Deception: make them pay for what they think they saw

Lesson

11 of 14

Watch

The lecture, 79 minutes, or a 60 minute cut

Practise

A checkpoint și a plan for your next game

A fake that costs nothing fools nobody. Deception works because it is expensive: you really do make yourself weaker for a while, and a smart opponent believes you because nobody would pay that price for a lie. After this lesson you will be able to price a bluff before you make it, tell when one is worth it, and read the ones played against you.

What you will learn

• Why every working deception is a purchase, and how to write its price down.
• The two kinds: pulling armies to the wrong place, and making them buy the wrong answer.
• Signalling: why a costly signal is believed and a free one is ignored.
• Probing: send a small force in, read what comes back, and punish the wrong answer.
• How to plan a game against one specific opponent before it starts.

Every deception is a purchase

There are two kinds, and they cover almost everything.

• Army deception. Pull their defending army to where it is not needed, or pull their attack to where it can do no damage.
• Build deception. Make them buy the wrong answer to a question you are not asking.

Both work for the same reason. Nobody puts themselves in a weaker position on purpose, so when you do, your opponent reads the weakness as real. They are reasoning correctly. That correct reasoning is what you are buying.

So price it. Every fake has two columns: what it cost you, and what it made them do. Here are seven from pro games. The lecture chip on each opens the video at that play.

Read the right column. Very few of these buy a win. Most buy time, clicks, or a wrong purchase. That is normal. A good fake does not win the game. It makes the rest of the game cheaper.

What to take from each

• Lurkers on hold. Strong players stop watching a fight once they think they know how it ends. The moment your opponent switches screens to macro is free, and their own skill created it.
• The empty dropship. The cheapest bluff is a unit that already has a job. A dropship that is only seen, never unloaded, still moves their army.
• Two barracks, no gas. Show a threat that needs an expensive answer and a cheap question. 150 minerals made them spend about 500.
• Tech on three paths. When you cannot hide what you are doing, keep enough options open that no single answer works. You pay in speed.
• The hidden factory. Give them an innocent explanation for what they can see, so they stop looking for what they cannot.
• Mines on the recall spot. Spend actions early, when you have spare ones, to save actions later, when you will not.
• The lifted base. The limit of the idea. If the fake costs more than any answer they could give, it is not a trick, it is a loss.

Signalling: why a costly fake is believed

A signal between opponents only works under three conditions.

1. It has to mean something to them. A move they cannot read is not a signal, it is exercise.
2. Both of you play to win. True in a 1v1. Not always true in a team game or a free for all, where someone may play for revenge.
3. It has to cost you something. If lying were free you would do it all the time, and they would learn to ignore it.

Condition three is why every fake in the table has a price. In nearly every case the price was delayed tech: the extra barracks, stim researched for five marines, the dropship with nothing in it.

Biology has the cleanest example. A gazelle that sees a lion sometimes jumps straight up on the spot, wasting energy. The jump means: I can afford to waste energy, so do not bother chasing me. The lion believes it because it is costly, and both animals save a chase. Your extra barracks is the same jump.

Probing: ask a question, grade the answer

Here is a technique you can use tonight. Send a small force into their base (eight speed zealots, four hydralisks in an overlord, a pair of vultures) and watch what comes back. Their response is the information.

Pro examples: in a Protoss mirror, one zealot answered by one zealot instead of two lets the attacker walk into the mineral line and win soon after. Savior dropped four hydralisks on Stork again and again because the right answer never came. And a shuttle can fly a dark templar toward a Terran without dropping it, just to see whether the tanks can unsiege and move in time.

The cost is real. The probing force can die for nothing, and you have to watch the reply while doing damage somewhere else. Keep the probe small enough that losing it is cheap.

Planning a game against one opponent

The lecture ends with a model of pre game planning, using Bisu against Savior on Blue Storm. Almost all of it happens before the game starts, and you can do the same before a series.

• What they have done before. Their favourite build, and whether they keep using it after everyone has seen it.
• Reputation and format. A well known player will want to show their signature build. In a short series, beating it in game one is worth more than one game.
• The map. Where the chokes are, how wide the entrances are, where defence goes. Cannons placed forward stop zerglings cold but can be outranged by hydralisks. Cannons placed back let the buildings soak the first hits. So denying their scout forces them to prepare for the wrong one.
• The counter chain. Corsairs beat overlords. Hydralisks beat corsairs. Dark templar beat hydralisks. Follow the chain and the whole game can come down to one timing.

Common mistakes

• Faking for free. If the fake costs nothing, it tells them nothing. They will ignore it.
• Faking what you cannot afford. A bluff that loses to their correct answer is just a bad build.
• Repeating it. A bluff works once per opponent. After that it is information for them.
• Not watching the answer. A probe is wasted if you do not look at what came back.
• Over-answering their probes. When they poke you, send exactly enough. Sending everything is the second wrong answer.

Watch the lecture

The full lecture, or the short cut that keeps the pro games and the theory.

Practise

• Price one bluff. Before a game, choose one fake you can afford. Write what it will cost. After the game, write what it bought.
• Probe three times. In one game, send a small force in three times. Write down each answer and whether it was nothing, everything, or just enough.
• Watch for the screen switch. In a replay of your own, find the moment your opponent stopped watching a fight. What could you have done with it?
• Run the numbers. Use the Battle Lab to see how many units your opponent really needed to answer a probe, and compare it with what they sent.
• Plan a series. Pick an opponent you play often. Write their favourite build, the map, and one thing you will fake to make them prepare wrong.

See also the guide Deception is bought, and lesson 7, which is the other half: scouting reads their plan, deception makes them read yours wrong.

Checkpoint