Labs

APM budget

A fixed number of clicks a minute, more jobs than clicks, and a list of what goes unattended.

The short version

  • Actions a minute is not a score. It is a budget, and a budget is only interesting when it does not cover everything.
  • Nobody actually runs out of clicks. They run out of attention, and the first thing that stops is always the thing with no deadline attached to it, which is making workers.
  • Micro is the most expensive thing on the list by a long way, and it is the one everybody buys first.
  • Buying work you will not have to redo is the cheapest APM there is: a rally point, a control group set once, a building placed where you will not have to walk round it.
  • Decide what you are not going to do. The alternative is finding out in the replay.

What your hands can actually pay for

Nobody runs out of actions. They run out of attention, and then they stop making workers. Untick a job to see what the rest of the minute buys instead.

120actions

You have

135actions

The jobs want

15actions

Short by

Everything below the line gets nothing.

1jobs

Unattended

At 120 actions a minute you are 15 short of doing everything, so Harassment gets whatever is left, which is not enough. Take something off the list on purpose, before the game takes it off for you.

JobWantsGetsWhat it is
12 12 One worker a hall every 17 seconds, plus the clicks to find the hall.
3 3 Enough to never be blocked, which is the cheapest thing on this list.
20 20 Selecting, queueing, and rallying.
5 5 Placement, worker pulls, and the walk back.
6 6 One unit kept alive and moving is not free.
24 24 Keeping groups where they should be, before anyone shoots.
45 45 Only while fighting, and it is the most expensive thing you will ever buy.
20 5 A drop or a runby is a second game played with the same hands.

The list is funded from the top down, so the order is your priorities. A row that gets less than it wants is a row you are doing badly.

Why the budget is the right way to think about it

An action is not a unit of skill, it is a unit of time. Every job in a game of Brood War has a rate: a hall wants a worker every seventeen seconds, supply wants a building before you are blocked, six gateways want six clicks whenever they finish.

Add those rates up and you get a number. Compare it to your own number. If the first is larger, something is not being done, and which thing it is was decided by you, in the moment, under pressure, without thinking about it. Doing that on purpose instead is most of what improving looks like.

Why raising your APM helps less than reordering it

Two players with the same hundred and twenty actions can be a league apart, because the jobs in this list do not pay the same. Workers compound: a click now is income for the rest of the game. Micro pays once, at the moment of the fight, and only if the fight happens.

So the return on an action is not flat, and the right question is never "how do I click faster". It is "which of these can I make cheaper". A control group is a discount on every future move order. A rally point is a discount on every unit that building ever makes. Both are bought once.

Attention is the scarcest resource

Lesson 1 puts attention at the top of the list of resources: minerals are the least interesting thing you spend, and the scarcest resource is the one you cannot expand. Lesson 10 and lesson 11 come back to it from the other side, as something you do to the opponent. Overloading them means making their list longer than their minute.

That is worth holding on to, because it changes what harassment is for. A drop that kills four workers is worth four workers. A drop that makes them watch a corner for the next three minutes is worth whatever they stopped doing in order to watch it.

What this lab assumes
AssumptionValueWhy, and how sure we are
Every cost on the list our estimate None of these is measured. They are what the jobs feel like they cost, written down so you can argue with them. Change them in app/models/learn/apm.rb if you have better numbers.
Priority order top to bottom The list funds itself down the order shown. Real players do not allocate cleanly; they interleave and drop things at random, which is worse than this model.
Micro scales with the fight share of the minute A minute with no fight in it costs nothing in micro, which is why the same player looks fine in the early game and falls apart at the first engagement.
No spells, no drops not on the list Casters and multi prong attacks are the two most expensive things in the game and neither is priced here.