UC Berkeley Starcraft Class, Week 1
Thursday, January 29, 2009 at 9:45PM 
A screenshot of how StarCraft looked early in development.Tonight I attended the much-talked-about StarCraft class at UC Berkeley as an observer. (Insert StarCraft joke about Observers.)
The main lecturer is the young Alan Feng. Mr. Feng is a physics student who says he's been playing StarCraft "for 2.5 years, 6 months on the pro level." He also had help leading the class from a guy named Yosh (I forget his real name, but I call people by their chosen names anyway), and a third guy who I only remember as Mumbling Guy. I would call Feng by his gaming name too, but I forgot what it was because he only said it once.
Feng and Yosh are an interesting contrast. Feng is endearingly highfalutin while Yosh is an old-timer (StarCraft-wise) who tells the young-uns how it used to be. Feng began the class this way:
There are not more than five musical notes, yet the combinations of these five give rise to more melodies than can ever be heard.
There are not more than five primary colors, yet in combination they produce more hues than can ever been seen.
There are not more than five cardinal tastes, yet combinations of them yield more flavors than can ever be tasted.
In battle, there are not more than two methods of attack: the direct and the indirect; yet these two in combination give rise to an endless series of maneuvers.
--Sun Tzu
And he added:
In Starcraft, there are only three races, but more gameplay remaining than can be explored.
There was then a long stretch of administrative debris about notecards we were to turn in, about what percentage of the final grade the homework is worth, and other such banalities. Notably though, 40% of the final grade comes from the final project where students must attempt to make a new contribution to the StarCraft community in the form of an analysis of some part of the game. These final papers will be public and subject to peer review--no doubt incredibly merciless peer review, given the tone of most gaming communities.
Feng then gave us a short history lesson about the release of StarCraft. It was announced in 1995, though it didn't release until 1997. Feng showed us graphs and stats of how many people had computers back then, what power they were, how many had internet access, and so on. His point was that StarCraft had a dramatically larger chance for success in 1997 than it did in 1995, so their delay was fortuitous.
As an aside, I'll point out that this involved Microsoft Powerpoint slides. One student asked if the slides would be available and Feng said no, that the slides don't contain anything useful except pictures anyway. That's an interesting statement and he's right. I hope presenters will learn that Powerpoint slides are a generally terrible way of conveying information. Especially if they have terrible typography and blocky graphs as these did. (Apple Keynote can at least look nice.) But whatever, let's move on.
Yosh then gave us 20 or 25 minutes of reminiscing about the history of the best StarCraft players. Almost everyone he mentioned is Korean, of course. I felt I had something in common with Yosh as he told us he's been playing and following his game for 10 years now, competing in tournaments and trying to improve.
He explained how various players evolved or changed the game. Boxer's initial dominance gave hope for Terran players in the early days. In fact, when asked who in the room is a Terran player because of Boxer, several students raised their hands. (Nerdy joke: is Boxer overpowered in every game?) Apparently Boxer went to the army for 2 years, and although he didn't get to play as much there, he still did play and the army cadets created a special army StarCraft team, just so he could keep playing. When he returned to the game, he made up for his generally weaker game by becoming much more bold, and pulling off insane strategies that no one else would use, like a fake base in the middle of the map.
Yosh told us about the personalities of several players. One of them he said never smiles or frowns or makes any expression at all...except for the one in the picture he showed us. Another has bravado, another was extremely effeminate. Some were known for their micro-management skills, others for their creativity, others for their consistency. One top player is called "cheater Terran" because he always seems to have more units than you'd think he'd be able to at any given time. It seems that "every gaming community is a weird mirror image of every other gaming community."
After this walk down memory lane of Korean Starcraft champions, Yosh let Feng take over for the last leg of the lecture. Feng talked about the different kinds of resources in the game. There are raw resources, which he defines as those that the Starcraft game construct knows about. Minerals, gas, population limit, creep/pylon fields, energy (for casting psionic storm, etc). There are also physical resources, which he defines as things outside the game that exist in the physical world (perhaps a misnomer?). These are things like attention (arguably the most important one in StarCraft), APM: actions per minute (arguably the one that a supposed strategy game should NOT focus on at all), physical endurance, state of mind, knowledge of the game, analysis, etc. I asked him to add yomi to the list, the ability to read the opponent's mind. He did not know the term, but I had earlier given him my book, so I'm sure he will soon. Yes he said, ability to read the opponent is another resource to draw on that exists outside the game construct.
