Hell-Pan Roulette

Name Hell-Pan roulette
Summary Abbreviate extended!
Proposer regret-index
Version 0.19

The structure of the extended endgame, inherited from near the very creation of Crawl, has some key design problems to it.

  • It's very long. Twelve more runes after the initial three required for a default win shows a clear backloading of play length. For those additional runes here's a lot of establishment filler, which has some unique dynamics but is presented in quite the excessive melange. Six floors each for four Hells to dive through, uncountable floors of Pan to find a hidden rune and dive for announced special runes, the winding design of Tomb's layout and threat- followed by another mean rune vault. Most branches in the main game are much shorter. When paired with high character strength / extracting tactics and Crawl's innate unreliable randomization (rolls for actions and damage), death in extended feels even more unsatisfyingly sudden than how one could have already won normally.
  • It's redundant. A lot of work has been done to introduce monster variety, but the fundamental overlaps- elemental themes, a spray of unholiness, guaranteed threatening mechanics like torment and damnation- are still present and defining. While there are monsters shared throughout earlier branches, when paired with the other two points they manage to keep vastly separate from one another in threats (compare, say, each Hell to each Lair roulette branch).
  • It's static. There's no level variation like the Lair branch roulette or the Slime/Abyss/Vaults third rune choice, since a fifteen-rune game has to visit every branch. Character builds solidify by extended standards- once a main tactic of devastating monsters is established and defenses are built up enough, players stop using new tactics for the majority of the encounters and experience goes towards frivolous ends (or mass conjurations for Ziggurat completion). This pairs nastily with the enduring length.

With all three problems side by side, it feels reasonable to try and repair the situation here. The proposal here addresses all three problems at once without cutting any content, and features discrete changes that can be implemented in testing segments.


  • Place only one of Pandemonium or Hells per game. Max unique runes per game is now 11.
  • Announce a specific placement of the demonic rune in Pan in new scary vaults.
    • Rune chances (demonic or otherwise) increase with other runes claimed and/or Pan lords slain.
  • Add a single-floor, refurbished holy branch to the Hells, keeping rune count even.


Place only one of portals to Pandemonium or the Vestibule in per-game. The max number of unique runes gathered in a game shrinks from fifteen to eleven, with one instead of two demonic branch sets. Extended runs are shorter and have a clear variation point from one another, while thematic / mechanical overlap (such as Tartarus and Tomb or Cerebov's realm and Gehenna) is severely reduced. For stronger characters and experienced players it's a shorter slog and a shorter victory lap, for weaker characters there's no ease of powering up for Hells through doing Pan first, and for newer players it's more variety and wonder attached to not play both when learning and winning full-rune games.

One note, though: Pandemonium is generally held to be weaker than the Hells- the less elaborate vaultwork for the runes, the reliance on increased level spawns and the attrition of demonic rune vaults pale in the face of randomly devastating hell effects pressure. Some further buffs are probably warranted to the monster set (e.g. lorocyproca, blood saints) if it's going to take over Hell's presence in half of all games.

Straggler runes

Since Pandemonium and the Hells have an uneven amount of runes, some more structural work has to be done beyond simply setting one to spawn and one to not, though.


The demonic rune is uniquely frustrating and awkward in the way it's hidden. Three choices, of the slim random chance in the vault on every floor, the spoiler-craving guaranteed recognizable guaranteed rune vaults, or the eventual landing in the additionally-spoiler-grasping holypan level, still outlines a spectrum of caution or reckless abandon but no matter the scenario requires going through many vaults and many floors of Pan to find it. There's not enough agency involved, no potential means of progress to the diving beyond choosing whether or not to take e.g. hellion_isle or holypan on a character when they show up.

It would be easy to argue that in the midst of this change to the greater branch that the demonic rune could be cut, but there are still positive qualities to the rune. It means directly engaging with something different from the specific themes of the other four Pan runes- the branch's set of mixed greater demons and classed demonspawn. It's a rune with high variability that mirrors its branch mixing together such a set of monsters, which works much more than, say, the out-of-branch threats for Swamp rune vaults. It also makes active use of a very large set of vaults and the constantly-iterated-upon random Pandemonium Lord designs, which would have next to no function with the removal of the demonic rune (beyond, say, forcing descending by placing most stairs inside the vaults, which introduces a different set of aggravations). Without the demonic rune, there's little point in the floors and contents in Pan outside of the other four Pan runes.

I've had drafts for a while now (see the Laplace branch) on an idea which tries to make it easier and meaner to find the demonic rune, which can be applied outside of this proposal: converting the guaranteed demonic rune vaults into announced serial vaults, and pairing up two per floor. Alongside a plotted-out second announced arrangement of four or five regular random demonic rune vaults placed on one floor, and with the option of increasing the chance for the serial vaults to spawn with each Pan rune obtained, this amplifies the choice dynamic- by making it much easier (with spoilers or not) to find the more dangerous demonic rune vault.

