DonJon Random Dungeon generator. Random Encounters are insane

[DontEatRawHagis]

So I found this resource for creating random generating:
https://donjon.bin.sh/5e/dungeon/

The problem I'm having is that the random encounters for 4 level 10 PCs are coming out like this:

1 4 x Otyugh (mm 248, cr 5), consumed by disease and madness
2 Half-Red Dragon Veteran (mm 180, cr 5) and 14 x Thug (mm 350, cr 1/2), scavenging for food and treasure
3 Mind Flayer Arcanist (mm 222, cr 8) and 12 x Intellect Devourer (mm 191, cr 2), wielding bizarre eldritch powers
4 1 x Cloaker (mm 41, cr 8), scouting from another part of the dungeon
5 1 x Cloaker (mm 41, cr 8), scavenging for food and treasure
6 4 x Umber Hulk (mm 292, cr 5), lost and desperate

Anyone familiar with this resource know why a CR8 and 12 CR2s is a balanced encounter? The math has the encounter coming out as 24,900 XP in Kobold Fight Club. For 4 level 10 PCs, the XP Budget for a Deadly Encounter is 11,200. I didn't find a support/comment link on the site, but I found it interesting/thought someone might have experience with Donjon here.

[Inevitability]

My guess: it's not factoring in XP increases for encounters consisting of multiple monsters. The described encounter is 12 x 450 + 3900 XP, that's 9300 XP without adjusting for the large number of creatures. For 4 level 10 PC's, that would be a Hard encounter.

[Daehron]

Meh, what the calculator says is "hard" PCs chew up in less than 2 rounds.

Up the difficulty, it's a good thing.

[DontEatRawHagis]

Meh, what the calculator says is "hard" PCs chew up in less than 2 rounds.

Up the difficulty, it's a good thing.

Agreed. My players chew through crap too easily.

[xanderh]

Meh, what the calculator says is "hard" PCs chew up in less than 2 rounds.

Up the difficulty, it's a good thing.

While that may be true, it's throwing 12 intellect devourers at the players. This can TPK in one round, and is going to be tough even at high levels, because of what is essentially a 2-turn save-or-die, where the only way to be safe is having an int of at least 19. It's probably only going to be the wizard that has that. And if they kill you by devouring your brain, they get control over your body,which means your party members are in big trouble.

That type of encounter is basically guaranteed to kill at least one PC, and will likely end in a TPK.

[Firechanter]

Int Devourers are terrible design and should _never_ be used. Not at level 2, not at level 10, not one and most definitely not twelve.
You might as well say "okay guys, I've had enough of this campaign, let's play something else."

[kaoskonfety]

Very cute tool... but at a glance all of the encounters its making need some tweeks. Something is clearly wrong on the force multiplier end and while 12 intellect devourers and *any* number of mind flayers are HILARIOUS that's the kind of encounter you put in front of combat happy PC's to make them rethink their lives and start negotiating (begging for mercy).

I'd see this more as a general time saver (map, rough encounter layout and some inspiration) than something I'd use "out of the box" on a lazy module night where I'm doing zero writing and a little reading.


edit - toying with it some more there are some issues with the encounter difficulties in general, an Umber Hulk random encounter for 3rd level PC's is going to be a bit dicey and the vampire spawn "bloodied and running from a greater threat" made me chuckle at how paranoid that's going to make the players.

Edit edit... and it REALLY likes intellect devouerers... like a creepy amount.

[DontEatRawHagis]

While that may be true, it's throwing 12 intellect devourers at the players. This can TPK in one round, and is going to be tough even at high levels, because of what is essentially a 2-turn save-or-die, where the only way to be safe is having an int of at least 19. It's probably only going to be the wizard that has that. And if they kill you by devouring your brain, they get control over your body,which means your party members are in big trouble.

That type of encounter is basically guaranteed to kill at least one PC, and will likely end in a TPK.

