Ch.9: RTS Game Design: Does High APM Make You Better?

0:00 Take one player. Hands flying, control groups cycling, 300 actions a minute, and they just lost. And here's the trap: they didn't lose to a bigger army. A worker line sat idle, reinforcements never got built, one deadline slipped, and the number kept climbing the whole time. So a giant number, and still a loss. That's the part I can't square. When we watch a pro's hands blur across the keyboard, we just assume we need faster hands. I lived that assumption, honestly. I used to grind hotkey drills to watch the readout climb, and I couldn't figure out why my games stalled.

0:37 So what are we really answering here? Does a higher number mean a better player, or is it just spam? That's the fight. The counter tracks activity; winning means handling the right decision in time. We'll separate the measures, find where hand speed earns its place, then build one better practice loop. So start with the number itself. Where does it even come from? Blizzard's original patch notes, linked in the description, give us the cleanest historical split. In their old StarCraft 2 readout, classic APM counted every click and command.

1:09 Then EPM filtered that down to actual commands. So the raw one includes all of us, nervously reclicking an army five times. The filtered one strips a lot of that out. But read the fine print: "effective" only means it registered as a command. It does not mean the command was good. Huh. So we already have three different things. How much I did, how much registered, and whether the move was right. Three layers, and no single readout sees all three. Oh, it's a step counter, basically. The raw one logs every step, the filtered one drops the shuffling around the kitchen, and neither one knows whether I walked 10,000 steps toward the fire exit or in circles.

1:52 Yeah, and that's the gap everything else hangs on: the counter you stare at can't tell activity from judgment. Now, before we bury the counter, here's the question I actually have. Forget the readout. When does moving fast genuinely matter? Sometimes fast hands plainly do win. For example, you're a StarCraft player whose army fight just started at the front. Production needs an order, a worker cycle needs attention, a scout just spotted something, and if production waits, your reinforcements arrive a unit short.

2:24 Four things, one you. Right, and I might know exactly what to do on all four. Sure, but you can understand every decision and still fail to execute in time. That's when speed matters: it closes deadlines you already know are open. So speed isn't the skill by itself. It's what lets the decisions I already have actually land in time. Exactly, which points at what actually matters. Not the number, but the path from your head to the game. From that head-to-game path, here's one concrete case: I need another villager in Age of Empires.

2:58 The slow way is to hunt for my Town Center on the map, click it, click the villager button, then go find my army again. That's a lot of travel for one unit. And you're repeating that round trip for the whole game? Yeah, and the game's own tutorial shows the compressed version. A key jumps straight to the Town Center, a key queues the villager, a key finds an idle one, and a control group snaps me back to the army. Same intent, but two map hunts and half a dozen clicks collapse into four key presses. That actually flips it: a strong player isn't adding actions there.

3:35 They're actively trying to do less. That's the reframe. The hotkey doesn't create attention out of nowhere. It removes selection and navigation steps, so more limited focus stays on the fight. The number might tick up as a side effect, but that's a by-product, not the target. And building on that, it isn't only hotkeys. StarCraft 2's control guide documents a whole set, and they all do the same job: protect your next decision. Such as? For example, control groups keep a useful selection ready. Saved camera spots remove the trip to a place you visit constantly.

4:09 A queued command lets a unit finish a known sequence while you look elsewhere. Okay, but a mouse-first RTS player would push back: "I never save camera spots, and I don't feel slow." Do those actually bite? Think about the last fight you reached late. What did the camera trip cost? A few seconds, maybe the whole first volley. You can know the fight is coming and still arrive after it's decided. That's the nightmare version. Each tool removes a different overhead: selecting, moving the camera, or waiting.

