D&D.Sci Dungeonbuilding: the Dungeon Tournament

post by aphyer · 2024-12-14T04:30:55.656Z · LW · GW · 11 comments

Contents

  STORY
  DATA & OBJECTIVES
  OPTIONAL HARD MODE
  SCHEDULING & COMMENTS
None
11 comments

This is an entry in the 'Dungeons & Data Science' series, a set of puzzles where players are given a dataset to analyze and an objective to pursue using information from that dataset.

Estimated Complexity [LW · GW]: 3.5/5  (this is a guess, I will update based on feedback/seeing how the scenario goes)

STORY

The Dungeon Tournament is held regularly, calling dungeon creators from around the kingdom to create the most challenging dungeon for adventurers to battle through.  The winner is crowned with glory (plus they receive a large dungeon-building grant, and are often hired to design labyrinths, pyramids, mazes, tombs, and many other dungeons).

The current competitors, however, have no conception of how to employ Data Science to optimize their dungeons!  After gathering data on the past dungeons and how they were scored, you entered the tournament, confident in your victory and the attendant glory and improved career prospects.  You planned a terrible dungeon, forcing adventurers through a deadly gauntlet of traps and ferocious monsters.

Unfortunately, managing the various monsters you need to stock a dungeon turned out to be more difficult than you expected.

Your dragons got into a fight that smashed one of your golems, and ended with one of the dragons leaving and flying away to sulk.

Your goblins have broken half of your traps while trying to play with them.

Your hag...hasn't done anything wrong exactly, but whenever you check in on her she cackles at you in a very unsettling way while muttering ominously at you[1].

You barely have enough encounters left to fill every room in your dungeon, much less to fill them with carefully-chosen challenging encounters.  The adventuring teams that will be exploring your dungeon for the competition[2] pass by the disaster area around your workshop and snigger at you.  All you can do is hope that clever use of the data can let you recover and avoid your dungeon being a complete disaster.

DATA & OBJECTIVES

OPTIONAL HARD MODE

You cannot stand these Goblins being around.  They break everything.  They steal everything.  They smell awful.  They constantly make terrible, terrible jokes.[3]  And they aren't even all that big a threat to adventurers!  Is there no way you could make do without them?  Or at least make do with fewer of them?

To pursue this bonus objective, use fewer Goblins in your submission.  Since you only have 9 encounters available to begin with, you will need to leave a room empty for each Goblins encounter you leave out.

You'd rather put up with these Goblins for a bit longer than embarrass yourself in the Dungeon Tournament, though: only leave them out if you think you can do that with minimal effect on your performance in the Tournament!

SCHEDULING & COMMENTS

Since we're in the holiday season, I expect many players (and also myself) to be busy, and so I'm leaving substantially more time for this scenario than usual so that anyone who is e.g. out for holidays can still have time to try it. I'll aim to post the ruleset and results on January 6th.

As usual, working together is allowed, but for the sake of anyone who wants to work alone, please spoiler parts of your answers  that contain information or questions about the dataset.  To spoiler answers on a PC, type a '>' followed by a '!' at the start of a line to open a spoiler block - to spoiler answers on mobile, type a ':::spoiler' at the start of a line and then a ':::' at the end to spoiler the line.

  1. ^

    "The bane of the mountains, the princess's hand.  The loyal protector all at her command.  Her lovely, her darling, oh now where are you?  Are you gone?  Are you lost?  No, you're boiled in a stew!"  

    At this point you made a hasty retreat, because you did not like the way the Hag was eyeing you while talking about boiling you in a stew.

  2. ^

    Four different adventuring teams have signed up as judges to run this tournament's dungeons; thankfully, Dungeon Tournament officials will be healing your monsters and resetting your traps between adventuring teams, so you won't have to worry about that.

  3. ^

    "My uncle Gobbo has no tongue!"  "How does he talk?"  "...Terrible!  Ahahahaha!"  And you don't even want to mention the one about the Mind Goblin.

11 comments

Comments sorted by top scores.

comment by abstractapplic · 2024-12-14T21:45:47.963Z · LW(p) · GW(p)

I still have a bunch of checking to confirm whether this actually works, but I'm getting my preliminary decision down ASAP:

CWB/OOH/XXD (where the Xes are Nothing or Goblins depending on whether I'm Hard-mode-ing)

On the basis that:

Adventurers should prioritize the 'empty' trapped rooms over the ones with Orcs, then end up funelled into the traps and towards the Hag; Clay Golem and Dragon are our aces, so they're placed in the two locations Adventurers can't complete the course without touching.

Replies from: abstractapplic
comment by abstractapplic · 2024-12-15T02:18:42.627Z · LW(p) · GW(p)

On further inspection it turns out I'm completely wrong about

how traps work.

and it looks like

Dungeoneers can always tell what kinds of fight they'll be getting into: min(feature effect) between 2 and 4 is what decides how they collectively impact Score.

It also looks like

The rankings of effectiveness are different between the Entry Square, the Exit Square, and Everywhere Else; Steel Golems are far and away the best choice for guarding the entrance but 'only' on par with Dragons elsewhere.

Lastly

It looks like there's a weak but solid benefit to dungeoneers having no choice even between similarly strong creatures: a choice of two dragons and a choice of two hags are both a bit scarier than hag-or-dragon. (Though that might just be because multiple of the same strong creature is evidence you're in a well-stocked dungeon? Feature effects are hard to detangle.)

