How to scout, and keep them guessing

Lesson

8 of 14

Watch

A written lesson, no video

Practise

A checkpoint and a plan for your next game

Lesson 7 taught you why you scout. This lesson is the how, on both sides: where to look, what each look tells you, when a trip is worth its cost, and how to keep your opponent guessing for almost nothing.

What you will learn

• How to put a price on a scouting trip, and when it is worth paying.
• Which facts stay true for a minute and which stay true all game.
• Six ways to scout, including ones that never go near their base.
• How to deny their scout, and why denial is the cheapest edge in the game.
• How deep to cut into economy for safety, using what you saw.

Price the trip first

A scouting trip is not free and it is not expensive. Put a number on it so you stop arguing with yourself.

A worker mines about 42 minerals a minute. Sent out early and kept alive, it is off the minerals for about a minute and a half: roughly 60 minerals of mining. If it dies, add the 50 it cost. Call the worst case 110 minerals.

Now the other side. What does being wrong cost you?

The break even, in one line

Send the scout when

where $p$ is the chance you are wrong about something you would act on. Put the numbers in. Against a 450 mineral misbuy and a 110 mineral trip, the scout pays if there is better than a one in four chance you guessed wrong.

You guess wrong more than one time in four. Send the scout.

The same line tells you when to stop. A second look straight after the first tells you almost nothing, because $p$ just dropped to near zero. Look again when enough time has passed for them to have built something new. That is when $p$ has climbed back up.

Not every fact lasts as long

Two facts you learn at the same second can be worth very different amounts a minute later.

The rule: spend your scouting on the slow facts. Where their army stands is false in ten seconds, so it is only worth looking at if you are about to fight. Whether they took gas, and when, still tells you something five minutes later. One worker peeking at a geyser early is worth more than three minutes of an observer following an army around.

Scouting pays best against either or choices. A building for air units has no middle setting. One look at it settles the question: air is coming, so you need detection and anti air now. A slightly bigger army, by contrast, tells you almost nothing you did not already assume.

Six ways to scout

You do not always need a unit in their base. Most of these cost less than a worker trip.

1. Scout by elimination. On a map with several start positions, an empty one tells you where they are not. Their scouting worker's direction and timing tell you where it came from.
2. Scout by intention. An army standing between two of your bases has already told you it is about to attack. Where their army stands is a plan, not just a position.
3. Scout with a sentry. Park a cheap unit, a burrowed zergling or a unit on a hill, on an expansion you want to know about. Patrolling an invisible unit through their side of the map does the same job without your attention.
4. Scout by denial. The more you contain them, the less you have to look. If they cannot leave their base, you know what they are not doing.
5. Scout by timing. A building that takes a long time to finish tells you what they gave up to have it now. Early tech means a smaller army or fewer workers. Go and hit the side they cut.
6. Scout the gas. How much gas they have mined, and from when, limits every tech unit they can have for the rest of the game.

One trick worth testing in a practice game: build your own refinery, assimilator or extractor on their geyser before they take it. It costs them time to kill it, and it tells you exactly when they wanted gas.

Keep them guessing

Everything above works in reverse, and denying a scout is cheaper than making one.

The last row is the one nobody does. If a tech building is the fact that lasts longest, the cheapest anti scouting move in the game is building it where their worker does not walk past.

Every bit of scouting you deny raises their $p$. It costs you almost nothing. So against a player who scouts well, denial is how you get back to even.

Use what you saw: how deep to cut

You would always rather spend everything on workers and bases. The army you build is the cut you take so you do not die. Scouting tells you how deep that cut needs to be.

After every look, ask three questions in order:

1. Can I win a fight right now? Then attack, and keep producing behind it.
2. Am I losing if they attack right now? Then build units, and build them now.
3. Neither? Then spend on economy. This is where games are won.

Cutting deeper than you need is how players beat themselves. Two cannons you did not need at four minutes are 300 minerals your opponent has in workers, and those workers are still mining at twelve minutes.

Common mistakes

• Scouting once and never again. Your first look expires. Plan a second look for when they could have built something new.
• Watching the army instead of the buildings. Army position lasts seconds. A tech building lasts all game.
• Letting their worker wander your base. Chase it off, or at least block the corner where your tech is.
• Building for the worst case every game. Defence you did not need is a gift to your opponent. Scout, then cut only as deep as you have to.
• Seeing something and not acting. A scout is worth nothing unless it changes what you build next.

Practise

Then do these in your own games:

1. Price your own scout. Time how long your scouting worker is away in one game. Multiply by 42 minerals a minute, and add 50 if it died.
2. Find the slow facts. In one replay, list every look you took in the first six minutes and mark each one fast or slow. Count how many of your scouting seconds went to fast facts.
3. Deny one thing. Play a game where your only anti scouting move is one cheap unit on the path into your base. Write down whether their scout came, when, and what it saw.
4. Hide the tech. Put your first tech building in the least visible corner of your base, and check in the replay whether their scout ever found it.

Checkpoint