Then there are what Feng calls transformational resources. These are things you convert raw or physical resources into other resources. The most common one is simply your "army." You use your APM (clicking speed skills) along with minerals and gas and time, and you convert all that into units that compose your army. That army is capable of taking over territory or killing enemy units or defending a new expansions, etc.
Feng's point here is a good one. He's trying to get the students to think of the game as a big collection of resources and your decisions are about how to shift those resources around. It's easy to overlook how many resources are really involved in a decision, and if you overlook some, you aren't understanding the real implications of your decision. For example, if your population limit is 131/131, what do you do? As it stands, you cannot build more units. Should you build pylons? That means spending minerals and time. Should you attack with units you already have? That means spending units and possibly more of your attention resource. How long will it take the units to attack and trade with the enemy units? Did you scout enough to know what you'll be up against and what important thing you could attack?
Another example he gave was using raw resources to cover for a lack of physical resources. If you have very bad reaction time and you know this, then you are aware that in a surprise attack on your peons (resource gatherers), you might lose more than you really should. It might be worth it to spend minerals to build some cannons back there so that less depends on your slower reaction time. It's a tradeoff that might be worth it depending on your particular play skills.
The last example he gave was that of defending a choke point. If you control a choke point and put some cannons near it, but the enemy does not attack there, what have you spent and what have you gained? You spent time and minerals of course, but Feng was saying we shouldn't be so hasty in saying that we gained nothing. We did gain some resources here. If there is a pylon there, we increased our population limit. We also have vision to that part of the map. That means we have slightly better overall information about where the enemy is (or isn't, in this case). We prevented the enemy from scouting here, so the enemy has a slightly worse mental picture of the map. We control some territory that might not otherwise control (whatever is behind the choke point). So really there are a lot of resources to consider here, even in this very simple example where no one even attacked anyone.
And that was it for week one. A class about StarCraft at UC Berkeley.
--Sirlin
Reader Comments (238)
@INTangiBLZ: "Characters that are difficult to use, actually increase character variety in high level tournaments. Look at low tier characters that win in tournaments, usually they're hard to use low tier characters instead of easy to use low tier characters. *my initial post explains why*"
There's nothing wrong with characters that are hard to use. Heck, they can be top tier. Dhalsim is 'hard to use' but he's not hard to use because he randomly jumps when you meant to set up a fireball trap, he's hard to use because he has a massive decision tree available to him at any given game step, and the wrong decions result in getting punished. HOWEVER, when you make the right decision, you get rewarded.
However, what you seem to be preaching is an arbitrary physical dexterity barrier. If I guess that my opponent is going to use a fireball, and, I do a standing roundhouse with dhalsim, to stuff the fireball, would the game be made deeper if I have to suddenly answer trivia questions in order to preform the move I intended to do. "I'm sorry, White Horse is not the capitol of Canada, instead of standing roundhouse, you jumped forward!"
You are correct that some 'low-tier' characters show up and win tournaments in the hands of experts. And you are correct in saying that the reason these characters may be percieved low tier is because of the difficulty in playing them. But the difficulties have nothing to do with 'special move execution barriers' (at least not ideally). If the jab button preformed SPDs (but required you to buffer the move), Zangief would still have all the same difficulties he does: getting past a fireball trap, and actually getting into a position where an SPD is possible. Why, when you do everything correctly should you still play the 'spin the joystick' minigame. Sure the game is balanced currently around difficulty of execution but... doesn't that seem like lazy design. Why can't every character be like dhalsim.
Lets be lazy. We can make every move fei-longs development version of short flame kick (i.e. safe on block, invincible on start-up), so that there isn't a question: flame kick if you can, because it's always correct. Seems unfair right? Well it's not because we're going to give it a random timing window!1!11 What's more we made the input insasnely difficult (think Geese's super special, deadly rave). Heck, it's so hard to preform consistently, we can let it juggle an unlimited number of times. Now we removed that pesky decision making that is so often required, but still have a 'pro' consistently beat a non-pro. you also get big, splashy, crowd pleasing finishes when, at the last round of a tournament, some one managed to, with a sliver of life left, juggle the opponent 40 times in a row, doing a 100% damage with fei-longs short flame kick. Man, that would be spectacular and rare! I'd much rather watch that than watch a player, with a sliver of life, anticipate every move an opponent was going to make, react accordingly, and get rewarded for it.
steve: ok fair enough, but there's still something not right there. You and I both agree that a "real-time strategy game" is not turned based. We both agree that you are expected to operate at some fairly fast speed, or greater. I think we also agree that being forced to operate at a fairly fast speed creates interesting situations because of time pressure, and adrenaline rush, and catching opponents when they are too wrapped up in thinking about the wrong thing during that critical split-second. So far, we both are talking about the same kind of game and enjoying it. I think we would both happily classify such a game as "real-time strategy."