A further step in the above, which would be even more unusual but specifically provide a feeling of progress alongside more player control, would be to further increase the odds of the serial vaults (or even all of the Pan runes!) placing with each slain Pandemonium lord. Reducing bad luck through risk and high character strength seems to work for Abyss and is the principle of the above announced serial rune, anyway.

Hell's fifth

Keeping the demonic rune in a Hell-Pan roulette means an uneven rune count, though. There's a neglected monster set which can be put to work for this- the Holy monsters, currently mostly regulated to one vault in Pan and a lot of cameo / summon appearances. With constant reductions over the versions and years since the first big push to implement them contained many finnicky mechanics, Holypan has the two reliably functional and oldest angels alongside many straggler monsters not yet cut- this seems wasteful.

A quick-and-dirty single-floor branch draft is listed over in Purgatory. A single-floor subvaulted branch placed in the hells, after a long further set of changes and buffs to the monster set, provides one last nasty encounter and a severe shift in monster threats before finishing the Hells. The proposal for this branch could also use work, though- feedback is desired there, too.


  • Scoring is going to be screwed up moving from fifteen runes to eleven, and some changes to the formula plus a split in measurement is going to be required to accommodate this. I'm no master of the scoring formulas myself, so I welcome others to manage this.
  • This also disturbs a couple of tournament banners- requiring multiple games through extended to get all possible runes or uniques as fixed ideal accomplishment, alongside the lesser specific challenges running into not reliably having Dis in a rune-branchless game or it being unusual to slay across two separate games a Hell and a Pan lord. Not very important stuff, as this has been disrupted before and tournaments shouldn't influence the design of the base game, but it can be briefly noted for future reference.

Alternate proposals

Some aspects of these approaches held by others are orthogonal to the above Hell-Pan roulette proposal, and they're usually of much last drastic changes. They could probably get pages of their own if others wish to defend and expand on the rationales / ramifications involved, but the aforementioned problems of extended and the segmented manner solutions can come in mean it's worth mentioning the general gist of side-proposals for further approaches in abbreviating extended. Feel free to fit in feedback, arguments, and further proposals here, with signatures.

  • I'm not fond of long-held proposals to make Pan finite. The demonic rune changes / lord-slaying benefits above, the replacement of ziggurat raiding medium, and the lack of Hells in a game with Pandemonium should make visits to Pandemonium less aggravating in context, and the infinite nature of the branch curiously mirrors and separates itself from Hells in non-fixed depths diving. Still:
    • [Make Pandemonium all of 15 to 20 floors], with all runes guaranteed in some fashion in this set of floors. Having to dive an infinite plane for the random chances of seeing the highlight goals of the runes is an annoying matter to do for the same lack of agency and progress problems the demonic rune holds. There are also few benefits to allowing people to infinitely traverse the planes once they've gathered all of the runes.
    • The demonic rune can no longer be randomly hidden in this arrangement (but can use the above serial demonic rune vault announcement), necessitating the placement of stairs within Pan lord vaults to keep their value- which may be less but still somewhat aggravating with a fixed lowered amount of floors, in putting pressure on always searching for and visiting many Pan lord vaults. Maybe if there were two Pan lord vaults per floor, with the stairs split between them, to allow a little more choice?…
      • Alternatively, cut the demonic rune, which removes the need for the corresponding holy rune branch. I've defended the demonic rune earlier in my proposal, though, and the holy rune concept is of a short level with mostly revised monsters compared to imposing new content.
  • In the same, less prominent vein is the capability to cut down on Hell floors. Interim Hell floors work better than Pan floors because they're a potentially short known quantity (e.g. hatches, shafts) instead of an random unknown quantity, and the exploratory pressure of hell effects is much higher than Pan's demonic rune hunt or infinite varied floor spawns. They also have discrete themes and unique layouts to work with compared to Pan's combinations.
    • [e.g. Cut down the Hells to a single nine level branch, with runes on levels 3 / 5 / 7 / 9 – Rast.] It has to be acknowledged interim hell floors are readily frustrating because of the luck associated with getting a generous layout / minimal vault presence as one dives through each of them, and how one might get bogged down by hell effects and floor spawns during that diving that has to be done at least four different times.
      • Cutting out 19 floors seems rather drastic and with the combined branches makes the Hell-to-Pan comparison too close. I'd rather see, say, four or five floor branches with guaranteed vaults in a middle floor to still eek out uses of previous assets and hell effects through regular floors as opposed to the fixed, hostile, winding encompass vaults on their last floors.
  • Of course, we could just keep extended as it currently is- extended has a specific character of being a long and somewhat dangerous endeavour. Experience seems to positively correlate with getting bored of extended even faster than the spite and irreverence normally generated from playing through regular games with all character builds and combinations, though, and the problems above don't disappear when people advise / vow to not bother with extended very often.
Logged in as: Anonymous (VIEWER)
dcss/brainstorm/dungeon/hell-pan_roulette.txt · Last modified: 2016-05-19 05:17 by Claws
Recent changes RSS feed Donate Powered by PHP Valid XHTML 1.0 Valid CSS Driven by DokuWiki