Pages

Friday, March 20, 2026

Minimalist treasure 2: generating hoard types

Continuing from the previous post, I wanted a way to generate treasure without complicated tables. The goal: a single roll that tells you not just how much treasure there is, but what kind — and how much of it is worth carrying.

Introduction (designer's notes)

Modern economists, following Aristotle, identify (at least) three qualities of sound money: portability, divisibility, and durability. In D&D, coins are certainly durable, but portability and divisibility are more situational. Few millionaires keep a significant share of their wealth in one-dollar bills — and no one with fifty thousand dollars in assets holds it as a single diamond. The denomination also has to match the transaction. Nobody buys a Ferrari with piles of dimes.

Put simply, the wealthier a person is, the more efficient their treasure tends to be per pound.

A handful of silver coins offers more divisibility than gold, but the gold is more portable. A gem offers more portability still, at the cost of divisibility, since dividing a ruby makes it lose most of its value.

In D&D, perhaps only a dragon — a compulsive hoarder by mythological nature — would accumulate the kind of mixed, inefficient pile that the encumbrance system is quietly designed to punish. Everyone else, given the choice, gravitates toward the top of the table.

This system takes all of this into account. If I may say so, it ended up being a lot easier to use than I expected.

The system

You already know the total value of the hoard (roughly 100 gp per HD, as discussed before; a lair of 100 bandits has a hoard worth about 5,000 gp). This roll tells you the composition — what the treasure is made of, and how efficiently it's packed.

Roll d20, then add +1 for every 1,000 gp in the hoard, up to a maximum of +20 (corresponding to a 20,000 gp hoard). Larger hoards should be broken into smaller parts and rolled separately. The result determines treasure type from the table below.

RollTreasure typeValue density
0 or lessLow-value or heavy objects (junk, pelts, tools)= gold / 1,000
1Reroll -10
2–9Copper= gold / 100
10–11Reroll ±10
12–19Silver= gold / 10
20–21Reroll ±10
22–29Gold= gold
30–31Reroll ±10
32–39Platinum= gold × 10
40–41Reroll ±10
42–49Gems & Jewelry~ gold × 100
50–51Reroll ±10
52–59Magic items~ gold × 1,000
60Reroll +10
61+Legendary item or artifact~ gold × 10,000

A quick note on the value density column. Each tier is roughly 10× more efficient per pound than the tier below — a clean order-of-magnitude progression that makes the table easy to understand.

And this progression looks very reasonable and not that far from the original. Consider:

Gems and jewelry in B/X are worth about 1,000 times their weight in gold (the 100× figure in our table is a deliberate simplification). In the real world, everyday jewelry is worth only slightly more than its gold content, but gems span an enormous range — common stones are worth a fraction of gold per gram, while fine rubies and emeralds can reach 2,000–5,000×.

Magic items vary immensely in price and weight, but 1,000× their weight in gold is not absurd as a round number. The spread is enormous: a Ring of Invisibility is roughly 3,300 times more valuable per pound than +1 plate armor, which works similarly to the gem range — the same category contains both a pebble and a diamond.

Because of that, gems and magic items might deserve their own sub-tables (which are already in B/X, AD&D etc.).

Treasure composition

The digit at the end of your result — the ones place, from 2 to 9 — tells you how heterogeneous the treasure is. This is where Pareto comes in: if a player takes only 20% of the treasure by weight, how much of the value do they recover?

  • Digit 9: 90% of the value is in the top 20% by weight. Highly varied — gems scattered among coins, a magic item wrapped in cloth at the bottom of a chest of silver.
  • Digit 5: roughly 50% of the value in the top 20%. Moderately mixed.
  • Digit 2: 20% of the value in the top 20%. Nearly homogeneous — a chest of nothing but copper pieces, uniform all the way down.

A result of 24 (gold, digit 4) means: the treasure is worth its weight in gold on average, and taking the best 20% by weight recovers about 40% of the value (this 40% is mostly platinum pieces, some gems, etc.). A result of 27 (silver, digit 7) means: it's heavy and low-value on average, but picking carefully gets you 70% of the value in 20% of the weight — there's some gold in there.