It does not logically follow that such a game continues to reward faster and faster and faster and faster clicking. In StarCraft, it does. In Puzzle Fighter, you reach a limit maximum useful APM more quickly (still not quickly enough, but that's a minor point). So we can have a real-time strategy game with or without this bound. The problem is, if we do not have a bound and more fast clicking = you are definitely at advantage, all the way up to 400APM, then it kind of stops being a "real-time strategy game." It starts being DDR/IIDX with some strategy.
I started playing Puzzle Fighter with the plan of "make the best decisions possible, try to make them faster as I get better." This was a very unsuccessful method. Then I switched to "play as fast as possible...and given than I am playing that fast, make the best decisions I can." That was a very successful strategy and I'm good at the game. I have to admit that it favors fast play over strategy though. A little too much emphasis is put on fast-play relative to strategy. If it shifted a tiny bit towards lowering the upper bound, it would STILL be a very fast game and have all the good strategy stuff it has. But if it instead went the opposite way--if it removed the ceiling of how fast you can play and simply said faster is better always--then it would slide pretty far away from even being a strategy game.
So in short, I think it is necessary to say StarCraft is "RTS-with-unbounded-APM" and it's not redundant. An actual "real-time strategy game" would only be about the things in the first paragraph of my comment, it would be fast and strategic but not willing to shift so far towards fast that it gets valued to the point that coaches scout for it.
Also, I think you're selling StarCraft short yet again by saying that it wouldn't have an interesting and evolving metagame if it had a max useful APM of 200 or something. Is it really so shallow that only really high APM evolves the metagame? I really doubt that. It seems like a pretty complicated game that allows players so much room for creativity, tactics, and strategy that it would evolve if played at 200APM, 100APM, or with a direct brain-link.
@ Sirlin
Sirlin wrote "It's not like super high APM is needed to make that [psi storming several tanks] difficult."
Huh? All I was trying to say about the psi storming is that players who can do that need to act and click very fast because its hard to do, and storming in such a way SHOULD be hard to do and not be dumbed down. But I guess this was a moot point because your theoretical SC HD Remix only makes your peons auto mine, which I think wouldn't make or break the game.
I refer to SF HD because you draw examples from it to prove that you made it a better game, and through the same examples, you can make SC a better game, but I argue that you can't. Talking about the fact that it is limited by frame rate (where SC is not) and by controlling only a single character (where SC is not) is valid because I'm just trying to show that they are not the same or even remotely similar games.
BtW, i don't know anything about Dune 2 and I didn't post about it at all. But for you to boast about wishing to make a better game out of Starcraft without even playing it very much might be a bit forward.
Sirlin wrote "... an intentionally bad interface."
Starcraft does NOT have an intentionally bad interface. The interface is indeed very simple and pretty much nothing is automated but that doesn't mean it's bad. It gives the player more options and it feels very raw, which is one of the feelings that makes Starcraft so great and Warcraft III so bad.
Sirlin wrote "I don't even know what we are talking about anymore."
Let me explain what we're talking about. Basically, you claim that APM is too highly rewarded in pro SC. I claim it is not because players with lower APM (like Savior) can still be the best players out there.
Automining might not be a super bad idea and I never argued against it in this thread. All I'm saying is that too much automation and simplifying tasks in Starcraft might not be a good idea. I'm not sure what your SC HD remix entails anyway. If I'm not mistaken, this is the first time you wrote about what might go into it.
I'm basically arguing that Blizzard has balanced SC in such a way that the things that require your attention, whether it be "boring" base management or strategic army movement, requires just the right amount of attention. There doesn't need to be much tweaking here and there... the interface doesn't need to get worse or better to make it a better or more strategic game.
On a side note, have you ever played X-Com UFO Defense? It's a great game where you can basically do anything with relative ease, no APM required at all. But it's designed this way because it is a turn based game.