What if they have a Circlet of Intelligence? It sets their INT to 19.

??? Infinite Int Devourer food?

[Joe the Rat]

I always look at overkill encounters as an option for non-combat and non-sequitur encounters. A mind flayer and 12 intellect devourers? I'm thinking circus act.


Edit edit... and it REALLY likes intellect devouerers... like a creepy amount.Almost as if the programmers were under the control of some sort of brain eating para-

No, that way lies nonsense. Intellect Devourers are fictional and cannot transmit themselves across the internet and consume new hosts by lurking in RPG forums. And the number of intellect devourers suggested are too low, if anything. Yes, more intellect devourers!
>burp<

[kaoskonfety]

I always look at overkill encounters as an option for non-combat and non-sequitur encounters. A mind flayer and 12 intellect devourers? I'm thinking circus act.

"Over protective mother/creator Illithd terrified the PC's will hurt its 'babies' " - don't corner them and you're fine
"Dog walker - doesn't want to risk combat with several peers (superiors) intellect devouerers at risk, but has limited personal investment in their survival and is not getting paid nearly enough" - 'take them spare me!'
"A high elf and 12 dwarves under illusions to fit in and not be bothered in the Underdark"
"All of them are feeble minded by a passing vengeful wizard and will only act in self defence"
"the Illithd is a human wizard transformed into a Mind flayer, the Devourers are body guards deemed more useful in this form and also transformed"
"They are fleeing some other threat - the threat is a far more reasonable challange for the PC's but is immune to being un-brained/pshycic attacks (a golem or something), the Illithd offers a decent magic item if the party deals with it, and if they think to ask, also will agree to let them live"

[Democratus]

While some of the encounters generated can be whacky...DonJon is a treasure.

That site has saved me so many hours of prep time!

[MaxWilson]

So I found this resource for creating random generating:
https://donjon.bin.sh/5e/dungeon/

The problem I'm having is that the random encounters for 4 level 10 PCs are coming out like this:

Anyone familiar with this resource know why a CR8 and 12 CR2s is a balanced encounter? The math has the encounter coming out as 24,900 XP in Kobold Fight Club. For 4 level 10 PCs, the XP Budget for a Deadly Encounter is 11,200. I didn't find a support/comment link on the site, but I found it interesting/thought someone might have experience with Donjon here.

That's awesome! I would totally play that random dungeon. It seems like about the right difficulty level to me.

[DontEatRawHagis]

That's awesome! I would totally play that random dungeon. It seems like about the right difficulty level to me.

Just wait till you get the 10 young blue dragons encounters

[MaxWilson]

Just wait till you get the 10 young blue dragons encounters

Hahaha, awesome. Yes, I'd play that.

[xanderh]

What if they have a Circlet of Intelligence? It sets their INT to 19.

??? Infinite Int Devourer food?

The problem with that is that it trivialises the encounter for that player. Basically, the encounter is either guaranteed deadly without the item, but trivial with the item. That's not good creature design, unless the creature is basically the BBEG or close to it, and the magic item is some plot-related artifact. This is neither.

[darkelf]

Anyone familiar with this resource know why a CR8 and 12 CR2s is a balanced encounter?

step 4, modify total XP for multiple monsters. when making this calculation, don't count any monsters whose CR is significantly below the average CR of the other monsters in the group unless you think the weak monsters significantly contribute to the difficulty of the encounter.

since whether or not the weak monsters significantly contribute is a pure judgement call, the code can't consider that, and is simply discounting the CR 2 devourers as insignificant minions compared to the CR 8 mind flayer. thus, the total XP is calculated to be 3900 + (12 x 450) = 9300, which is comfortably hard.

also, random encounters are random.

[darkelf]

also, +1 sins for capitalizing donjon incorrectly.

[darkelf]

Just wait till you get the 10 young blue dragons encounters

juvenile delinquents out drinking and golem-tipping, no doubt.

