Eyes, ears and hands: the four jobs a unit does

Newcomer Learner

The short version

• A unit has four jobs, and fighting is only the loudest one. It also sees, denies sight, and carries things.
• Damage in Brood War is a small multiplication table: three damage types against three unit sizes. Learn the table and half of the matchup knowledge you were memorising becomes obvious.
• What kills a unit is hits, not damage, and hits are whole numbers.
• So an upgrade is worth nothing until it removes a hit, and then it is worth a great deal at once. Those moments are called breakpoints.
• Their armour upgrade can cost you more hits than your damage upgrade gains you.

The situation

You know that zealots beat zerglings. You have known it since your first week. Then you get plus one carapace zerglings against your zealots and the fight goes worse than it used to, and you cannot say why, because you are still counting units.

The reason is a rounding error with teeth. A zealot was killing a zergling in three swings. It still is. Meanwhile the zergling was killing nothing, and now it is killing slightly more of nothing. That specific fight did not change much. Another one did, and if you only know which units beat which, you cannot tell those two cases apart.

The four jobs

Lesson 2 starts from a frame worth having: a unit is a sense organ and a limb before it is a weapon. It sees, it hears, it does work, and only then does it shoot.

Eyes. Everything you know about the map, you know because a unit is standing somewhere. That is why the cheapest unit in the game is often the most valuable one on the map, and why killing a scout is not a small win.

Ears. Denying sight is the same job, mirrored. An overlord parked on a ramp, a turret on a cliff, a zealot at a watchtower: none of these are fighting, and all of them are doing work.

Hands. Units build, repair, heal, carry and block. A worker holding a choke, a shuttle that never drops anything, a medic that never heals: all of them are converting your attention into something on the map.

And fighting, which is the job the interface shows you and the job most players budget for exclusively. The practical consequence of the other three is that a unit's value is rarely its combat value alone, and the units that look worst on paper are usually the ones doing the other three jobs.

Local imbalance

The name for this idea is David Sirlin's, from his write up of the class: local imbalance.

The game is balanced in the aggregate and wildly unbalanced in specific places. A siege tank outranges a lurker and a reaver, so in that particular pairing there is no counterplay available to the shorter ranged unit except to not be there. Terran has no melee ground unit at all, which means a whole category of interaction simply does not exist for one of the three races.

These are not bugs and they are not oversights. They are the texture of the game, and the reason matchups feel different from each other rather than like the same game with different sprites.

The way to use the idea: stop asking whether a unit is good, and start asking where it is good. Every unit in Brood War has a small region of the game where it is close to unbeatable and a large region where it is bad. Winning is mostly about arranging to fight in your region.

Gas is a tell

The same lesson makes a point that belongs in every scouting checklist, so it is worth repeating here rather than in T4 only.

Gas is a costly, visible commitment. A player who takes gas early has told you a great deal: about which half of the tech tree he is buying, about how many of the cheap unit he will not have in two minutes, and about when the expensive thing arrives. A player who does not take gas has told you the same amount in the other direction.

You cannot read the tech tree. You can read the assimilator, and the number of workers on it, and that is most of the same information for the price of looking.

The damage table

Here is the part you can memorise in a minute and use for the rest of your life.

Every weapon in Brood War has a damage type. Every unit has a size. The type against the size gives a multiplier, and that is the whole system.

Marines, zerglings, zealots, workers, vultures and hydralisks are on the small and medium end. Dragoons, tanks, goliaths, archons and ultralisks are large.

Now three sentences that used to be a hundred lines of matchup memorisation:

• A vulture does twenty concussive. Against a zealot that is twenty. Against a dragoon it is five. The vulture is not bad against dragoons, it is doing a quarter of its printed damage.
• A dragoon does twenty explosive. Against a zergling that is ten, so it takes four shots for a thirty five point body. Against a hydralisk it is fifteen.
• A siege tank in siege mode does seventy explosive. Against a marine that is thirty five, which is still a marine. The tank is not aimed at marines, it is aimed at everything large.

Hits, not damage

The second half of the idea is arithmetic and it is where the real leverage is.

A unit does not die when it has taken enough damage over time. It dies on a specific hit. So what matters is how many hits, and hits are whole numbers, and whole numbers do not move smoothly.

A zealot hits twice for eight, so sixteen a swing. A zergling has thirty five hit points. Sixteen, thirty two, forty eight: three swings.

Give the zealot plus one attack. Nine twice is eighteen. Eighteen, thirty six: two swings. One upgrade took a third off the time it takes a zealot to kill a zergling, and nothing about the numbers on the unit card suggests that it would.

Now give the zerglings plus one carapace instead. The zealot's eight becomes seven, so fourteen a swing. Fourteen, twenty eight, forty two: still three swings. That upgrade bought the Zerg nothing at all in this particular fight.

Those two facts live one line apart in the same table, and neither of them is visible if you are thinking in percentages.

The working

Attacks that are not meant to win

One more idea from that week, because it is the bridge to T11 and T5.

Not every attack is trying to kill something. An attack can be trying to make them react, to make them stop mining for twenty seconds, to make them build a defensive thing they did not want, or simply to see what is at home. Those attacks are measured on a different scale, and measuring them by kills is how players talk themselves out of making them.

If you have ever thought "I should not send that, it will just die", the question you skipped is what it would have made them do.

Go further in the course

Lesson 2 teaches this with worked problems: medics against sunken colonies, what a probe is good for, and how many tank shots kill a hydralisk. The efficiency matrix it sets up is worked through in the first twenty seven minutes of the lesson 3 lecture, up to ►27:00. T8 is about it.

Careful with one common answer. A sieged tank does 70 explosive damage, and explosive lands at three quarters against a medium unit, so one shot takes 52.5 off a hydralisk's 80 hit points. That is two shots at every upgrade level, not one. A zergling is exactly on the line: half of 70 is 35, and a zergling has 35 hit points, so one point of Zerg carapace turns a one shot kill into two.

Still true today