Also

It seems like there's a weirdly strong interaction between the penultimate obstacle and the ultimate obstacle?

Replies from: abstractapplic
comment by abstractapplic · 2024-12-15T02:21:24.718Z · LW(p) · GW(p)

Oh and just for Posterity's sake, marking that I noticed both

the way some Tournaments will have 3 judges and others will have 4

and

the change in distribution somewhere between Tournaments 3000 and 4000

but I have no clue how to make use of these phenomena.

Replies from: kave, abstractapplic
comment by kave · 2024-12-16T20:07:31.140Z · LW(p) · GW(p)

Maybe sometimes a team will die in the dungeon?

comment by abstractapplic · 2024-12-15T12:31:54.159Z · LW(p) · GW(p)

On reflection, I think

my initial guess happened to be close to optimal

because

Adventurers will successfully deduce that a mid-dungeon Trap is less dangerous than a mid-dungeon Orc

and

Hag-then-Dragon seems to make best use of the weird endgame interaction I still don't understand

however

I'm scared Adventurers might choose Orcs-plus-optionality over Boulders

so my new plan is

CBW/OOH/XXD

(and I also suspect

COW/OBH/XXD

might be better because of

the tendency of Adventuring parties to pick Eastern routes over Southern ones when all else is equal

but I don't have the confidence to make that my answer.)

comment by kave · 2024-12-14T21:14:32.110Z · LW(p) · GW(p)

I'm guessing encounter 4 (rather than encounter 6) follows encounter 3?

Replies from: aphyer
comment by aphyer · 2024-12-14T21:18:31.984Z · LW(p) · GW(p)

The dungeon is laid out as depicted; Room 3 does not border Room 4, and does border Room 6.  You don't, however, know what exactly the adventurers are going to do in your dungeon, or which encounters they are going to do in which order.  Perhaps you could figure that out from the dataset.

(I've edited the doc to make this clearer).

comment by kave · 2024-12-14T23:04:21.392Z · LW(p) · GW(p)

So I did some super dumb modelling.

I was like: let's assume that there aren't interaction effects between the encounters either in the difficulty along a path or in the tendency to co-occur. And let's assume position doesn't matter. Let's also assume that the adventurers choose the minimally difficult path, only moving across room edges.

To estimate the value of an encounter, let's look at how the dungeons where it occurs in one of the two unavoidable locations (1 and 9) differ on average from the overall average.

Assuming ChatGPT did all the implementation correctly, this predictor never overestimates the score by much. Though it frequently, and sometimes egregiously, underestimates the score.

Anyway, using this model and this pathing assumption, we have DBN/OWH/NOC

We skip the goblins and put our fairly rubbish trap in the middle to stop adventurers picking and choosing which parts of the outside paths they take. The optimal path for the adventurers is DONOC, which has a predicted score of 30.29, which ChatGPT tells me is ~95th percentile.

I'd love to come at this with saner modelling (especially of adventurer behaviour), but I somewhat doubt I will.

comment by Christian Z R · 2024-12-18T10:24:01.044Z · LW(p) · GW(p)

 A few notes about strange phenomenons in the scores:

 

 1) As already pointed out there is a clear jump in scores by around one point around tournament 3400 (the jump is too small for me to be quite certain when it happened). This might be because of a small change in the rules. So conclusions drawn by data from before this point might be flawed.

2) The scores are either whole numbers, or fractions ending in halfs, thirds or quarters. So they might be from taking the mean of either 1, 2, 3 or 4 whole numbers. It is inherently tricky to find out how a given whole number came about. Also the half points might be the mean of either 2 or 4 numbers. 
However there seems to be some time dependence in this. The scores ending in .333... or .666... are slightly (but statistically significantly) more likely to be found in the earlier tournaments than the scores ending in 0.25, 0.5 or 0.75. (40% of thirds happen before tournament 3400 compared to 35% of quarters.). This might very well be related to the rule change happening around tournament 3400, so later tournaments were more likely to have 2 or 4 judges instead of 3.

3) The mean score of Scores ending in 0.33.. and 0.66.. seems to be almost the same as the mean for scores ending in 0.25, 0.5 or 0.75. so it can not be only the change in number of judges that led to the change in mean score.

Replies from: Christian Z R
comment by Christian Z R · 2024-12-19T09:31:46.980Z · LW(p) · GW(p)

Some more insights:

I assume that Adventurers can't walk diagonally. In that case we can try to look at dungeons where the same encounter is present in room 2 and 4 (or in room 6 and 8), so the adventurers must pass through that exact challeng. I then make a linear model on this encounter + include room 1 and 9 in order to control for the fact that dungeons with strong encounters in one room is more likely to also have them in others.

Looking at both the case where Enc2==Enc4 and the case where Enc6==Enc8 I got an agreement between the difficulties of the encounters (Nothing < W < G < B < O  < S < H=D ). Clay Golems however differed.

It also seemed that in general meeting a tough challenge later would make a bigger difference than meeting it early. 

So my plan will look the same as Abstractapplics for pretty much the same reasons, except I will place the Boulder Trap after the Whirling Blade, since Boulder Traps seems to be more challenging, and so should be encountered later.

Replies from: Christian Z R
comment by Christian Z R · 2024-12-19T09:32:36.463Z · LW(p) · GW(p)

So my plan is:

CWB / OOH/ XXD