Steve wrote: "Calling a RTS a "RTS-game-with-unbound-maximum-APM" is as redundant as calling a FPS a "FPS-game-were-you-control-a-character-from-a-first-person-view-and-shoot-stuff.""
That's a great point, Steve. Sirlin basically wants to limit the maximum useful APM in Starcraft so that everyone can compete. I personally like the idea of Starcraft being without bounds. Remember that simply spamming fast clicks doesn't make you better, but that fast clicks with meaning behind them make you better.
There are tons of players out there that play at like 300 APM but suck a whole lot more than people who have a clearer mind and play at 200 APM.
You're still talking about frame rate in an incorrect way. This has little to do with anything so it hurts to sidetrack, but oh well. This also isn't a value judgment about what is better or worse, just a statement of how thing are. Street Fighter takes place at 1/60th of a second. It does this very reliably and was designed for a no-lag environment. It has a lot more precise timing than StarCraft (NOT a value judgment!) with combos that require 1-frame links, reversal attacks that require 1-frame timing, and the difference between your dragon punch winning or the other guy's winning coming down to whether you did yours 1-frame early or not.
The game measures 3,600 APM. Stop saying that it's "limited by frame rate compared to StarCraft." That's just a nonsense statement. It's worth pointing out that even though the game registers your 3,600APM, the game design itself doesn't give you more and more and more rewards as you approach 3,600. Just standing there doing nothing is sometimes smart. Sitting a block for a second can be smart. While you are locked into a move, you'll be committed to that move for 20 frames or something sometimes, so there's no need to do anything else. It happens to have very good properties of being a real-time game where you must make fast decisions, and having a huuuuuge possible APM, but actually a pretty low maximum useful APM.
I talked about frame rate in 1 out of like 140 lines of text. Odd that you'd make a response post only talking about frame rate...
"Is it really so shallow that only really high APM evolves the metagame?"
No, but the metagame would evolve differently without the APM and I have no reason to believe it would be better that way, and my hunch is that it wouldn't.
"if we do not have a bound and more fast clicking = you are definitely at advantage, all the way up to 400APM, then it kind of stops being a "real-time strategy game." It starts being DDR/IIDX with some strategy."
looks like another redundant way to describe the RTS genre as we know it. I agree however, that you are technically right in saying that a game can be both "real-time" and have "strategy" with an upper bound APM, and I suppose you would have to lump such a game in with the "RTS" genre, and it could even be a really great game, but that just isn't what I am looking for in a starcraft game.
Would you mind terribly if this thread became about HOW one would implement an RTS game with either an APM cap or diminishing returns on increasingly higher APM?
"Would you mind terribly if this thread became about HOW one would implement an RTS game with either an APM cap or diminishing returns on increasingly higher APM?"
Steve, I'm glad you asked. The answer is very simple, because it's a non-answer. To come up with something new, I would first state a premise, then come up with ideas to make that premise more tangible. I have already stated my premise many times here. (Does it need restating? I don't even know.) As is usual on the internet, we did not get to the part about how to achieve the premise because people show up who say the premise is wrong, or that no possible game could exist with that premise. That is why creating new things in a public discussion on the internet is usually bad.
The only way we could answer that question is to stop debating the premise, ignore people who disagree with it, and come up with constructive ideas about how to make a real-time strategy game that's fast but not so fast as to diminish the strategy. I submit auto-mining as the first of a large number of changes that would be needed.
Note to Lore: when I say "why would we make an intentionally bad interface?" what I'm talking about is, let's say we are making a new game that is to be "better" than StarCraft. The UI designer says "I'm putting in this feature where you can select multiple bases and build peons simultaneously." If we tell the UI designer "no, we want you to exclude that obviously user-friendly feature" then we are making an intentionally bad interface. My premise is that we can make the best interface possible, something Google or Apple might make, and come out the other side with a better game, not a worse one. I think steve and I would rather simply ignore you if you say it can't be done, because really, who cares what you think can't be done? If you think it SHOULDN'T be done, then ok fine you could say that. But even that can end up being a cloaked form of "can't be done." Remember that the premise is there would be no less strategy with a good interface, so there is nothing worthwhile lost.
Now that we're ignoring the haters, the rest of you lurkers are free to post ideas that make the premise come true.