Results ending in 0 or 1 are rerolls: 0 means roll again and add 10, 1 means roll again and subtract 10. This preserves the natural 20 as a potential windfall and the natural 1 as a setback, while keeping the table open-ended in both directions. The +20 cap means most large hoards cluster in the gold-to-platinum range, with gems and magic items requiring either a lucky roll or a genuinely exceptional hoard.

Treasure appearance

Homogenous treasure (i.e., digit 2) is easy to describe. For example, result 22 is basically a big pile of gold, etc. Mixed treasure, however, will look mostly as one tier below. E.g.,  result 25 is maybe almsot half silver, around 20% gold, around 20% copper, and only a bit of platinum.


Pocket money

Individual creatures carry roughly 1% of the lair's total value on their person, in the same denomination as the main hoard, provided they can carry it. A gnoll lair worth 3,000 gp in silver means each gnoll carries about 30 gp worth of silver coins — enough to be worth mentioning, not enough to change the logistics. This also gives the party a small preview of what's coming: creatures carrying gold suggest a gold hoard ahead; creatures carrying gems suggest something more interesting.

Outliers

Even a very small hoard can contain gems and magic items. This isn't usually a problem — as noted above, both categories vary enormously in value. A potion might be worth 50 gp, a semi-precious stone even less.

If you roll gem and magic item values separately, you face a choice: fix the results to match the hoard's overall scale, or let the dice fall and accept that a lone goblin might be carrying a ring of inestimable value. Maybe that deserves a table of its own... but that's a post for another day.

Carrying mixed treasure

The digit makes it possible to record treasure concisely and make decisions about it later. A player might note "30 pounds of platinum (34) Treasure" — meaning the hoard is platinum-tier (so, it is worth 3,000 gp) and the digit is 4. Back in town, or at a bottleneck in the dungeon, they can decide to keep only the best 20% by weight: that recovers roughly 40% of the value, or about 1,200 gp, at one fifth the encumbrance. If they have a cart, they take it all. If they're fleeing through a collapsing corridor, they know exactly what to grab first and what to leave behind — and they have a number to justify the decision at the table rather than an argument.

Pareto to infinity?

The system can recurse. A gold-tier hoard contains a platinum-tier sub-hoard — roll again to find what share (the digit, as before). That platinum sub-hoard may itself contain a gem-tier portion, and so on, until you've identified the single most valuable in the pile. Stop whenever the detail stops being useful, or when the digit is 2 (meaning a uniform pile of coins, homogeneous all the way down). In theory you could choose the best gem in a handful... But few adventurers are wealthy enough to leave any gems behind!

Time is money

One thing I haven't analyzed here (nor have I seen it addressed in any D&D rulebook) is the time required to sort a hoard. A disorganized dragon hoard could take hours to sift through properly. Most human-administered treasures, by contrast, will have at least some organization and can probably be assessed and selectively looted in a few minutes, depending on size. In a rush, however, PCs might be forced to carry a few random bags and trust their luck!

But does this actually make sense?

Yes!

For example, if you rolled 24, this is what a treasure could look like. This is mostly AI-generated but fixable by adding more copper and gems, and probably a human with excel could do a similar job.

Hoard: ~10,000 gp total, result 24 (gold tier, digit 4)

ItemWeightValue% of value
30,000 cp in copper coins300 lb300 gp3%
40,000 sp in silver coins400 lb4,000 gp40%
3,000 gp in gold coins300 lb3,000 gp25%
120 pp in platinum coins12 lb600 gp6%
5 gems (avg 80 gp each)0.1 lb400 gp4%
2 pieces of jewelry (avg 200 gp)0.2 lb400 gp4%
1 magic item0.5 lb1,000 gp10%
Total~1,013 lb~9,700 gp100%

The top 20% by weight is ~203 lb. That's:

  • the magic item (0.5 lb, 1,000 gp)
  • the jewelry (0.2 lb, 400 gp)
  • the gems (0.1 lb, 400 gp)
  • the platinum (12 lb, 600 gp)
  • and about 190 lb of gold coins (1,900 gp)

Total: ~203 lb, ~4,300 gp — about 44% of the value in 20% of the weight.

