The resource model of StarCraft Brood War

Newcomer Learner Competitor Advanced

The short version

A resource is anything limited and contested. That is a much bigger list than minerals and gas.
The game counts some of your resources for you. It does not count the important ones.
Every decision you make is a conversion of one resource into another, and every conversion costs time.
Attention is the scarcest resource you own. It does not accumulate, you cannot bank it, and you spend it whether you meant to or not.
A bank is not power. Minerals in the bank are minerals that have not become anything yet.

The situation

You lose a game and open the score screen. You mined more than they did. You built more workers than they did. Your army supply was higher at three separate moments. And you still lost, at about eleven minutes, to an attack you saw coming.

The score screen is telling you the truth about the things it can count. The game it counted is not the game you played. What you actually spent, in the order you spent it, was: a decision to take gas early, four seconds of not looking at your mineral line, one probe that walked the long way, the ground in front of your natural, and a click budget that ran out about a minute before the fight did.

None of that is on the score screen. All of it was a resource.

See it

Pick a chain and follow it. Every arrow is a conversion, and no arrow is free or instant.

Spent: 50 minerals and the time the probe takes to walk out. Got: a machine that pays the 50 back and then keeps paying. Cost: everything else that 50 could have been, for as long as the payback takes.

Nothing on this map is free and nothing is instant. Every arrow is a conversion, every conversion has a price on the left and a delay in the middle.

The map is not a strategy. It is a checklist. When you are about to spend something, the two questions are the ones the class left its students with in lesson 1: what did I spend, and what did I get. If you cannot answer the second one in a sentence, you are not making a decision, you are reacting.

The four kinds of resource

The lecture sorted them into four boxes, and the boxes are more useful than the list.

Raw resources are the ones the game counts on your behalf: minerals, gas, supply, energy, buildings, units, and time. Two of the three races carry two more. Protoss have power, which is the right to place a building at all. Zerg have creep, which is the same thing wearing a different hat. Raw resources are the only ones with a number in the top right of your screen, which is exactly why they are the ones players over-weight.

Physical resources exist in your room, not in the game. Your reaction speed. Your actions per minute. Your eyesight, which is a real constraint the lecturer made a joke about and then took seriously: you genuinely cannot see what is under a dark swarm. Your practice time. Your stamina at game forty of a session, and your morale after losing three in a row. Nothing in the game engine knows these exist, and they decide more games than gas does.

Transformation resources are the concepts that turn one resource into another. An army is the machine that turns units into power. Position, and high ground in particular, turns the same army into more army. Territory turns into economy. Macro is the name for the whole conversion plant.

Assistant resources do nothing at all on their own and multiply what you already have. Upgrades. Tech. Support units. Building placement. Micro. An armour upgrade kills nobody. It makes every body you own and every body you will own harder to kill, which is why its value grows with the size of your army and why buying it early against a small army is a real mistake rather than a cautious one.

The taxonomy earns its place the moment you notice that most players only budget the first box.

Every decision is a conversion

Here is the move that makes the model do work. Stop thinking of a decision as a choice between options, and think of it as an exchange with a price on both sides.

You build a photon cannon behind your mineral line. In the first box that is 150 minerals gone. In the fourth box, you have bought back a physical resource: the reaction speed you do not have. The cannon is standing there so that you do not have to be. You have converted minerals into attention, at a fixed exchange rate, permanently.

You hold a choke with three units and never push out of it. That choke has cost you something: the map behind it, the expansion you did not take, and the initiative. It has also bought you something: a flux cap on everyone who wants to come through, which is worth more than three units are. Both halves are real, and the habit worth building is naming both out loud.

You take gas at four minutes. You spent worker time that was mining minerals, and the ability to build the cheap thing right now. You bought a branch of the tech tree, and, whether you wanted to or not, you also sold information: your gas count is one of the loudest tells in the game, which is a point lesson 2 makes and which T4 picks up.

Conversions cost time

Every arrow on that map has a delay on it, and the delay is the part players forget.

A mineral is not a zealot. A mineral is a zealot in about forty seconds, if a gateway already exists, and in about two minutes if one does not. During that delay the mineral is doing nothing for you at all, and neither is the thing it is turning into. This is why a bank is not power: a thousand minerals in the bank is not a thousand minerals of army, it is a thousand minerals of army that starts existing a minute from now, and the fight is at eleven minutes.

