Donnerstag, 10. Februar 2011

Really long and good explaination of how the macro changed (+balancing)


This was originally going to be a part of the “Oh Micro, Where Art Thou?”-thread, but as the aforementioned thread grew too big I decided to chop the material into two separate threads. As time passed and with the game fast approaching launch, I got discouraged with the non-existent prospect of anything I was going to propose ever being considered or implemented. Thus I totally gave up on writing anything.

The thread was initially going to address macro mechanics and the need for a possible revision of their early game role. When I picked the idea up again in November, it would eventually develop into a monster 23 page word document detailing how macro mechanics were speeding up the game and deflating the worth of scouting information. That document is still sitting on my desktop, but it will likely never see the light of day. When trying out the ideas in practice, I realized I was largely wrong in where I was placing the blame and at the same time sort of stumbled upon a different approach to the subject. Macro mechanics in the end turned out to be one small part of a broader combination of factors pushing SC2’s gameplay in a certain direction.

Whether the issues about to be presented are even real or merely figments of my imagination I’ll leave to the 3,1k master leaguers and true scholars of TL to decide in the engaging and civilized discourse that will be sure to follow this post.

The article will make a bold, probably all too outrageous, attempt at explaining how SC2’s economy impacts and affects gameplay through the merging of SCIENCE™ and ESPORTS™.

Chapter I: Macro Analysis

The inspiration for the experiments conducted in this post come from a thread dating back to 2008. It didn’t receive very much attention back when it was posted but was nonetheless very informative: CDRdude's Worker Saturation thread. The experiment measured the mining speed of X workers over 5 minutes and graphed the results. What I did was basically just replicate his experiment in an attempt to make a comparison between Broodwar and SC2.

+ Show Spoiler [Method detail] +

[image loading]
Measuring how many minerals were mined over 5 minutes per number of workers.

Before conducting these experiments, I had always gotten the feeling that SC2 had a surplus of minerals in its early game compared to Broodwar. This experiment, while providing some interesting results, didn’t really confirm those suspicions. What’s most interesting about SC2’s graph, is its purely linear growth from 9 to 16 workers, and how little effect there is from making more than 22 workers on one base. The mining speed is pretty much constant from ~24 workers and upwards. The graph is linear at start, moves onto an exponential decline for a short while, and finally becomes constant.

This is of course to be expected as worker AI has become smarter in SC2 in combination with Blizzard lowering the time that each worker spends mining at each mineral. They simply don’t disturb each other very much when mining anymore and they wait more patiently in line for their turn to mine – minimizing the wandering phenomenon that would inevitably occur in Broodwar.

To be fair, we should point out that purely linear growth would have been found in Broodwar as well if only the experiment would have measured the mining rate of the 4th worker to the 8th worker. In SC2 the same thing can instead be observed from worker nr 6 to worker 16. The truly interesting result is discussing what effect reaching max saturation at around 3x workers per patch, instead of 4x workers per patch as in broodwar, has on gameplay.

As we all know, TL scholars like to engage in highly metaphysical debates surrounding the true nature of cheeses and all-ins in SC2 – and whether they even exist at all. Perhaps the graph can shed some light on this hotly debated subject. It seems like we can make the claim that build orders in SC2 will reach their final and most developed one base state quicker than in Broodwar. We can probably also say that after ~22 workers mining minerals in an 1base vs 1base situation, there is no differentiating between a cheese and a “normal” build until an expansion is up and operational. Does this imply that expanding is more dangerous in SC2 as opposed to Broodwar?

I don’t really know, that might be stretching it a bit too far; though there is certainly less of an effect of expanding before you are beginning to supersaturate your first base. Also: supersaturating your first base against someone who cuts worker production will provide you with no other real benefit than having workers to maynard. Using this logic one could claim that expanding is in fact more dangerous. If the races reach their fully saturated states quicker in SC2 as opposed to Broodwar, and if a cut in worker production after a certain point doesn’t reflect on your income at all, then a continued worker production will only really mean you are cutting your army size by the amount you invest in workers and in an expansion.