I'm curious how you could design Starcraft with a low bounded useful APM. Starcraft is designed as a fast-paced game with lots of units and areas of importance to be attacked and guarded, and units are extremely responsive. OK, suppose you implement auto-mine and MBS. Then people will spend extra actions to focus fire individual units of the massive enemy army one at a time in three different areas on the map, and APM becomes a tax again. The only ways to reduce this is to 1) move away from the vision of Starcraft by reducing the game speed, reducing the number of units, or making units less responsive, or 2) implement some stupid anti-focus fire mechanic that would likely cheapen even lower-level micro.
If it's impossible to put an upper bound on useful APM, then it becomes a matter of where the APM is required. Do we want a game with tradeoffs between micro and macro, or do we want to hold a certain level of macro constant and only have people work on micro?
"I submit auto-mining as the first of a large number of changes that would be needed."
interesting, but there is no effect on the boundless APM.
The ideas of Automining, multiple building select, and the sort are age old ideas that Blizzard has had countless discussions internally and with people concerned about starcraft 2. Of course it can be done, I never said it can't be done. But there must be a good reason why Blizzard is even considering having it not be done.
Once again, I never said automining is a bad thing that shouldn't be implemented. I actually think it might be a nice feature that won't impact the game too much.
There's a pretty damn good reason google and apple don't make games and Blizzard does.
Something worthwhile that could be lost if you have all that automation implemented into SC, then a macro-esque player who comes out with a huge army would not be as impressive as before, and people who used to be able to macro in such a way would fall in line with the rest of the crowd because now EVERYONE can macro that way. I'm not saying that leveling the playing field in this way is a bad idea, i'm just saying its a valid concern and it could definitely detract from what makes Starcraft a great, quite possibly the best RTS game ever made.
" But there must be a good reason why Blizzard is even considering having it not be done."
Isn't that reason because their fanbase hates change and they'd rather make Starcraft 1 with better graphics, than avoid alienating the core with a better, but different game?
Also, I hereby submit that Mario Party is the deepest strategic game known to man. They took a boring turn based game and threw on arbitrary dexterity tasks. You can ACTUALLY row a boat by preforming not 1, but 100s of 360 motions.
Also, Starcraft has an event loop (i.e. Frames). Every game does. I think a lot of people get tripped up because street fighters old school animation has far fewer frames than the game it's self.
Lore, it seems safe to ignore you at this point. Nice jackass comment about Google and Apple. If you are really want to make sure we have bad UI, yes we should steer clear of those companies. Seriously though, ignoring them is just saying "I don't want to learn about UI."
Also it's not a valid argument to say that if Blizzard is considering making a bad decisions, that is some kind of proof of anything. Blizzard makes lots of good things and bad things. They made an obviously flawed PvP honor system in world of warcraft. It's entirely possible they arent' interesting in making a strategy game at all with StarCraft series, I have no idea what they want. Maybe they want to make a real-time click game with 5% strategy, who knows, and who cares when it comes to the question of how to do this right. I suspect that their actual goes are these: 1) appeal to current starcraft players even if doing so has vastly bad effects like rewarding 400APM and 2) given point 1, make the game as deep as possible.
You also lost me at "but then a huge army wouldn't be impressive!" Again, you disagree with the premise. If building a huge army is hard because normally the opponent would smash you for trying, then yes it will still be impressive. Or if everyone builds large armies, then no it won't be. This is neither here nor there.
Steve: automining doesn't reduce maximum APM, correct, but it does help the other side of the spectrum which is a real problem that also needs solving. That's the APM tax to even get to the real game. So we need auto-mine to help with that.
Perhaps to think about maximum useful APM in an outside-the-box way, think about why you might want to do nothing. As I keep saying, in Street Fighter it's sometimes perfectly fine to do nothing. Remember that Street Fighter is an action game that is generally fast, and the game we are aiming for is an action game that is generally fast. If there is a reason to occasionally do nothing in a game, that's probably a good sign that it isn't about more, more, more clicks. Also, I've heard that the game Company of Heroes already answers this entire question by having a much lower maximum useful APM, but I never played it and I don't know how they did it.
yeah the problem isn't that there is too much do do, that's a good thing because the point having a trade-off, when you do one thing you don't do another thing. The problem, from the perspective of desiring a limited APM game (which i still maintain starcraft should never be) is that if you are fast you get to do both. I think the best solution here is the obvious one: a cap on APM. it literally counts the actions you made in the last minute and if that number hits the limit, you can't make another action until you fall back below the limit. That way every action is literally a decision not to do something else, and you can't circumvent the trade-off by going faster.
Hmm...
