Micro: win the fight you are already in

Lesson

3 of 14

Watch

The lecture, 46 minutes, or a 42 minute cut

Practise

Homework 3, a checkpoint och a plan for your next game

This lesson makes you better at fights. You will learn to price a fight before you take it, to decide what your army shoots first, and to control your units by one rule that covers every situation: keep the most weapons firing for the longest time.

What you will learn

• How to count what your army can kill, unit by unit, and see whether a fight is won before you take it.
• Why the choice of target matters as much as the size of your army, with a worked example you can check.
• How to reason about an army you have not scouted: what he can afford, and what the clock rules out.
• The one sentence definition of micro, and five rules that follow from it, each one a thing to practise.
• What to watch for in three famous pieces of professional micro.

Price a fight before you take it

Make a small table. Give each of your unit types a row and each of his a column. In each cell, write how many of his unit one of yours kills before it dies. That table is the efficiency matrix, $E$. The cell $E_{ij}$ reads "one of my unit $i$ kills this many of his unit $j$".

You can fill some cells straight from the unit stats. A zealot swings twice for 8 each, so it deals 16 to a zergling, and a zergling has 35 hit points. Three swings kill it. A zergling hits a zealot for 5, which comes off the 60 shields first and then off the hit points at 4 a hit past the zealot's 1 armour. That is 37 zergling hits to kill one zealot. Nobody fights one on one in a real game, so the cell is a judgement, not a formula, but the stats tell you which way it leans.

Price it in what you are short of

The same fight gives different answers depending on the currency you count in. These are illustrations, not measurements:

Use the currency you are about to run out of. Late in a game with 6,000 gas banked and 700 minerals left, you do not have a gas problem. Price your army in minerals, and the trades that matter become obvious.

Your army, and where you point it

The first line is what your army could do. The second is what it does, once you decide where it shoots. That decision, the appropriation vector $\mathbf{a}$, is the part most players never think about. The matrix belongs to the game. The vector is yours.

A worked example: ten zealots and an archon

You have ten zealots and one archon. He has 25 zerglings and eight mutalisks. Say a zealot kills three zerglings and no mutalisks, since it cannot hit air, and an archon kills eight zerglings or six mutalisks.

Now the choice. If you let the fight split itself the way his army is built, 25 parts zergling to 8 parts mutalisk out of 33, this is what happens:

Careful with the arithmetic when you do this yourself: the denominator is 25 plus 8, which is 33. It is an easy slip, and a wrong denominator can flip a close verdict.

The better split is not a fraction at all. It is per unit. Zealots cannot hit air, so every zealot goes at the zerglings: ten zealots kill 30, and there are 25. The archon goes at the mutalisks: it kills six, and there are eight. You are two mutalisks short, and no split fixes that. The fight is lost, and knowing it before you walk up the ramp is the difference between losing the army and sitting under your cannons for another ninety seconds.

Add three cannons and the answer turns over

Give each photon cannon six zerglings or two mutalisks. With three cannons, the mutalisk column reads 6 plus 6, which is 12, against eight. Now point the cannons and the archon at the air and the zealots at the ground. Zealots kill 30 of 25 zerglings. Archon and cannons kill 12 of eight mutalisks. You win both halves.

Notice what did not win: splitting fire in his ratio. That still loses, even with cannons, because it wastes zealot attention on a column zealots cannot touch. The target you choose is a decision, and it is often the whole fight.

What the table cannot tell you

Three limits, and each one is a lesson in itself.

1. Real targeting is messier than a share per type. Which units shoot which targets, from where, at what moment, is far more than one number per column. The table gives you the direction, not the choreography.
2. The numbers move. An upgrade, a ramp or a bad arrival changes every cell. A table measured at one upgrade level is wrong at the next.
3. Support units break it. A medic kills nothing, so its row is all zeroes. Yet a marine alone barely trades with a zergling, and a marine with a medic behind it kills three or four. A table that adds units up one at a time has nowhere to put that.

The fix is neat. Keep the ordinary table, and add a small correction that is empty almost everywhere:

The medic's cell becomes a function of how many marines stand near it. The science vessel's cell becomes a function of time, because irradiate builds up damage rather than dealing it at once. The practical lesson: when you count an army, count its support units as multipliers on what stands next to them, never as bodies.

Fighting an army you have not scouted

When you have not seen his army, it is not a list of units. It is a set of possibilities, and you can narrow it a lot without seeing anything.

Three rules you can use tonight:

• He cannot have more than he could afford. Resources mined so far, divided by the unit's cost. At six minutes a zerg has mined a known amount. He cannot also have ten mutalisks and three bases.
• The clock rules some things out completely. He cannot pop an ultralisk at six minutes. That is not unlikely, it is impossible. Cross it off and stop worrying about it.
• Scouting narrows it, and time widens it again. Your scout sees a spire and the question is settled. Every second after that, he banks more, and the possibilities spread out again.

That third rule is why one scout is never enough. See the whole method in lesson 7.