4 comments:

  1. This is a very interesting and fruitful way of thinking about generating hoards because it focuses on the key trade-off between value and encumbrance; and it recognizes the other trade-off between value and time spent sorting. Weight and time are, of course, the most important limited resources in OSR games.

    However, the analysis feels difficult to use at the table. Once the DM generates the total value of the hoard, it seems to me that the 2 key questions facing adventurers are (1) what is the weight of the entire hoard, and (2) assuming we have ___ encumbrance available to carry, how much value of the hoard can we carry away?

    There are really three variables here: (a) what is the total value of the hoard; (b) what is the value per weight of the hoard, i.e. is it primarily copper, silver, gold, platinum, gems, etc.; and (c) how dispersed is the hoard.

    (a) and (b) are fairly straightforward, and would be determined by the system/difficulty of the hoard. Determing how dispersion feeds into the question is hard question of math.

    Recalling my college intro to statistics, I think the best way to think about dispersion is a gini coefficient, i.e. the relative amount of the total value captured by a percentage of the coins (or equivalently, coin-weights if you assume all coins weigh the same). If the gini coefficient is 0, then the hoard is perfectly homogenous and each coin is worth 1/total value; if the gini coefficient is 1, then all value is concentrated in a single coin.

    The actual formulas are very hairy (and to be honest, I needed AI help to derive them). But we can pick some reasonable values for the gini coefficient, and pre-compute the results. The tables are linked below, with the following variables:

    V = value of hoard, measured in median coins
    N = number of coins, or equivalently coin-weight of treasure
    G = Gini coefficient

    https://docs.google.com/spreadsheets/d/1zLY9r-Bd6OkvonNTj30l84c5cVyuMDtYIs4hw2Ck0wk/edit?usp=sharing

    These tables allow the GM to concisely specify a treasure hoard, and quickly determine the weight of the hoard, based on the hoard type (as represented by G). The G values of 0 and 1 are not really useful in practice, but helpful for intuitions.

    For example, suppose the GM determines that the hoard is primarily silver coins and objects, that the value of the hoard is 1,000 gp or 10,000 sp (measured in the primary metal); and that the hoard is normal, mixed denomination with G = 0.6.

    From the table, the number of coins in the hoard is 49% of V, or 4,900 coins of mixed denominations. The party can only carry 1,000 coins of weight, so approximately 20% of N. From the second table, if they used all their carry weight, this would allow the party to take 64% of the value of the hoard, or 6,400 sp in value (but not actually 6,400 coins; there are only 4,900 coins in the entire hoard!). But the party is greedy, they want to grab at least 8,500 sp in value (85% of the value of the hoard). From the third table, the GM tells the party they would need about 44% of N, or about 2,156 coins worth of carrying capacity to carry away that much value.

    Whew, that was an interesting rabbit-hole.











    ReplyDelete
    Replies
    1. Wow, great stuff!

      Which I had considered the Gini coefficient.

      Well, I did really like it at first but maybe it is still too complicated for a minimalist game.

      As I tried to generate actual treasures I've realized that, since some items make other items worthless (i.e., a single gem which is worth 100 gp makes all copper meaningless in comparison), maybe I can limit treasure to three or four types per hoard, assuming other types are worth no more than 1% of the total value.

      So maybe a party III of the series is coming soon...

      Delete
    2. I agree; it seems complicated. If you're interested, I updated the Google spreadsheet to incorporate the actual formulas as a new sheet. You can enter 3 values and the spreadsheet does all the calculations for you. Maybe still too complicated to actually use, but its a tool I put in my permanent toolbox.

      Delete