The data can probably be interpreted in a variety of ways. But as I’m the author of this thread I get to showcase mine: Due to the lower max saturation cap SC2 builds will tend to conform into one standard or one mould much quicker than Broodwar builds. They will also tend to be less punished when cutting workers in favor of “cheesing”. Merely defending a cheese won’t win games, but rather getting the superior unit composition and securing the expansion without dying will win you the game. Of course this interpretation is somewhat exaggerated and SC2 is a lot more dynamic than I will have it sound, but I still think it is somewhat evident that Broodwar builds develop and evolve forth in more distinct stages where scouting information has a chance to play a bigger role in the game. A Broodwar build will simply take longer to reach its final and most developed one base state (which pretty much will look identical to and support as many production buildings as their SC2 counterparts), and go through more intermediary stages before getting there. On top of these facts, there is a slight mineral surplus in the cheesy stages of a game in SC2 compared to Broodwar.

Alright. Enough chattering, I’m starting to tread on dangerous grounds here. Let’s get on with this thread. After the initial test, I had a feeling that distance mining from your expansion might prove to be cost effective after reaching a certain point of saturation. So that’s what I examined next.

[image loading]
Tests were conducted on the 12 o’clock position of Lost Temple. Initial travel distance was included in the measured time period (i.e. I started measuring as the probes were sent out from the main base, not when they started mining

Woah! Surprising results. Perhaps these results will have some application in PvP and in certain cases maybe even in PvT. Hell, why discriminate against terrans? Might have an application to any race that’s forced to stay on one base and is supersaturated on said base. After 22 workers mining minerals there is actually a gain from sending your workers distance mining. At least if the expansion is at a similar distance to that of LT’s 12 o’clock position.

Moving on to the effectiveness of each worker.

[image loading]

I don’t think there were any real surprises here. The results could pretty much be deduced from looking at the previous graphs.

I had yet to take into account the effects of macro mechanics on income though, so naturally that’s the direction I headed next. First was the MULE, which was a fairly easy macro mechanic to measure and graph. The only reservation I want to add, is that the graph might be a bit misrepresenting because building an Orbital Command puts you 2 workers behind. To simulate that, I would have to change the x-axis from number of workers to a time axis, which initially presented some problems to me but which I nonetheless attempted in the graph following this one.

[image loading]
Although the graph is technically correct, imagine it lowered about 400 minerals in the mid ranges.

At first, upon seeing this, I thought my sentiments and suspicions of SC2 being a game overflowing with minerals in its early and midgame confirmed. The starcraft community have often expressed their need and want for better scouting capabilities in SC2. They have also long thought there to exist magical fixes such as simply increasing overlord base speed.

What if the explanation of the deflation in the worth of scouting information simply lay in the fact that SC2 strategies were made so extreme in their strength and timings through the respective races all having bursts of mineral surplus at varying stages of one and the same match? I believe this to be an important observation, because much of the recent balancing of the game appears to have gotten stuck in fighting mineral surplus disguised as imbalance.

I followed Blizzard’s balance panel from Blizzcon with great interest, and I want to start this paragraph off by clarifying that I think they’re doing as good a job as they could possibly do in balancing the game without making fundamental game changing alterations (which would make absolutely no sense to implement in normal patches). They seem to be more aware than I ever thought they’d be (compliment, not diss) of the issues plaguing the game. This thread is not meant to school them on balance, but rather provide an alternative interpretation to some of the issues they’ve expressed concern about – through the perspective of a fresh pair of eyes.

One of the things discussed in the balance panel was the community’s whine about “stim being overpowered”. The people at Blizzard expressed some concerns as well, but were sensibly reluctant towards meddling with something as fundamental as stim. They thought there might be some unnamed and difficult to define combination of factors accounting for the problem. Concerns were also expressed about matchups possibly being undynamic, with races rolling each other over at specific timings and different varying stages of one and the same matchup.

I believe that analyzing the economic system of Starcraft 2 might provide a better explanation model to these phenomena than would searching for the answer in unit and build/research time tweaks. Protoss’ most weak timing in PvT is undoubtedly in the early mid game, when trying to expand while dealing with Terran’s stim timing. Coincidentally this is also the exact timing in the game when Terran experience a mineral surplus surge compared to a fully saturated and capped Protoss player. Meddling with unit balance due to perceived imbalances because of mineral surplus fluctuations in the earlier stages of a game might have unwanted effects in the potency of certain unit combinations in the later stages of the same matchup.