[Tanarii]

since whether or not the weak monsters significantly contribute is a pure judgement call, the code can't consider that, and is simply discounting the CR 2 devourers as insignificant minions compared to the CR 8 mind flayer. thus, the total XP is calculated to be 3900 + (12 x 450) = 9300, which is comfortably hard.I've been through this exact discussion before, and the most reasonable way to account for the creatures is to include their seperate values added up. So 12 IDs value times three for their numbers, plus a mindflayer. Obviously that's a house-rule (of something that's a guideline in the first place), because the DMG just says to ignore the much lower CR creatures completely if they're not contributing. If you don't ignore them, they should get full value.

So either way, donjon is wrong. It should be (3900 + 12*450) * 3 size multiplier = 27900 (2.5 * Deadly) , or it should be 3900 (Easy). By my house-rule it'd be 3900 + (12*450)*3 = 20100 (1.8* Deadly). The way you calculated it, it'd actually be easier than facing the 12 IDs by themselves.

[darkelf]

So either way, donjon is wrong.

uhm, no. the process is laid out clearly in the DMG. if it sometimes provides odd results, that's a different matter and not my concern.

[MaxWilson]

The problem with that is that it trivialises the encounter for that player. Basically, the encounter is either guaranteed deadly without the item, but trivial with the item. That's not good creature design, unless the creature is basically the BBEG or close to it, and the magic item is some plot-related artifact. This is neither.

I wouldn't call 12 Intellect Devourers "trivial" even if you have Int 19.

(1) They can still kill your companions, eat their brains, and use your buddies' spells against you. That's non-trivial.
(2) They can still kill you to death with either psychic damage or their sharp claws.
(3) If they can knock you unconscious with their claws they can still eat your brain.


step 4, modify total XP for multiple monsters. when making this calculation, don't count any monsters whose CR is significantly below the average CR of the other monsters in the group unless you think the weak monsters significantly contribute to the difficulty of the encounter.

since whether or not the weak monsters significantly contribute is a pure judgement call, the code can't consider that, and is simply discounting the CR 2 devourers as insignificant minions compared to the CR 8 mind flayer. thus, the total XP is calculated to be 3900 + (12 x 450) = 9300, which is comfortably hard.

also, random encounters are random.

As a general rule, it's a good bet that lower-CR monsters always contribute significantly because bounded accuracy. The odd corner case can happen (one ancient dragon and two kobolds is not significantly harder than one ancient dragon) but in the general case, XP multipliers should always be applied, IMO.

[Tanarii]

uhm, no. the process is laid out clearly in the DMG. if it sometimes provides odd results, that's a different matter and not my concern.The process is laid out clearly: either count it all, or discount the low CR creatures completely.

Technically though my statement was wrong: donjon was right if it discounted the XP completely, as it's an Easy encounter if you discount the XP for the IDs completely. It's just your math that was wrong. And clearly so. It results in an adjusted encounter XP lower than just the 12 IDs by themselves.

[MaxWilson]

The process is laid out clearly: either count it all, or discount the low CR creatures completely.

That's not what step #4 says to do. You would only discount insignificant creatures for purposes of the multiplier. An ancient red dragon and two kobolds would be 62,050 x1 = 62,050 XP, not 62,000.

[Tanarii]

That's not what step #4 says to do. You would only discount insignificant creatures for purposes of the multiplier. An ancient red dragon and two kobolds would be 62,050 x1 = 62,050 XP, not 62,000.It says not to count the monsters. They don't add to the adjusted XP/difficulty at all if you follow the instructions.

[MaxWilson]

It says not to count the monsters. They don't add to the adjusted XP/difficulty at all if you follow the instructions.

That's not what step #4 says. It says not to count insignificant monsters while making "this calculation", i.e. computing the multiplier. It doesn't say to remove them from prior calculations.

[Tanarii]

This calculation is the difficulty calculation. Not some subset of it.

