice Elementalist before I should say "So many books and Sif still hasn't given me OCR yet. This character isn't worth playing anymore." and quit?

In fact, if branch-length were determined at game-start, like Demonspawn mutations are, branch-length size could be determined to have a constant total. That is to say, you'd know that Snake+Shoals+Slime+Vault+Elf+Orc+Crypt would have a total of, say, 33 floors, and so if your Snake and Shoals are each only 3 floors, you'd know that your other branches would guaranteed not be the minimum size.

Also, your last point is ridiculous. Do players walk around branch-ends now with an amulet of stasis just in case they hit a teleport trap? I've never heard of such a thing! It has certainly never happened to me.

[GermanJoey]

galehar: The same could be said about certain vaults, however. We probably all know the early gnoll/kobold castles -- getting one of those means significantly more xp than you'd have gotten otherwise.

Regarding the global structure, I believe in this:
0. Keep cutting as before, in small steps.
1. Better grip on monster generation. Use this to make monster progression smoother.
2. The dungeon builder should accept target sizes smaller than 80x72.

You're a tough man to convince! Do you have any thoughts on the variable-length?

[dpeg]

GermanJoey: I don't follow? My first line was a counter-argument to galehar's point.
And it is true, I didn't address your idea directly. Here's more background: the point I gave are crucial (in my opinion), so should be undertaken no matter what. Your proposal is interesting, but optional.

[mageykun]

The way I see it, variable branch lengths are only really interesting if the length is hidden- so you don't know when you're going to find the "boss" as it were. But the ^o screen gives away lengths.

What I dislike about this approach is it randomized difficulty. Usually the biggest factor in a given games difficultly is up to the player- your chosen build and playstyle. What the RGN throws at you is secondary. But if this is implemented, a unluckly "short" dungeon could suddenly trigger hardmode on someone who wasn't looking for it.

You could kind of solve these problems by trying monster density to the number of floors. If you get less floors, you get more monsters packed in to fix the xp loss. ...but that still increases the difficulty in certain games, as denser monster crowds are generally harder.

I remember nethack had a variable length for the main dungeon. Mostly it was just annoying when you rolled a long game and had to spend more time in Gehennom. Different situation though, because the variability just effected end game length, not useful growth and progression.

[Tiber]

The thing is, avoiding danger is supposed to be smart play, is it not? I think if the problem is people hold off on the last level, the solution isn't to trick them into doing the last level without realizing it. If anything, I'd say it's a matter of smoother progression of difficulty. And if you want more variable length, Add more portals.

[GermanJoey]

The thing is, avoiding danger is supposed to be smart play, is it not? I think if the problem is people hold off on the last level, the solution isn't to trick them into doing the last level without realizing it. If anything, I'd say it's a matter of smoother progression of difficulty. And if you want more variable length, Add more portals.

Avoiding danger *is* smart play, but the funny thing about danger is that you never know where its gonna pop up! For instance, I played a mummy ice elementalist about a week ago, and after I finished Lair and Orc, I decided to next explore either Swamp or Snake. I decided on Swamp first. However, Swamp 2 ended up having Margery and her band of Merry Men, Mara, Wiglaf, a crapload of slime creatures and ugly things, and a couple of hydras to boot. I killed Wiglaf and managed to escape by the skin of my teeth by burning through several teleport and blink scrolls. I then proceeded to snake, which I annihilated with ease. If I had known Swamp would have been such a disaster (I later learned that Swamp:5 contained the Lernaean Hydra, from which I escaped a certain-death when I was caught by surprise with 1 mana by a luckily-working fear scroll), I would have never have ventured there at all. How is the unknown danger of situation like this any different than the unknown danger of where the branch end is?

Knowing the floor of the branch-end itself is no guarantee of danger; compare the difference between the hound lair ending to the one filled with demons and a gate to hell or the one with the rampaging dire elephants. The swamp endings are similar in this regard, as they are so variable that different character builds may find them anywhere from very easy to extremely difficult. I also recall the 0.6 ending of Elf that had a lava/deep water moat in front of the final vault, which made it trivial for even a low-level character to kill 90% of the elves.