This is starting to get long-winded; I’d better hit you with the next graph or I’ll likely lose your interest. In the next graph I tried to change number of workers to a time axis. In SC2, building 5 workers takes about 61,2 seconds. In Broodwar it takes only slightly longer. For the sake of practicality, I’ve rounded them both down to 60 seconds. The x-axis now depicts number of minutes elapsed after starting off with 9 workers. There might be a more elegant way to simulate the effect of the chrono boost, but this is what I in the end was able to come up with.

[image loading]
Probably needs some explanation: I tried to take into account that the build time of the Orbital Command is 34 in game seconds, while the build time of an SCV is 17. I simulated 3 chrono boosts for Protoss, the first at 10 workers, second at 14, third at 16 workers.

We would probably like to believe Blizzard are stupid and never foresaw any of these issues, but more likely is that they did in fact foresee some problems. By adding a second gas geyser to each base, they were able to delay the advent of max saturation, evening out mining speed and partly curbing the extreme effects of macro mechanics in the early game.

Since the last graph was in no way indicative of a real game, my next project was trying to simulate how the graph would look like in a real game, with people using real build orders. I watched white-ra play a game on scrap station against some terran, and took note of all the chrono boost and gas timings. In that particular game, white-ra was super greedy and used up 5 quick chrono boosts on his nexus. His terran opponent opted for a 2 refinery tech build, so that’s what I simulated: a special case not necessarily indicative of all games.

[image loading]
Chrono boosts at 10, 14, 16, 20 and 26 workers. First gas finishes at 15 workers for Protoss. Second gas finishes at 23 workers. Those are the “dips” in the graph – workers being pulled off mining minerals. For Terran, I assumed they have 1 worker off of mining at all times for the construction of buildings. Orbital was started at 15 workers, after which refinery would soon finish. Second refinery finished at 20 workers. Possible scouting timings weren’t taken into account for.

I don’t know if it would be presumptuously assigning too much meaning to the graph in claiming that it helps explain some of the ebbs and flows of the PvT matchup. So I won’t even try.

I tried to include zerg into this, but since they’re on 2 bases and a highly irrational race to boot, I was left with nothing but a headache. Sorry.

Chapter II: The bottleneck and the ceiling

Does an inherent 2 base bottleneck exist in SC2? When is the optimal timing to take a third base? Are nexus/cc first builds viable?

[image loading]
Data was extrapolated and calculated from previous existing values.

The above graph is probably unnecessary and overkill. It’s just a stretched out copy of the first graph posted in the thread. In a way though it’s still important to post – if only to illustrate how the effectiveness of additional bases will scale with the amount of workers you’ve got.

If nothing else, the graph helps show why nexus first or CC first were such powerful builds in BW while a slightly more dubious opening in SC2.

The really interesting implications though, are when you start brainstorming about how the graph would look like on 3, 4, 5 and 6 bases and compare that to how many workers and bases you could possibly support with a 200 supply cap in each of the respective games.

It is safe to say that players in SC2 are not as likely to venture out into taking third bases before starting to supersaturate their two already existing bases. If you assume 12 workers harvesting gas on 2 bases, and linear growth up until 32 workers mining minerals on 2 bases, you’d have to reach 44 workers total before even experiencing any positive effects whatsoever of taking a third base. Assuming you go up to 22 workers mining minerals at each base, you’d be up at 56 workers before taking a third base – and the additional gain from spreading workers out evenly on those 3 bases would be a mere ~1100 minerals over 5 minutes.

In this light, it seems slightly foolish of the community to expect the metagame of SC2 to eventually evolve into something resembling that of Broodwar. You will likely never see players opt for as early of a third base as used to be the norm in Broodwar. Rather players will tend to be bottlenecked on 2 bases for longer (especially on Blizzard-sized maps).

The 3 base ceiling

How many bases can you really support in SC2? If you assume 3bases with 16 workers mining minerals at each of them, and with 18 workers assigned to harvesting gas you’re already up to 66 workers total. The common consensus seems to be that the “optimal” number of workers is somewhere around 70-75.

Let’s do a comparison between Broodwar and SC2:

Assume we have 54 workers mining minerals equally divided on 6 bases. How many minerals will those workers mine in 5 minutes? And how many minerals would those workers mine if instead confined to 3 bases?

Protoss, BW, with 54 workers equally distributed on 6 bases: 18120 minerals over 5 minutes.
Terran, BW, with 54 workers confined to 3 bases: 13200 minerals over 5 minutes.