I'm not sure where the comment about google and apple came off as a jack ass comment. It's true that Blizzard has made plenty of bad things but I think it's safe to say that the current Starcraft UI isn't a bad creation. Starcraft was and probably still is the best RTS ever made and Blizzard realizes this, so their consideration to keep the original UI is, to me (and to some others), a valuable consideration.
Anyways, I seem to have frustrated Sirlin to the point where he wants to ignore me. I hope some others reading this forum will continue to find some logic in my arguments.
"automining doesn't reduce maximum APM, correct, but it does help the other side of the spectrum which is a real problem that also needs solving. That's the APM tax to even get to the real game. So we need auto-mine to help with that." just because you don't go back to your base as often doesn't mean the tax before you see the real game is lower. If your opponent can harass, defend your harass, and fight a battle all at the same and you can only choose one or two of the above, you are at a severe disadvantage.
Isn't having the highest APM ALWAYS going to be advantageous though? It seems inevitable given the very nature of a RTS... It happens in real time and speed is a factor, and the faster you are, the bigger your advantage, with no cap. That doesn't mean that fastest will always win, but fastest will always have an advantage... And if you removed the "mindless" clicks that add to the high APM wouldn't they just be replaced by "Mindful" clicks that would ultimately add to a high APM? Let's say theoretically that, if, presently in SC, 100 of the 300 APM are "mindless" and removed from the game completely... Now you have a baseline of 200 APM with no "mindless" clicks... As the game evolved and players got better and better and faster and faster wouldn't their APM go up anyway, as they figured out new things to split their attention with? In essence, wouldn't you simply be robbing mindless-Peter to pay Mindful-Paul? And what you are criticizing, that uncapped speed/APM is always rewarded, wouldn't that remain exactly the same?
rfresh: that's exactly what I am saying. the guy with faster fingers will always win in a traditional RTS (not a value judgement). there are very few mindless clicks. take for instance, the act of building supply depots. where you put it is sometimes strategic, 90% of the time you put it exactly where you would have expected to put it a minute ago and an advantage goes to whoever is better at going back to their base in the middle of a battle, telling an SCV to make a supply depot, and then going back to the battle.
if we want to make a game where an advantage wasn't conferred to the guy who can hit f2,*select scv*,b,s*click where to build*,*hit hotkey of a unit in battle twice to bring your screen back to the battle*, we'd need strategic reasons why you might not always want to be performing so many actions.
we are talking about a game very different from starcraft.
Might I suggest, Sirlin, that to avoid the points you make against starting to develop an idea in a public forum on the internet, you perhaps write a new article/blogspot on the subject? Something to the effect of "I take beef with X Y Z etc mindless actions in SC, I would make them better this that and the other way"? I know I personally would be very interested in such an article, it would allow time away from the clear conflict that has arisen in this discussion and a proper place for further discussion (we're pretty off topic of the original post: the coolest college course ever) with a clear beginning point.
Maybe it would be on better accessibility in games(lower "Tax" to play) being fundamentally good for strategic competition (though I believe something to this effect is in another of your postings). Maybe it would be specifically about RTS's, maybe just SC, maybe on a clear explanation of the necessity of some kind of bounded MUAPM. Your call. I think any of the above would be good starting points for furthering this discussion, though, as we do seem to be going in circles a bit.
Only one quick note from me since I don't have time to write a long post.
Starcraft has these APM and execution barriers because they NEED to be there to balance the damn game.
This is not at all correct. Starcraft had APM barriers long before it was balanced. It was balanced around those APM barriers. There are multiple ways to balance things, and an APM barrier should not be the first thing a developer looks at when trying to balance a game.
In many ways, APM barriers just make the game harder to balance for a wider number of players. For example, consider a hypothetical Race A that has few APM barriers and is balanced around the (relatively good) stats of its units. Now consider another hypothetical race, Race B, that has worse unit stats overall, but that has some really powerful tricks (like stacked Mutalisks and such, but much harder to execute) available past a certain very high APM barrier. Below that critical value of APM, Race A is stronger than Race B due to better unit stats, but above it, Race B is either equal to Race A or more powerful than Race A.
So if you balance this game around the top level of play, where the APM level is very high, things might end up pretty even. But at anything below the topmost level, the game is simply unplayable due to imbalances. Starcraft only avoided this fate because players were lucky enough to discover engine bugs and other tricks that gave all the races powerful APM-based tactics, instead of just one or two.