[darkelf]

This calculation is the difficulty calculation. Not some subset of it.

barring official examples, we'll have to agree to disagree then. the donjon code isn't changing in that regard. i should probably revisit determining what's insignificant, maxwilson makes a good point regarding the effect of bounded accuracy.

[MaxWilson]

Can I ask a question about donjon? I notice there are a lot of rooms with flavor text a la 'The south and east walls have been engraved with incoherent labyrinths, Someone has scrawled "The Obsidian Orb is wreathed in flames" in draconic script on the north wall'. Is donjon smart enough to give those things meaning enough of the time to make players pay attention to them, or is that up to the DM who's using donjon? For example, if placing a mind flayer and twelve intellect devourers in the dungeon results in a corridor nearby with "A thick slime trail coats the passage floor, and a terrifying odor of doom envelops you", then the flavor text is not just flavor text at all, it's a (hidden Markov) clue!

Looking over the generated dungeons, I haven't been able to determine yet whether the flavor text is completely random or linked.

[DontEatRawHagis]

Can I ask a question about donjon? I notice there are a lot of rooms with flavor text a la 'The south and east walls have been engraved with incoherent labyrinths, Someone has scrawled "The Obsidian Orb is wreathed in flames" in draconic script on the north wall'. Is donjon smart enough to give those things meaning enough of the time to make players pay attention to them, or is that up to the DM who's using donjon? For example, if placing a mind flayer and twelve intellect devourers in the dungeon results in a corridor nearby with "A thick slime trail coats the passage floor, and a terrifying odor of doom envelops you", then the flavor text is not just flavor text at all, it's a (hidden Markov) clue!

Looking over the generated dungeons, I haven't been able to determine yet whether the flavor text is completely random or linked.

It is random. Though its easy enough to change the flavor text. I put in a "Don't believe everything your read" early on so the players knew that some was true(i.e. I updated) and some I left random.

[Tanarii]

barring official examples, we'll have to agree to disagree then. the donjon code isn't changing in that regard. i should probably revisit determining what's insignificant, maxwilson makes a good point regarding the effect of bounded accuracy.
Ha. Didn't realize you'd programmed it.

If you want to make any adjustment, I'd suggest that it do a check of if the smaller CR alone would be > the calculated CR.

Basically, if A is the big CR XP, and B is the smaller CR XP, n is the number of small, and x is the adjustment multiplier, you run into an issue as soon as A + n*B < x*n*B, or A < (x-1)*n*B. Since x is gated based on n, this is actually a function of n, although I'm not sure how to write that out mathematically off the top of my head. But if you can code it, it'd prevent weird situations where the encounter is actually easier than the low CR creatures alone. (Note here I'm using yours and Max's assumption of how the difficulty formula works. Also this assumes solo big guy plus little guys, so there's no multiplier on the left side of the formula.)

[darkelf]

let's try this... for a mixed group, select the highest CR monster and calculate the adjusted total XP for that monster alone. in our current example, that's CR 8, one mind flayer, adjusted XP = (3900) x 1 = 3900 xp. divide by 4 to obtain the minimum significant XP = 3900 / 4 = 975 xp. for each additional group of monsters, calculate the adjusted total XP for that monster alone, and compare it to the minimum significant XP to determine whether or not to account for it in the final encounter multiplier. this lets small numbers of lower CR monsters remain insignificant, but larger numbers may become significant by weight. one ID at 450 xp would be insignificant, final encounter XP = (3900 + 450) x 1 = 4350 xp. two IDs at 1350 xp becomes significant, final encounter XP = (3900 + (450 x 2)) x 2 = 9600 xp.

i'm a little concerned that the divide by 4 step might yield a threshold which is too low, but its probably fine.