The same delay is what makes an investment an investment. A worker costs fifty minerals now and returns about forty a minute forever. It pays for itself in something over a minute, and every minute after that is profit. The reason you build workers early and stop building them late is not a rule from a build order, it is that the payback window closes as the game shortens.

Attention is the one you cannot bank

Minerals accumulate. Gas accumulates. Attention does not. Every second you are alive in a game you are spending it, and if you spend it on nothing it is still gone.

That single property changes how you should treat it. You cannot save attention for the fight. You can only decide, in advance, what you will not be looking at. A player with a hundred and twenty actions a minute who has decided that he will not look at his mineral line between eight and nine minutes has made a real plan. A player who intends to do everything has made none, and will find out which half he dropped afterwards.

This is also why the cannon example above is not a gimmick. Buying back attention with minerals is one of the few exchanges in the game that runs in that direction, and the reason experienced players build static defence in places a stronger player would not need it is that they are pricing their own attention honestly. T10 turns this into a budget you can actually allocate.

The working

Show the maths

Take a worker that costs $c$ minerals and mines at $m$ minerals a minute. It is under construction for $\tau$ , and the game has $T$ minutes left in it. Its lifetime profit is

\Pi(T) = m\,(T - \tau) - c

(1)

Everything it mines after it exists, less what it cost to exist.

Set that to zero and you get the moment the worker stops being a loss:

T^{*} = \tau + \frac{c}{m}

(2)

With c of 50 and m of about 40 a minute, that is a little over a minute and a half of remaining game.

Now the interesting version. Suppose you spend every mineral on workers. Income is $m W$ , each worker costs $c$ , so the worker count grows as

\frac{dW}{dt} = \frac{m}{c}\,W \quad\Longrightarrow\quad W(t) = W_0\,e^{(m/c)\,t}

(3)

This is the exponential the lecture put against micro's quadratic.

It is exponential, and it is also a lie after about thirty seconds, because a base can only build one worker at a time. Write that cap as $P$ workers a minute. Growth is really

\frac{dW}{dt} = \min\!\left(\frac{m}{c}\,W,\; P\right)

(4)

Exponential until the production line saturates, linear after.

The exponential phase ends at $W^{*} = Pc/m$ . With one base producing about three workers a minute, a worker costing 50 and mining about 40 a minute, that is under four workers. In other words: on one base you are production capped almost immediately, and the compounding you were promised is not available to you. The only way to buy back the exponential is another base, which adds production capacity and mineral patches at the same time. That is the whole economic argument for expanding, and it falls out of two lines.

The mining rate here is an assumption, not a measurement. Roughly forty minerals a minute per worker on a close patch is the figure the community has used for twenty years, and the Economy Lab exists to replace it with a measured one. Nothing above depends on the exact value: change $m$ and the shape of the argument does not move.

What the lecture said

The lesson 1 lecture spends its last twenty minutes on this and nothing else. The framework arrives at 1:06:00, the four categories are built on the whiteboard in red marker over the following ten minutes, and the closing test, what did I spend and what did I get, lands around 1:25:00.

Two details are worth having. The course outline itself was built live from a question put to the room at 18:00: what is the most basic thing in StarCraft. The answer the room converged on was not minerals. It was buildings, on the grounds that losing all of them loses you the game. Everything else in the syllabus was derived from there, which tells you what kind of course it was going to be.

And the lecturer was candid at 14:00 about what the material could and could not do. It was not going to raise anyone's rank by itself. It was going to make them able to say why they were building the pylon.

Still true today

Still true

The taxonomy holds, and it works beyond the game

This is the part of the course that has aged best, because it is not a claim about StarCraft. It is a claim about how to account for anything scarce, and the game is the worked example.

Two things have moved. The first is that the physical resources box is now partly buyable: a good mouse, a steady frame rate and a hotkey layout that suits your hand are real purchases of reaction speed. The second is that you do not have to take the mining figures on trust: the Economy Lab lets you run them. Neither moves the model.

One refinement worth making: attention is at least two resources, not one. The attention it takes to notice something is not the attention it takes to execute something, and players fail far more often at the first. Lesson 10 goes further into this, and T10 carries it.

Checkpoint

Which of these is a physical resource in the class's taxonomy?
You have 1,400 minerals banked at nine minutes. What do you actually have?
Why is an armour upgrade an assistant resource rather than a raw one?

Sources for this page

Lesson 1 lecture, on YouTube, channel VioleTAK The resource taxonomy runs from 01:06:00 to the end of the recording
David Sirlin's write up of week 1 He was in the room that evening and published his account the same day
The class website (archived copy)