Zerg, SC2, with 54 workers equally distributed on 4, 5 or 6 bases: ~15384 minerals over 5 minutes.
Protoss, SC2, with 54 workers confined to 3 bases: 14586 minerals over 5 minutes.

I cannot help but find a major contradiction in Blizzard’s conceptual outline of how the zerg race is supposed to be played in SC2 with what the game’s economical system actually allows for. Zerg are supposed to keep outexpanding, outmacroing and outproducing their opponents.

Based on these data, the only way to secure a macro lead in SC2 seems to be by rushing to 3 fully saturated bases as quickly as humanly possible. The entire objective for zerg in SC2 seems to have been reduced to recklessly rushing to a macro lead as quickly, stupidly and foolishly as possible before the game caps the chance for any macro lead to develop.

Will larger maps save Starcraft 2?

[image loading]

This is an interesting question to pose with the new and giantly oversized maps GSL have introduced. I believe these large maps are an anti reaction to the volatile and unpredictable play that plagued “Blizzard-sized maps”. The unmanageable strategic extremes (due to unnamed factors that may or may not have been attempted to be explained in this article) on small and medium sized maps simply created the need for a party to step in and introduce a buffer zone for rushes and timing attacks.

With that said: what will larger maps achieve apart from increasing rush distances?

I would say absolutely nothing. What need do players have for 14 expansions in a game like Starcraft 2? Absolutely none. Zerg’s play will be centered around saturating 3 bases as quickly as possible and launching suicide attacks at the opponents’ thirds. Protoss’ play will be centered around camping and delaying until they’ve reached their invincible end game composition on 3+ bases. Terran’s play will… no idea.

Large maps will simply and frankly favor the race that currently has the pleasure of being dominant when maxed out in a 3base vs. 3base late game situation. That race, as you’ll see, will be Protoss. And please don’t mistake this for whine; it’s merely stating what should be obvious. On the other end, the same maps will likely disfavor the previous most stable performing tournament race on blizzard-sized maps: Terran.

[image loading]

The colossi and immortal are units that absolutely and critically need to be
as strong as they are for Protoss to survive terran stim timings and zerg onslaughts
in the mid game. But in the lategame, used in combination with templars/voidrays/phoenixes,
they become a headache for Blizzard to balance.

The game is balanced around small maps – not large maps. Units are primarily balanced to withstand the effects of mineral surplus on said maps, not to remain balanced throughout all stages of a game. Blizzard’s first priority is to prevent shit from dying instantly to other shit. Second comes worrying about whether these changes prove to provide dynamic mid- and lategames (and it’s here-in that the real challenge lies).

300 supply cap?

If I’ve learned anything from observing Blizzard the past year, it’s that it is largely pointless to suggest anything that would require alterations to the game engine itself. Whether it be about moving shot or built-in delay for firing projectiles (tank AI). If it can’t be achieved through the map editor, it likely won’t be “fixed” in the way you imagined. Thus I’m not even going to attempt to discuss changing worker AI.

If the future of SC2 is to be played out on GSL-sized maps, one proposition would be increasing the supply cap of the game so you can support ~110 workers and about 5 bases. One of the greatest proponents of an increased supply has long been day[9] himself. My main argument for an increased cap is that the strategies in the game likely will become streamlined and predictable very quickly if kept back by a 3 base ceiling. The main counterargument? It wouldn’t be balanced at all in the game’s current state, and would likely require a lot of rebalancing.

I think Blizzard have to make a decision soon about whether they want to balance the game for GSL-sized maps or for their own tiny sized maps.

Chapter III for this thread was supposed to go more in depth about specific strategies and rant about a conceptual flaw in the design of the zerg race, but I decided including it would likely detract from the whole of the article. I’m already steering far off topic as it is. Plus, merely listing a bunch of problems as opposed to sticking one’s neck out and proposing solutions makes this thread look that more impressive and impervious to the critique of TL scholars.

Thanks to Pholon for helping me host the pictures. I hope my 1000th post was an enjoyable read to you all. The economic system and the macro mechanisc of Starcraft 2 are in sense its fourth race. As much detail and attention should be spent on understanding and balancing their effects on the game as goes into balancing the races.

Perhaps I’ll post the rant after Assembly.


  1. Well my head is now spinning I guess I need to construct additional pylons to absorb more information. thanks for the info though!

  2. Jesus Christ, you must be a high level diamond player. I really hate the bigger maps like Kulas Ravine.