[MaxWilson]

let's try this... for a mixed group, select the highest CR monster and calculate the adjusted total XP for that monster alone. in our current example, that's CR 8, one mind flayer, adjusted XP = (3900) x 1 = 3900 xp. divide by 4 to obtain the minimum significant XP = 3900 / 4 = 975 xp. for each additional group of monsters, calculate the adjusted total XP for that monster alone, and compare it to the minimum significant XP to determine whether or not to account for it in the final encounter multiplier. this lets small numbers of lower CR monsters remain insignificant, but larger numbers may become significant by weight. one ID at 450 xp would be insignificant, final encounter XP = (3900 + 450) x 1 = 4350 xp. two IDs at 1350 xp becomes significant, final encounter XP = (3900 + (450 x 2)) x 2 = 9600 xp.

i'm a little concerned that the divide by 4 step might yield a threshold which is too low, but its probably fine.

Overall, that sounds fine to me. And I agree that the exact value of the threshold doesn't much matter; what matters is that there's a proportionate threshold.

The one problem I see, which might or might not be a problem, is that you're considering every subsequent group of monsters separately when they really should be aggregated. Otherwise you get different results for

1 Mind Flayer (3900 XP)
8 Orcs (2000 XP--significant!)

than for

1 Mind Flayer (3900 XP)
4 Orcs (800 XP--not significant!)
7 Drow (875 XP--not significant!)
1 Ogre (450 XP--not significant!)

The second encounter is clearly harder, but by calculating everything separately it "looks" easier to the proposed heuristic. Whether this is a problem or not in practice will depend upon whether donjon actually ever generates encounters with the second "shape", but perhaps you could avoid it by making the significant test monotonic in the number of monsters somehow. Maybe compare the cumulative total to the threshold instead of only the per-group total?

Edit: of course, since you're writing a computer program, you could actually use more sophisticated algorithsm than the DMG. For example, convert CRs to canonical HP/DPR ranges (using the DMG monster quick-stats table) for all the groups, then sum them all up. If the HP/DPR for the biggest CR monster is within N% of the HP/DPR of the group as a whole, treat the other monsters as insignificant. N=25% seems reasonable but I could even see as high as 50%. You could even parameterize it.

Edit2: you could also consider the case where there isn't just a single biggest monster, but more than one. E.g. 2 adult red dragons and 5 kobolds. You could iteratively run the algorithm on the biggest M monsters in the group until you reach a point at which considering additional monsters increases "official" XP calculation by more than the proportionate increase in HP/DPR; call that the break-even point, and any monsters over that threshold don't count towards the multiplier. I think that algorithm will be capable of detecting the difference between the dragons and the kobolds and disregarding the kobolds.

[Tanarii]

I say use whichever is greater, the formula as written, or the addition of the "significant" and "insignificant" individually.

For Mindflayer + 12 IDs that's whichever is greater of:
3900 + (12 * 450) (Formula as written the way you guys interpret it)
3900 + 3*(12*450) (addition of two numbers individually)

That prevents situations where the difficulty is actually calculated as less than the "insignificant" creatures alone.

More generally, building off my previous formula and rearranging the variables to make more sense, greater of:
x*(n*A + m*B)
x*n*A + y*m*B

Where:
A = "significant" creatures XP
n = number of "significant" creatures
x = multiplier for number of "significant" creatures
B = "insignificant" creatures XP
m = number of "insignificant" creatures
y = multiplier for "insignificant" creatures

[xanderh]

I say use whichever is greater, the formula as written, or the addition of the "significant" and "insignificant" individually.

For Mindflayer + 12 IDs that's whichever is greater of:
3900 + (12 * 450) (Formula as written the way you guys interpret it)
3900 + 3*(12*450) (addition of two numbers individually)

That prevents situations where the difficulty is actually calculated as less than the "insignificant" creatures alone.