Micro, in one sentence

Ask ten players what micro is and most will say "controlling your units by hand". That describes what you are doing. It does not tell you what to do. Here is the definition this course uses:

It has two halves. Fire more weapons, so more damage lands. Keep them alive longer, so they keep firing. Picking off weak units fits inside it, because an enemy unit that dies stops firing at you. So does stutter stepping, spreading against splash, and pulling a hurt unit back.

Five rules that follow from it

Practise these one at a time. Each one is a single habit.

1. Keep your units alive. A unit on 10 hit points that walks back still fires in the next fight. A dead one never fires again.
2. Retreat, so there is a next battle. A losing fight you leave early costs you a few units. A losing fight you stay in costs you the army.
3. Regenerate more than he does. Shields, zerg regeneration, medics and repair all buy time. Stretch fights out when you heal faster, and end them fast when he does.
4. Maximise your splash and minimise his. Clump his units, spread yours. Against storm, lurkers or sieged tanks, spreading is the whole fight.
5. Dodge what you can. Walk out of a storm, run from a spider mine, never stand still under a reaver's scarab.

Three pieces of pro micro, and what to watch for

Each of these clips shows the five rules working. Watch each one twice: once for the story, once for one rule.

Garimto's zealots ►36:00 (the clip). One gateway against three. Watch for: he retreats every time the local trade turns bad, and when one of your zealots runs, he puts two or three of his on it so it takes several hits on the way out. Damage on a leaving unit is damage he does not have to take back. Six zealots against six ends with four of his alive.

Kingdom's dragoons ►39:00 (the clip). His nexus is gone, so he has no production at all, against dark templar, zealots and cannons. Watch for three things:

• He sends probes in first so the units that deal damage take damage last. That is rule 1.
• Shields regenerate, so he stretches the fight to buy health. That is rule 3.
• But every second he waits, his opponent is mining. Regeneration costs him the other player's income. That is lesson 1 in one decision: name what you spent and what you got.

He finishes on about 4 supply against 40, and his opponent leaves the game.

Marines against lurkers ►44:00 (the clip). Mobility against immobility. Lurkers must burrow to attack, so the marines go wherever the lurkers are not. Watch for: "running away" here sometimes means running toward his base. The word is about damage taken, not direction.

Common mistakes

• Splitting fire the way his army is built. It feels fair and it is usually the worst split there is. Point each unit at what it kills best.
• Counting medics, vessels or templar as army. They are multipliers. Count what they multiply.
• Staying in a lost fight to "get something back". Every second in a lost fight feeds him kills. Leave, and fight again where your flux is better (see lesson 4).
• Fearing everything you have not scouted. Cap it by what he could afford and cross off what the clock forbids. What is left is short.
• Microing one unit while the rest stand still. Rule one is more weapons firing. Keep the army shooting first, then save the hurt unit.

Watch the lecture

A 46 minute lecture on the efficiency matrix, micro and three pro clips. The short cut skips four minutes of class business.

The minutes worth your time: the worked example at ►9:00 , what the table cannot do at ►16:00 , the definition of micro at ►27:00 , and the clips from ►36:00 .

Practise

Homework 3 has six problems on exactly this lesson: damage breakpoints, medic positioning, leapfrogging, how healing changes a fight, and why mutalisks want a one volley kill. Two worked answers to get you started:

Breakpoints. A sieged tank hits for 70 explosive. Another sieged tank is large, with 1 armour and 150 hit points, so 69 lands and it takes three shots. Two attack upgrades add 5 each, the hit becomes 79, and two shots deal 158. The kill drops from three shots to two. That is the shape of every breakpoint: upgrades do nothing, nothing, then suddenly remove a whole hit. Marines against zerglings show it too. A marine deals 6, a zergling has 35 hit points, six shots. With +1 attack a marine deals 7, and five shots kill. One upgrade makes every marine a sixth faster at killing zerglings.

The one volley kill. A mutalisk's first bounce deals 9. A marine has 40 hit points, so five mutalisks kill one marine per volley with 5 damage to spare. Ten mutalisks kill two. Eleven kill two and waste the eleventh. Group your mutalisks in multiples that finish targets, because a marine left on 4 hit points gets healed and shoots back while you come round again.

Medics and the death rate. If damage arrives at rate $\beta$ and medics heal at rate $\alpha$, a unit with $H$ hit points dies at $\beta/H$ without medics and $(\beta-\alpha)/H$ with them. Against storm or lurkers, $\beta$ is huge and medics barely matter. Against light fire, $\alpha$ gets close to $\beta$ and your marines stop dying at all. That is why medics look useless in one fight and unbeatable in the next.

Do homework 3

Then test the numbers yourself:

• Damage Lab: pick any attacker and target and see shots to kill at every upgrade level. Find three breakpoints.
• Battle Lab: run ten zealots and an archon against 25 zerglings and eight mutalisks, then add cannons.
• What micro actually is: the longer guide to the five rules.

Checkpoint