4:39 Ouch, yeah. And then there's autocast, where the game just fires the ability on its own. It does, and I raise it for one reason: the game already automates some actions and hands others to you. That line is a design choice, not a law of the genre. Huh. It's easy to treat autocast as part of the genre, but somebody deliberately chose which repeated actions stay manual. Exactly. And to be fair, not every repeated input is waste either. A player might be keeping rhythm, or just checking state. The trap is treating the number as the objective.

5:12 So the fix has to live somewhere the readout doesn't even look. Okay, that leads to what I actually want. What do I practice, then? There's a nice piece of evidence here, and this is where it gets fun. One study looked at nearly a million StarCraft 2 action sequences from over 3,000 players, and found something intuitive: practiced actions clump into repeatable sequences. That's a huge sample. Oh, wow. So the pattern has a name? Motor chunking. The individual clicks fuse into one move you can just...

5:45 do. Like typing your password. You stop thinking in letters. Yes, except the paper only shows the letters grouping into a practiced sequence. It doesn't prove the rest of your attention is suddenly free or that your rank rises. Okay, so what's the version that stays inside the evidence? Try this instead. First: open a replay, find one recurring miss. For example, a worker I keep forgetting, or an army I park somewhere and lose. Pick exactly one. Then define the shortest legal path that prevents it, and bind or group it cleanly. Then run it in a real game and check whether the miss rate dropped. Not whether the readout moved.

6:26 So, concretely: my replay shows production sitting idle every time I'm in a battle. I put those buildings on one key, and the new habit is battle begins, tap it, queue one round, eyes back on the front. Then the only question is whether the idle time shrank, not what my counter did. Right. And next game, that one miss is the only score you're keeping. Now here's my pushback, and I think it's the strongest one. If I already know the right strategic move, why should I lose because I can't click a mouse fast enough?

6:57 Why not just automate the routine stuff? Yeah, that objection deserves a straight answer, and the best one I know comes from inside the genre. Take a look at StarCraft 2's own history: its designers actually built what you're asking for. In an early prototype you could select nine Hatcheries at once, hit one key, and queue up 1,000 Zerglings. That sounds like exactly what I want. And they pulled it back, because playtests showed it was producing worse decisions, not better ones. The shipped game makes the Zerg player spend each production larva separately. Their lead designer drew the line plainly: walking a worker to minerals got automated because it's never an interesting choice, and automation exists to give players good decisions instead of just a lot of clicks. Hold on, though. A fast-handed friend would say, "Be honest. That's lower-speed players rationalizing a gap." One scrapped prototype still doesn't settle it. Maybe Blizzard just gave up on the idea.

7:56 Fair. Years later, Blizzard tested that boundary again in the Legacy of the Void beta: Spawn Larva went on autocast. Wait, a public beta, not just another internal prototype? Exactly. Two other repeated base-management actions, Chrono Boost and the MULE, were removed because Blizzard said these mechanics were hard to perform and hard for viewers to see. The released game restored all three as player-triggered actions. So they tested the same boundary again, years later, and the player-triggered version still survived?

8:27 Twice, on the record. That doesn't prove one universal rule, but it shows Blizzard revisited the same boundary and returned to manual triggers both times. So the test isn't manual versus automatic. It's whether the action carries a decision. That's the boundary. In Age of Empires, every action is mouse-playable, and the keyboard path is simply faster. I still won't tell you one burden is correct for every game; the two games choose different burdens. So "why not automate everything" has a real answer: automate the walking, keep the choosing.

8:59 And you can see which is which in exactly one place: your own replay. So let's settle the fight we started with. Skill or spam? Take a replay and ask three questions. 1: what useful obligation did I actually miss? Okay, that's the miss. 2: which command path made handling it too slow? And 3: did shortening that path fix the decision before its deadline? And notice not one of those asks what my number was. Right. The readout might climb as you improve, but it's the exhaust, not the steering wheel.

9:32 Capacity, not the objective. Right. So don't practice the number. Find the decision that keeps arriving too late, and shorten the path to it. That's the whole chapter. Thanks for listening to Learning Podcasts.