More generally, building off my previous formula and rearranging the variables to make more sense, greater of:
x*(n*A + m*B)
x*n*A + y*m*B

Where:
A = "significant" creatures XP
n = number of "significant" creatures
x = multiplier for number of "significant" creatures
B = "insignificant" creatures XP
m = number of "insignificant" creatures
y = multiplier for "insignificant" creatures

Is there ever a situation where it would use the first version and not the second one? Since there's no negative multipliers, would it ever use the one without a multiplier?

[Tanarii]

Is there ever a situation where it would use the first version and not the second one? Since there's no negative multipliers, would it ever use the one without a multiplier?

Actually what I meant to suggest was the greater of:
x*(n*A + m*B)
y*m*B

So thanks for the question. :)

The other way I had it, the only time the top one came into play was is x > y.

[darkelf]

Whether this is a problem or not in practice will depend upon whether donjon actually ever generates encounters with the second "shape",

for sanity's sake, it does not. :)

[DontEatRawHagis]

for sanity's sake, it does not. :)

From what I can tell Donjon isn't truly random. In fact many of the generators I use seem to only use 2 different enemy types for an encounter.

I had to use another program to get more variability with encounters. One such I found had Mummies, Beholder Zombies, Whites, and Skeleton Minotaurs. So I've been using that instead of donjon.

[MaxWilson]

From what I can tell Donjon isn't truly random. In fact many of the generators I use seem to only use 2 different enemy types for an encounter.

I had to use another program to get more variability with encounters. One such I found had Mummies, Beholder Zombies, Whites, and Skeleton Minotaurs. So I've been using that instead of donjon.

It would be awesome to have a program that took maps into account. For example, someone in this thread mentioned 10 Young Blue Dragons, and I thought that was awesome... but then I realized that it would be very hard to fit 10 blue dragons into any given room of a donjon map, since they take up 4 squares each (assuming each square is 5' by 5').

I'm certainly not knocking donjon--I think it's awesome what it is doing already--but it kind of has me wanting to write my own cellular automata-based random dungeon generator so I can tweak it to my own satisfaction and make it actually place the monsters on the maps. Not that I'm likely to find time for that any time soon, but if anyone knows a random generator that already does that I'd be obliged to you for pointing it out. :)

[darkelf]

It would be awesome to have a program that took maps into account. For example, someone in this thread mentioned 10 Young Blue Dragons, and I thought that was awesome... but then I realized that it would be very hard to fit 10 blue dragons into any given room of a donjon map, since they take up 4 squares each (assuming each square is 5' by 5').

i assume 10' squares, old school.


I'm certainly not knocking donjon--I think it's awesome what it is doing already--but it kind of has me wanting to write my own cellular automata-based random dungeon generator so I can tweak it to my own satisfaction and make it actually place the monsters on the maps. Not that I'm likely to find time for that any time soon, but if anyone knows a random generator that already does that I'd be obliged to you for pointing it out. :)

please do, we need more generators in the gene pool.

[Tanarii]

i assume 10' squares, old school.I've been playing around with it for about a week or so, which makes the timing of this thread great ;). But the first thing I noticed was I constantly had to set the room size to "Small" to get rooms that did seem like ludicrously huge caverns (for 5e). What did you base your room size determinations on? Does it generate by the DMG random generation tables?

I'm not trying to bash the tool I think it's awesome, and obviously I'm talking about a personal preference, not a problem. Just commenting on the very first thing I noticed when using it.

[MaxWilson]

please do, we need more generators in the gene pool.

All right, maybe I'll have time this weekend.

[darkelf]

I've been playing around with it for about a week or so, which makes the timing of this thread great ;). But the first thing I noticed was I constantly had to set the room size to "Small" to get rooms that did seem like ludicrously huge caverns (for 5e). What did you base your room size determinations on?

the room sizes are based on nothing more than my own personal preference for ludicrously large rooms. :) i prefer them creatively, as they give me room to expand upon dungeon themes and random elements. and i prefer them tactically, as they give the monsters and PCs room to move around in, and not get wiped out by a single fireball spell.